diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,93016 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.22837441617185544,
+  "global_step": 15500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.058823529411763e-07,
+      "loss": 3.7871,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4117647058823527e-06,
+      "loss": 3.8352,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.117647058823529e-06,
+      "loss": 3.865,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8235294117647054e-06,
+      "loss": 3.6066,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.529411764705882e-06,
+      "loss": 4.0823,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.235294117647058e-06,
+      "loss": 3.6868,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.941176470588235e-06,
+      "loss": 2.8153,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.647058823529411e-06,
+      "loss": 4.0699,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.352941176470587e-06,
+      "loss": 3.5638,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.058823529411764e-06,
+      "loss": 3.0112,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.764705882352941e-06,
+      "loss": 4.061,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.470588235294117e-06,
+      "loss": 3.6791,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.176470588235294e-06,
+      "loss": 2.8934,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.88235294117647e-06,
+      "loss": 3.964,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0588235294117646e-05,
+      "loss": 3.7387,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1294117647058822e-05,
+      "loss": 2.6371,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 3.62,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2705882352941174e-05,
+      "loss": 3.4492,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3411764705882353e-05,
+      "loss": 3.6621,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4117647058823528e-05,
+      "loss": 3.2213,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4823529411764704e-05,
+      "loss": 3.297,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5529411764705882e-05,
+      "loss": 2.793,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6235294117647056e-05,
+      "loss": 3.1317,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6941176470588233e-05,
+      "loss": 2.8041,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.764705882352941e-05,
+      "loss": 2.6695,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8352941176470587e-05,
+      "loss": 1.602,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.905882352941176e-05,
+      "loss": 3.1175,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.976470588235294e-05,
+      "loss": 3.0261,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0470588235294115e-05,
+      "loss": 2.8206,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1176470588235292e-05,
+      "loss": 2.3562,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1882352941176466e-05,
+      "loss": 2.9143,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2588235294117643e-05,
+      "loss": 3.1259,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3294117647058824e-05,
+      "loss": 3.5127,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 2.9489,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4705882352941174e-05,
+      "loss": 2.8986,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5411764705882348e-05,
+      "loss": 3.1892,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6117647058823525e-05,
+      "loss": 2.9296,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6823529411764706e-05,
+      "loss": 2.9748,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.752941176470588e-05,
+      "loss": 2.5736,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8235294117647056e-05,
+      "loss": 2.5858,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.894117647058823e-05,
+      "loss": 2.8461,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9647058823529407e-05,
+      "loss": 3.4808,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0352941176470588e-05,
+      "loss": 3.1436,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1058823529411765e-05,
+      "loss": 2.9985,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.176470588235294e-05,
+      "loss": 3.3454,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.247058823529411e-05,
+      "loss": 2.9192,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.317647058823529e-05,
+      "loss": 2.6256,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3882352941176466e-05,
+      "loss": 2.7848,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.458823529411765e-05,
+      "loss": 2.9693,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.529411764705882e-05,
+      "loss": 1.9026,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 2.6622,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6705882352941175e-05,
+      "loss": 3.0343,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.741176470588235e-05,
+      "loss": 2.8375,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.811764705882352e-05,
+      "loss": 2.9581,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.882352941176471e-05,
+      "loss": 1.6797,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.952941176470588e-05,
+      "loss": 2.6291,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.023529411764706e-05,
+      "loss": 3.076,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.094117647058823e-05,
+      "loss": 2.8834,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.164705882352941e-05,
+      "loss": 2.0854,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2352941176470585e-05,
+      "loss": 1.866,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.305882352941176e-05,
+      "loss": 2.4675,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.376470588235293e-05,
+      "loss": 3.1492,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.447058823529411e-05,
+      "loss": 2.9545,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5176470588235286e-05,
+      "loss": 2.8734,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.588235294117647e-05,
+      "loss": 3.1196,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.658823529411765e-05,
+      "loss": 2.9104,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.729411764705882e-05,
+      "loss": 2.9532,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 2.5793,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.8705882352941175e-05,
+      "loss": 2.6191,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.941176470588235e-05,
+      "loss": 2.8935,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.011764705882352e-05,
+      "loss": 3.0042,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.0823529411764696e-05,
+      "loss": 2.7202,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.1529411764705876e-05,
+      "loss": 2.5353,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.223529411764705e-05,
+      "loss": 3.1932,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.294117647058824e-05,
+      "loss": 3.4131,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.364705882352941e-05,
+      "loss": 3.1395,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.4352941176470585e-05,
+      "loss": 2.2197,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.505882352941176e-05,
+      "loss": 2.5104,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.576470588235294e-05,
+      "loss": 2.6758,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.647058823529411e-05,
+      "loss": 2.9786,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.7176470588235286e-05,
+      "loss": 2.4571,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.788235294117646e-05,
+      "loss": 3.0308,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.858823529411764e-05,
+      "loss": 3.0852,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.9294117647058814e-05,
+      "loss": 2.596,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 2.95,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.0705882352941175e-05,
+      "loss": 2.4203,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.141176470588236e-05,
+      "loss": 2.5522,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.211764705882353e-05,
+      "loss": 3.0179,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.28235294117647e-05,
+      "loss": 2.6305,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.352941176470588e-05,
+      "loss": 2.4001,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.423529411764705e-05,
+      "loss": 2.665,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.494117647058822e-05,
+      "loss": 2.5192,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.56470588235294e-05,
+      "loss": 2.6758,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.635294117647059e-05,
+      "loss": 2.2262,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.705882352941176e-05,
+      "loss": 2.6959,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.776470588235293e-05,
+      "loss": 2.4712,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.847058823529412e-05,
+      "loss": 2.3038,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.91764705882353e-05,
+      "loss": 1.1007,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.988235294117647e-05,
+      "loss": 2.6221,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.058823529411764e-05,
+      "loss": 2.9487,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.129411764705881e-05,
+      "loss": 3.0938,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 1.4368,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.270588235294116e-05,
+      "loss": 2.5154,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.341176470588235e-05,
+      "loss": 2.7627,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.411764705882352e-05,
+      "loss": 2.7191,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.48235294117647e-05,
+      "loss": 3.0349,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.552941176470588e-05,
+      "loss": 3.0185,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.623529411764704e-05,
+      "loss": 2.7609,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.694117647058823e-05,
+      "loss": 2.5645,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.764705882352942e-05,
+      "loss": 2.683,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.764705882352942e-05,
+      "loss": 2.7233,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.835294117647058e-05,
+      "loss": 2.7719,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.905882352941177e-05,
+      "loss": 2.7533,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.976470588235293e-05,
+      "loss": 2.6326,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.047058823529411e-05,
+      "loss": 2.6293,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.117647058823529e-05,
+      "loss": 2.9133,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.188235294117646e-05,
+      "loss": 2.9437,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.258823529411763e-05,
+      "loss": 2.3026,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.329411764705882e-05,
+      "loss": 2.8023,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.4e-05,
+      "loss": 2.826,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.470588235294117e-05,
+      "loss": 2.9862,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.541176470588236e-05,
+      "loss": 2.3115,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.611764705882352e-05,
+      "loss": 2.8432,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.68235294117647e-05,
+      "loss": 2.9921,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.752941176470586e-05,
+      "loss": 2.9988,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.823529411764705e-05,
+      "loss": 3.0744,
+      "step": 126
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.894117647058822e-05,
+      "loss": 3.2687,
+      "step": 127
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.96470588235294e-05,
+      "loss": 2.5784,
+      "step": 128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.035294117647057e-05,
+      "loss": 2.7226,
+      "step": 129
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.105882352941176e-05,
+      "loss": 2.9783,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.176470588235295e-05,
+      "loss": 2.3864,
+      "step": 131
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.247058823529411e-05,
+      "loss": 2.9768,
+      "step": 132
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.31764705882353e-05,
+      "loss": 3.1438,
+      "step": 133
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.388235294117645e-05,
+      "loss": 2.5777,
+      "step": 134
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.458823529411764e-05,
+      "loss": 2.8765,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.529411764705882e-05,
+      "loss": 3.1712,
+      "step": 136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 2.7391,
+      "step": 137
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.670588235294116e-05,
+      "loss": 3.0247,
+      "step": 138
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.741176470588235e-05,
+      "loss": 2.7379,
+      "step": 139
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.811764705882352e-05,
+      "loss": 3.0205,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.88235294117647e-05,
+      "loss": 2.5504,
+      "step": 141
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.952941176470588e-05,
+      "loss": 2.6743,
+      "step": 142
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010023529411764704,
+      "loss": 2.7139,
+      "step": 143
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010023529411764704,
+      "loss": 2.8212,
+      "step": 144
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010094117647058823,
+      "loss": 2.9276,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010164705882352939,
+      "loss": 2.7264,
+      "step": 146
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010235294117647058,
+      "loss": 2.7908,
+      "step": 147
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010305882352941175,
+      "loss": 2.7132,
+      "step": 148
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010376470588235293,
+      "loss": 2.6468,
+      "step": 149
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001044705882352941,
+      "loss": 2.9542,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010517647058823529,
+      "loss": 2.574,
+      "step": 151
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010588235294117647,
+      "loss": 2.3304,
+      "step": 152
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010658823529411764,
+      "loss": 2.996,
+      "step": 153
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010729411764705882,
+      "loss": 2.3344,
+      "step": 154
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 2.6705,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010870588235294117,
+      "loss": 2.4769,
+      "step": 156
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010941176470588234,
+      "loss": 2.8841,
+      "step": 157
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011011764705882352,
+      "loss": 2.7279,
+      "step": 158
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011082352941176469,
+      "loss": 2.7129,
+      "step": 159
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011152941176470588,
+      "loss": 3.2001,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011223529411764705,
+      "loss": 1.2236,
+      "step": 161
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011294117647058823,
+      "loss": 2.8242,
+      "step": 162
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011364705882352941,
+      "loss": 2.8914,
+      "step": 163
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011435294117647057,
+      "loss": 3.1769,
+      "step": 164
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011435294117647057,
+      "loss": 2.8815,
+      "step": 165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011505882352941176,
+      "loss": 2.7103,
+      "step": 166
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011576470588235292,
+      "loss": 2.7477,
+      "step": 167
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011647058823529411,
+      "loss": 2.6239,
+      "step": 168
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011717647058823528,
+      "loss": 3.0246,
+      "step": 169
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011788235294117645,
+      "loss": 2.9505,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011858823529411763,
+      "loss": 2.8729,
+      "step": 171
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011929411764705882,
+      "loss": 2.7456,
+      "step": 172
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 2.6617,
+      "step": 173
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012070588235294116,
+      "loss": 3.0272,
+      "step": 174
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012141176470588235,
+      "loss": 2.7718,
+      "step": 175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012211764705882352,
+      "loss": 3.1359,
+      "step": 176
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001228235294117647,
+      "loss": 2.9468,
+      "step": 177
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012352941176470587,
+      "loss": 2.585,
+      "step": 178
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012423529411764706,
+      "loss": 2.5852,
+      "step": 179
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012494117647058822,
+      "loss": 2.8753,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001256470588235294,
+      "loss": 2.5209,
+      "step": 181
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001263529411764706,
+      "loss": 2.5307,
+      "step": 182
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012705882352941175,
+      "loss": 3.1038,
+      "step": 183
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012776470588235294,
+      "loss": 2.835,
+      "step": 184
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001284705882352941,
+      "loss": 2.611,
+      "step": 185
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001291764705882353,
+      "loss": 3.0948,
+      "step": 186
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012988235294117645,
+      "loss": 3.4929,
+      "step": 187
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013058823529411764,
+      "loss": 2.8255,
+      "step": 188
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001312941176470588,
+      "loss": 2.3364,
+      "step": 189
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 3.1101,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013270588235294117,
+      "loss": 2.7083,
+      "step": 191
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013341176470588233,
+      "loss": 3.3664,
+      "step": 192
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013411764705882352,
+      "loss": 3.12,
+      "step": 193
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001348235294117647,
+      "loss": 3.0959,
+      "step": 194
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013552941176470587,
+      "loss": 2.2597,
+      "step": 195
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013623529411764705,
+      "loss": 2.3383,
+      "step": 196
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013694117647058824,
+      "loss": 2.5271,
+      "step": 197
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001376470588235294,
+      "loss": 2.5915,
+      "step": 198
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001383529411764706,
+      "loss": 2.585,
+      "step": 199
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013905882352941175,
+      "loss": 2.1843,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013976470588235293,
+      "loss": 3.0202,
+      "step": 201
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014047058823529412,
+      "loss": 2.5887,
+      "step": 202
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014117647058823528,
+      "loss": 2.8987,
+      "step": 203
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014188235294117647,
+      "loss": 2.6396,
+      "step": 204
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014258823529411763,
+      "loss": 2.6767,
+      "step": 205
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014329411764705882,
+      "loss": 2.9199,
+      "step": 206
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 2.7167,
+      "step": 207
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014470588235294116,
+      "loss": 2.2925,
+      "step": 208
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014541176470588232,
+      "loss": 2.677,
+      "step": 209
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001461176470588235,
+      "loss": 3.9602,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001468235294117647,
+      "loss": 2.4081,
+      "step": 211
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014752941176470586,
+      "loss": 2.7713,
+      "step": 212
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014823529411764705,
+      "loss": 2.804,
+      "step": 213
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001489411764705882,
+      "loss": 3.1023,
+      "step": 214
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001496470588235294,
+      "loss": 2.6641,
+      "step": 215
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015035294117647058,
+      "loss": 3.3424,
+      "step": 216
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015105882352941177,
+      "loss": 2.4107,
+      "step": 217
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015176470588235293,
+      "loss": 2.4357,
+      "step": 218
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001524705882352941,
+      "loss": 2.5211,
+      "step": 219
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001531764705882353,
+      "loss": 2.5344,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015388235294117646,
+      "loss": 2.2503,
+      "step": 221
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015458823529411762,
+      "loss": 3.1444,
+      "step": 222
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015529411764705884,
+      "loss": 2.3456,
+      "step": 223
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000156,
+      "loss": 2.8629,
+      "step": 224
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015670588235294116,
+      "loss": 2.707,
+      "step": 225
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015741176470588232,
+      "loss": 1.4099,
+      "step": 226
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015811764705882353,
+      "loss": 2.8881,
+      "step": 227
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001588235294117647,
+      "loss": 2.7886,
+      "step": 228
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015952941176470585,
+      "loss": 2.5947,
+      "step": 229
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016023529411764704,
+      "loss": 2.8305,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016094117647058823,
+      "loss": 2.2743,
+      "step": 231
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001616470588235294,
+      "loss": 2.8999,
+      "step": 232
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016235294117647057,
+      "loss": 2.3118,
+      "step": 233
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016305882352941176,
+      "loss": 2.7522,
+      "step": 234
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016376470588235292,
+      "loss": 3.1311,
+      "step": 235
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001644705882352941,
+      "loss": 3.2037,
+      "step": 236
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016517647058823527,
+      "loss": 2.9395,
+      "step": 237
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016588235294117646,
+      "loss": 2.7352,
+      "step": 238
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016658823529411764,
+      "loss": 3.3759,
+      "step": 239
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001672941176470588,
+      "loss": 2.5393,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000168,
+      "loss": 2.4005,
+      "step": 241
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016870588235294118,
+      "loss": 2.941,
+      "step": 242
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016941176470588234,
+      "loss": 3.0195,
+      "step": 243
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001701176470588235,
+      "loss": 2.7674,
+      "step": 244
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001708235294117647,
+      "loss": 2.5207,
+      "step": 245
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017152941176470587,
+      "loss": 2.7573,
+      "step": 246
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017223529411764703,
+      "loss": 2.6872,
+      "step": 247
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001729411764705882,
+      "loss": 3.1836,
+      "step": 248
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001736470588235294,
+      "loss": 2.7105,
+      "step": 249
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017435294117647057,
+      "loss": 3.1815,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017505882352941173,
+      "loss": 2.8527,
+      "step": 251
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017576470588235294,
+      "loss": 2.7035,
+      "step": 252
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001764705882352941,
+      "loss": 2.6675,
+      "step": 253
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017717647058823526,
+      "loss": 3.0045,
+      "step": 254
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017788235294117645,
+      "loss": 2.7752,
+      "step": 255
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017858823529411764,
+      "loss": 1.6096,
+      "step": 256
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001792941176470588,
+      "loss": 3.5266,
+      "step": 257
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 3.3263,
+      "step": 258
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018070588235294114,
+      "loss": 2.9468,
+      "step": 259
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018141176470588236,
+      "loss": 2.5885,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018211764705882352,
+      "loss": 3.0821,
+      "step": 261
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018282352941176468,
+      "loss": 2.5829,
+      "step": 262
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001835294117647059,
+      "loss": 2.9061,
+      "step": 263
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018423529411764705,
+      "loss": 3.0802,
+      "step": 264
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018494117647058821,
+      "loss": 2.7815,
+      "step": 265
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018564705882352937,
+      "loss": 2.8549,
+      "step": 266
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001863529411764706,
+      "loss": 3.2559,
+      "step": 267
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018705882352941175,
+      "loss": 2.8647,
+      "step": 268
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001877647058823529,
+      "loss": 3.2318,
+      "step": 269
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018847058823529412,
+      "loss": 2.4071,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018917647058823528,
+      "loss": 3.263,
+      "step": 271
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018988235294117644,
+      "loss": 3.2332,
+      "step": 272
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019058823529411763,
+      "loss": 1.4374,
+      "step": 273
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019129411764705882,
+      "loss": 3.1374,
+      "step": 274
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 3.2776,
+      "step": 275
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019270588235294117,
+      "loss": 3.0381,
+      "step": 276
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019341176470588233,
+      "loss": 2.3407,
+      "step": 277
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001941176470588235,
+      "loss": 3.0066,
+      "step": 278
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001948235294117647,
+      "loss": 2.8311,
+      "step": 279
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019552941176470586,
+      "loss": 2.9353,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019623529411764705,
+      "loss": 2.9753,
+      "step": 281
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019694117647058823,
+      "loss": 3.0526,
+      "step": 282
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001976470588235294,
+      "loss": 2.9806,
+      "step": 283
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019835294117647055,
+      "loss": 2.9637,
+      "step": 284
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019905882352941177,
+      "loss": 2.7804,
+      "step": 285
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019976470588235293,
+      "loss": 2.843,
+      "step": 286
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002004705882352941,
+      "loss": 2.4779,
+      "step": 287
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020117647058823525,
+      "loss": 3.3983,
+      "step": 288
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020188235294117646,
+      "loss": 2.6167,
+      "step": 289
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020258823529411762,
+      "loss": 2.2951,
+      "step": 290
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020329411764705878,
+      "loss": 2.7711,
+      "step": 291
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000204,
+      "loss": 2.646,
+      "step": 292
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020470588235294116,
+      "loss": 2.9032,
+      "step": 293
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020541176470588232,
+      "loss": 2.9645,
+      "step": 294
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002061176470588235,
+      "loss": 2.9049,
+      "step": 295
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002068235294117647,
+      "loss": 2.9816,
+      "step": 296
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020752941176470585,
+      "loss": 3.0213,
+      "step": 297
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020823529411764704,
+      "loss": 2.7091,
+      "step": 298
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002089411764705882,
+      "loss": 2.9557,
+      "step": 299
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002096470588235294,
+      "loss": 2.4238,
+      "step": 300
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021035294117647058,
+      "loss": 2.5357,
+      "step": 301
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021105882352941174,
+      "loss": 2.7141,
+      "step": 302
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021176470588235295,
+      "loss": 2.4318,
+      "step": 303
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002124705882352941,
+      "loss": 2.8379,
+      "step": 304
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021317647058823527,
+      "loss": 2.6104,
+      "step": 305
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021388235294117643,
+      "loss": 2.9486,
+      "step": 306
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021458823529411764,
+      "loss": 2.3565,
+      "step": 307
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002152941176470588,
+      "loss": 2.4382,
+      "step": 308
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 2.8377,
+      "step": 309
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021670588235294118,
+      "loss": 3.028,
+      "step": 310
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021741176470588234,
+      "loss": 2.4513,
+      "step": 311
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002181176470588235,
+      "loss": 2.5736,
+      "step": 312
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002188235294117647,
+      "loss": 2.9127,
+      "step": 313
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021952941176470587,
+      "loss": 2.9574,
+      "step": 314
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022023529411764703,
+      "loss": 2.8154,
+      "step": 315
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022094117647058822,
+      "loss": 2.6469,
+      "step": 316
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022164705882352938,
+      "loss": 3.1019,
+      "step": 317
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022235294117647057,
+      "loss": 2.8446,
+      "step": 318
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022305882352941176,
+      "loss": 3.0441,
+      "step": 319
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022376470588235292,
+      "loss": 2.7581,
+      "step": 320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002244705882352941,
+      "loss": 2.8101,
+      "step": 321
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002251764705882353,
+      "loss": 2.8564,
+      "step": 322
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022588235294117645,
+      "loss": 3.0528,
+      "step": 323
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002265882352941176,
+      "loss": 3.3095,
+      "step": 324
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022729411764705883,
+      "loss": 3.4034,
+      "step": 325
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 1.2303,
+      "step": 326
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022870588235294115,
+      "loss": 3.0976,
+      "step": 327
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002294117647058823,
+      "loss": 3.0638,
+      "step": 328
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023011764705882352,
+      "loss": 2.7229,
+      "step": 329
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023082352941176468,
+      "loss": 2.7375,
+      "step": 330
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023152941176470584,
+      "loss": 3.2297,
+      "step": 331
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023223529411764705,
+      "loss": 2.7985,
+      "step": 332
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023294117647058821,
+      "loss": 3.0643,
+      "step": 333
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023364705882352938,
+      "loss": 2.9876,
+      "step": 334
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023435294117647056,
+      "loss": 2.6189,
+      "step": 335
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023505882352941175,
+      "loss": 2.9049,
+      "step": 336
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002357647058823529,
+      "loss": 2.8024,
+      "step": 337
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002364705882352941,
+      "loss": 3.2064,
+      "step": 338
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023717647058823526,
+      "loss": 2.9218,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00023788235294117644,
+      "loss": 2.7387,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00023858823529411763,
+      "loss": 2.6079,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002392941176470588,
+      "loss": 3.3716,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 2.419,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024070588235294117,
+      "loss": 2.9697,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024141176470588233,
+      "loss": 3.2017,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002421176470588235,
+      "loss": 2.4636,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002428235294117647,
+      "loss": 2.7026,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024352941176470586,
+      "loss": 2.9215,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024423529411764705,
+      "loss": 1.372,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024494117647058824,
+      "loss": 2.7926,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002456470588235294,
+      "loss": 2.6838,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024635294117647056,
+      "loss": 2.2892,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024705882352941174,
+      "loss": 2.7693,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024776470588235293,
+      "loss": 2.8793,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002484705882352941,
+      "loss": 2.9703,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024917647058823525,
+      "loss": 2.9536,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024988235294117644,
+      "loss": 2.8714,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002505882352941176,
+      "loss": 2.8116,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002512941176470588,
+      "loss": 2.8654,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 2.546,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002527058823529412,
+      "loss": 3.1375,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002534117647058823,
+      "loss": 2.7663,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002541176470588235,
+      "loss": 2.5768,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002548235294117647,
+      "loss": 3.1344,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002555294117647059,
+      "loss": 2.8528,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000256235294117647,
+      "loss": 3.1172,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002569411764705882,
+      "loss": 2.8231,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002576470588235294,
+      "loss": 2.7593,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002583529411764706,
+      "loss": 2.7105,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025905882352941176,
+      "loss": 2.8255,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002597647058823529,
+      "loss": 3.3087,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002604705882352941,
+      "loss": 2.4326,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026117647058823527,
+      "loss": 3.2367,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026188235294117646,
+      "loss": 2.3644,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002625882352941176,
+      "loss": 3.0984,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026329411764705883,
+      "loss": 2.2658,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 3.0065,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026470588235294115,
+      "loss": 2.7171,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026541176470588234,
+      "loss": 2.6675,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026611764705882353,
+      "loss": 3.0569,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026682352941176466,
+      "loss": 3.313,
+      "step": 381
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026752941176470585,
+      "loss": 3.1305,
+      "step": 382
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026823529411764704,
+      "loss": 2.9335,
+      "step": 383
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002689411764705882,
+      "loss": 2.6562,
+      "step": 384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002696470588235294,
+      "loss": 2.9591,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027035294117647054,
+      "loss": 2.9788,
+      "step": 386
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027105882352941173,
+      "loss": 2.8802,
+      "step": 387
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002717647058823529,
+      "loss": 3.0358,
+      "step": 388
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002724705882352941,
+      "loss": 3.4578,
+      "step": 389
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002731764705882353,
+      "loss": 2.6853,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002738823529411765,
+      "loss": 3.0115,
+      "step": 391
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002745882352941176,
+      "loss": 2.988,
+      "step": 392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002752941176470588,
+      "loss": 2.8369,
+      "step": 393
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000276,
+      "loss": 2.988,
+      "step": 394
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002767058823529412,
+      "loss": 2.9298,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002774117647058823,
+      "loss": 2.777,
+      "step": 396
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002781176470588235,
+      "loss": 2.9725,
+      "step": 397
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002788235294117647,
+      "loss": 2.6059,
+      "step": 398
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027952941176470587,
+      "loss": 2.5258,
+      "step": 399
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000280235294117647,
+      "loss": 3.0367,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028094117647058824,
+      "loss": 2.9706,
+      "step": 401
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002816470588235294,
+      "loss": 3.4093,
+      "step": 402
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028235294117647056,
+      "loss": 3.2761,
+      "step": 403
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028305882352941175,
+      "loss": 2.4709,
+      "step": 404
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028376470588235294,
+      "loss": 2.9922,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028447058823529407,
+      "loss": 3.0863,
+      "step": 406
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028517647058823526,
+      "loss": 3.2879,
+      "step": 407
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028588235294117645,
+      "loss": 2.8273,
+      "step": 408
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028658823529411763,
+      "loss": 3.2607,
+      "step": 409
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002872941176470588,
+      "loss": 2.983,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 3.2198,
+      "step": 411
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028870588235294114,
+      "loss": 2.2069,
+      "step": 412
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028941176470588233,
+      "loss": 2.6191,
+      "step": 413
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002901176470588235,
+      "loss": 2.8506,
+      "step": 414
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029082352941176465,
+      "loss": 3.0611,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002915294117647059,
+      "loss": 2.6387,
+      "step": 416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000292235294117647,
+      "loss": 3.2521,
+      "step": 417
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002929411764705882,
+      "loss": 2.8467,
+      "step": 418
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002936470588235294,
+      "loss": 2.6175,
+      "step": 419
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002943529411764706,
+      "loss": 2.8541,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002950588235294117,
+      "loss": 2.7329,
+      "step": 421
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002957647058823529,
+      "loss": 2.5677,
+      "step": 422
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002964705882352941,
+      "loss": 3.3226,
+      "step": 423
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002971764705882353,
+      "loss": 3.022,
+      "step": 424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002978823529411764,
+      "loss": 2.7992,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985882352941176,
+      "loss": 3.1778,
+      "step": 426
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992941176470588,
+      "loss": 3.2537,
+      "step": 427
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003,
+      "loss": 3.0302,
+      "step": 428
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999999886027863,
+      "loss": 3.4287,
+      "step": 429
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999995441114686,
+      "loss": 2.759,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299999897425087,
+      "loss": 1.3603,
+      "step": 431
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999981764461533,
+      "loss": 3.2724,
+      "step": 432
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999997150697439,
+      "loss": 3.0597,
+      "step": 433
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999995897004884,
+      "loss": 3.0236,
+      "step": 434
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999994415368678,
+      "loss": 3.4095,
+      "step": 435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999927057890467,
+      "loss": 2.79,
+      "step": 436
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999907682662494,
+      "loss": 2.8319,
+      "step": 437
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999886028005807,
+      "loss": 2.3699,
+      "step": 438
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999862093923704,
+      "loss": 2.9766,
+      "step": 439
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999835880419815,
+      "loss": 3.0011,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999980738749812,
+      "loss": 3.0665,
+      "step": 441
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999776615162964,
+      "loss": 3.1322,
+      "step": 442
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999743563419,
+      "loss": 3.2479,
+      "step": 443
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999708232271275,
+      "loss": 2.811,
+      "step": 444
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999670621725136,
+      "loss": 2.3276,
+      "step": 445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999963073178631,
+      "loss": 3.0397,
+      "step": 446
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999588562460865,
+      "loss": 2.32,
+      "step": 447
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999544113755195,
+      "loss": 2.4878,
+      "step": 448
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999949738567607,
+      "loss": 3.2312,
+      "step": 449
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999448378230576,
+      "loss": 3.0678,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999397091426167,
+      "loss": 3.0058,
+      "step": 451
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999343525270635,
+      "loss": 3.2505,
+      "step": 452
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999928767977212,
+      "loss": 3.2742,
+      "step": 453
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999922955493911,
+      "loss": 2.9875,
+      "step": 454
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999916915078044,
+      "loss": 2.9964,
+      "step": 455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999910646730529,
+      "loss": 2.8123,
+      "step": 456
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999904150452318,
+      "loss": 3.1045,
+      "step": 457
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999897426244398,
+      "loss": 2.9001,
+      "step": 458
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998904741077915,
+      "loss": 3.0634,
+      "step": 459
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998832940435547,
+      "loss": 2.7983,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999875886052779,
+      "loss": 2.2765,
+      "step": 461
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299986825013659,
+      "loss": 2.9776,
+      "step": 462
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999860386296147,
+      "loss": 3.1128,
+      "step": 463
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998522945326467,
+      "loss": 3.1824,
+      "step": 464
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999843974847318,
+      "loss": 2.7342,
+      "step": 465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999835427241425,
+      "loss": 2.9349,
+      "step": 466
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999826651716267,
+      "loss": 3.4529,
+      "step": 467
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999817648273177,
+      "loss": 2.7017,
+      "step": 468
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999808416913524,
+      "loss": 2.8509,
+      "step": 469
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299979895763871,
+      "loss": 3.0764,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997892704501726,
+      "loss": 2.6586,
+      "step": 471
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997793553493846,
+      "loss": 2.9459,
+      "step": 472
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999769212337852,
+      "loss": 2.6776,
+      "step": 473
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999758841417117,
+      "loss": 3.2865,
+      "step": 474
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999748242588754,
+      "loss": 3.0409,
+      "step": 475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999737415854375,
+      "loss": 3.032,
+      "step": 476
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999726361215625,
+      "loss": 2.8145,
+      "step": 477
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997150786741837,
+      "loss": 2.7541,
+      "step": 478
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997035682317657,
+      "loss": 2.9318,
+      "step": 479
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996918298901205,
+      "loss": 2.8952,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996798636510314,
+      "loss": 3.1036,
+      "step": 481
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999667669516317,
+      "loss": 2.8233,
+      "step": 482
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299965524748783,
+      "loss": 2.8306,
+      "step": 483
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999642597567459,
+      "loss": 2.5802,
+      "step": 484
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999629719757125,
+      "loss": 3.1175,
+      "step": 485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996166140587864,
+      "loss": 2.7378,
+      "step": 486
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996032804744337,
+      "loss": 2.996,
+      "step": 487
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999589719006094,
+      "loss": 3.2481,
+      "step": 488
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995759296558267,
+      "loss": 2.7967,
+      "step": 489
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995619124257287,
+      "loss": 2.9625,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299954766731793,
+      "loss": 2.7125,
+      "step": 491
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995331943345944,
+      "loss": 3.1921,
+      "step": 492
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995184934779215,
+      "loss": 2.9394,
+      "step": 493
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999503564750146,
+      "loss": 3.2307,
+      "step": 494
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999488408153536,
+      "loss": 3.1269,
+      "step": 495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999473023690395,
+      "loss": 3.142,
+      "step": 496
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994574113630606,
+      "loss": 3.0516,
+      "step": 497
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999441571173905,
+      "loss": 2.8965,
+      "step": 498
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999425503125336,
+      "loss": 2.7006,
+      "step": 499
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999409207219795,
+      "loss": 2.9097,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999392683459758,
+      "loss": 3.2064,
+      "step": 501
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999375931847737,
+      "loss": 3.1324,
+      "step": 502
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993589523862773,
+      "loss": 3.1423,
+      "step": 503
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999341745077958,
+      "loss": 3.026,
+      "step": 504
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999324309925395,
+      "loss": 3.7584,
+      "step": 505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999306646931238,
+      "loss": 2.9856,
+      "step": 506
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999288756098171,
+      "loss": 3.0831,
+      "step": 507
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999270637428912,
+      "loss": 2.8362,
+      "step": 508
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999252290926215,
+      "loss": 2.7216,
+      "step": 509
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999233716592868,
+      "loss": 3.1239,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992149144316934,
+      "loss": 3.3,
+      "step": 511
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999195884445548,
+      "loss": 2.7427,
+      "step": 512
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991766266373245,
+      "loss": 3.061,
+      "step": 513
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991571410099496,
+      "loss": 2.8559,
+      "step": 514
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991374275663827,
+      "loss": 2.5308,
+      "step": 515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991174863096216,
+      "loss": 3.0731,
+      "step": 516
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999097317242695,
+      "loss": 2.7813,
+      "step": 517
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999076920368669,
+      "loss": 2.6651,
+      "step": 518
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990562956906425,
+      "loss": 2.7211,
+      "step": 519
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990354432117496,
+      "loss": 2.9943,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990143629351595,
+      "loss": 3.0497,
+      "step": 521
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989930548640756,
+      "loss": 3.4212,
+      "step": 522
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989715190017357,
+      "loss": 2.4517,
+      "step": 523
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989497553514126,
+      "loss": 3.0936,
+      "step": 524
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989277639164136,
+      "loss": 3.0707,
+      "step": 525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299890554470008,
+      "loss": 2.5666,
+      "step": 526
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299888309770579,
+      "loss": 2.8641,
+      "step": 527
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029988604229369526,
+      "loss": 2.6201,
+      "step": 528
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029988375203970143,
+      "loss": 2.7033,
+      "step": 529
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029988143900894557,
+      "loss": 3.111,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998791032017792,
+      "loss": 2.551,
+      "step": 531
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029987674461855716,
+      "loss": 3.2151,
+      "step": 532
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029987436325963797,
+      "loss": 3.0486,
+      "step": 533
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029987195912538356,
+      "loss": 3.2665,
+      "step": 534
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029986953221615914,
+      "loss": 3.227,
+      "step": 535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029986708253233355,
+      "loss": 3.3644,
+      "step": 536
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299864610074279,
+      "loss": 2.5556,
+      "step": 537
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998621148423714,
+      "loss": 3.0634,
+      "step": 538
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029985959683698975,
+      "loss": 2.9071,
+      "step": 539
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029985705605851677,
+      "loss": 3.473,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998544925073385,
+      "loss": 3.2157,
+      "step": 541
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998519061838446,
+      "loss": 3.305,
+      "step": 542
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299849297088428,
+      "loss": 2.5745,
+      "step": 543
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029984666522148525,
+      "loss": 2.5416,
+      "step": 544
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998440105834163,
+      "loss": 2.4935,
+      "step": 545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029984133317462446,
+      "loss": 2.7963,
+      "step": 546
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029983863299551676,
+      "loss": 3.0142,
+      "step": 547
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029983591004650336,
+      "loss": 2.53,
+      "step": 548
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029983316432799815,
+      "loss": 3.2314,
+      "step": 549
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998303958404184,
+      "loss": 2.8089,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998276045841847,
+      "loss": 2.6452,
+      "step": 551
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029982479055972133,
+      "loss": 1.35,
+      "step": 552
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029982195376745583,
+      "loss": 2.5062,
+      "step": 553
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029981909420781933,
+      "loss": 3.3925,
+      "step": 554
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998162118812464,
+      "loss": 3.1486,
+      "step": 555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029981330678817495,
+      "loss": 2.5474,
+      "step": 556
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998103789290466,
+      "loss": 3.3907,
+      "step": 557
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029980742830430616,
+      "loss": 2.7161,
+      "step": 558
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998044549144021,
+      "loss": 3.1625,
+      "step": 559
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998014587597861,
+      "loss": 3.0058,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997984398409137,
+      "loss": 3.1333,
+      "step": 561
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029979539815824346,
+      "loss": 2.6858,
+      "step": 562
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997923337122377,
+      "loss": 1.7261,
+      "step": 563
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997892465033621,
+      "loss": 2.5086,
+      "step": 564
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997861365320858,
+      "loss": 2.9279,
+      "step": 565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997830037988814,
+      "loss": 3.296,
+      "step": 566
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997798483042249,
+      "loss": 3.2893,
+      "step": 567
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997766700485959,
+      "loss": 3.0527,
+      "step": 568
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029977346903247733,
+      "loss": 3.1612,
+      "step": 569
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029977024525635563,
+      "loss": 3.0516,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997669987207207,
+      "loss": 2.4951,
+      "step": 571
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997637294260659,
+      "loss": 2.6846,
+      "step": 572
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029976043737288806,
+      "loss": 3.1255,
+      "step": 573
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029975712256168736,
+      "loss": 3.1129,
+      "step": 574
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029975378499296764,
+      "loss": 3.1554,
+      "step": 575
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029975042466723605,
+      "loss": 3.1986,
+      "step": 576
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029974704158500324,
+      "loss": 3.061,
+      "step": 577
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997436357467832,
+      "loss": 3.1684,
+      "step": 578
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029974020715309367,
+      "loss": 2.9492,
+      "step": 579
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997367558044556,
+      "loss": 3.4715,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997332817013934,
+      "loss": 3.3562,
+      "step": 581
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997297848444351,
+      "loss": 3.1125,
+      "step": 582
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299726265234112,
+      "loss": 2.6965,
+      "step": 583
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029972272287095904,
+      "loss": 2.3935,
+      "step": 584
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997191577555145,
+      "loss": 2.5587,
+      "step": 585
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997155698883201,
+      "loss": 2.9283,
+      "step": 586
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997119592699211,
+      "loss": 2.9874,
+      "step": 587
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997083259008662,
+      "loss": 3.309,
+      "step": 588
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029970466978170745,
+      "loss": 3.4495,
+      "step": 589
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997009909130006,
+      "loss": 2.9679,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029969728929530454,
+      "loss": 2.6644,
+      "step": 591
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029969356492918185,
+      "loss": 3.0369,
+      "step": 592
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029968981781519855,
+      "loss": 2.6801,
+      "step": 593
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029968604795392394,
+      "loss": 3.2161,
+      "step": 594
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029968225534593093,
+      "loss": 3.0532,
+      "step": 595
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299678439991796,
+      "loss": 2.9876,
+      "step": 596
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029967460189209876,
+      "loss": 2.9152,
+      "step": 597
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029967074104742257,
+      "loss": 2.914,
+      "step": 598
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299666857458354,
+      "loss": 1.5739,
+      "step": 599
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996629511254834,
+      "loss": 3.1111,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996590220494043,
+      "loss": 3.8482,
+      "step": 601
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029965507023071373,
+      "loss": 2.8865,
+      "step": 602
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996510956700123,
+      "loss": 3.4969,
+      "step": 603
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029964709836790383,
+      "loss": 2.9596,
+      "step": 604
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299643078324996,
+      "loss": 2.9148,
+      "step": 605
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029963903554189954,
+      "loss": 3.7714,
+      "step": 606
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029963497001922885,
+      "loss": 3.3505,
+      "step": 607
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996308817576017,
+      "loss": 2.5661,
+      "step": 608
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996267707576395,
+      "loss": 2.7489,
+      "step": 609
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996226370199668,
+      "loss": 2.8872,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996184805452119,
+      "loss": 3.4748,
+      "step": 611
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996143013340063,
+      "loss": 2.9112,
+      "step": 612
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002996100993869852,
+      "loss": 3.3058,
+      "step": 613
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029960587470478705,
+      "loss": 3.1303,
+      "step": 614
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029960162728805395,
+      "loss": 3.2109,
+      "step": 615
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029959735713743125,
+      "loss": 2.9656,
+      "step": 616
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029959306425356787,
+      "loss": 3.1149,
+      "step": 617
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995887486371162,
+      "loss": 3.3315,
+      "step": 618
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029958441028873206,
+      "loss": 2.8778,
+      "step": 619
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995800492090747,
+      "loss": 3.2984,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029957566539880686,
+      "loss": 3.5744,
+      "step": 621
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029957125885859466,
+      "loss": 3.1336,
+      "step": 622
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029956682958910783,
+      "loss": 3.0419,
+      "step": 623
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029956237759101934,
+      "loss": 2.8516,
+      "step": 624
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995579028650058,
+      "loss": 3.1051,
+      "step": 625
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995534054117472,
+      "loss": 3.2025,
+      "step": 626
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029954888523192695,
+      "loss": 2.918,
+      "step": 627
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299544342326232,
+      "loss": 3.0873,
+      "step": 628
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029953977669535265,
+      "loss": 3.1171,
+      "step": 629
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029953518833998274,
+      "loss": 3.2263,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995305772608195,
+      "loss": 3.1878,
+      "step": 631
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995259434585637,
+      "loss": 2.8479,
+      "step": 632
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995212869339194,
+      "loss": 3.5257,
+      "step": 633
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029951660768759436,
+      "loss": 3.073,
+      "step": 634
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995119057202995,
+      "loss": 2.6146,
+      "step": 635
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995071810327495,
+      "loss": 2.2897,
+      "step": 636
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995024336256622,
+      "loss": 2.4825,
+      "step": 637
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994976634997591,
+      "loss": 2.8777,
+      "step": 638
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029949287065576515,
+      "loss": 2.5103,
+      "step": 639
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029948805509440857,
+      "loss": 2.9638,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994832168164211,
+      "loss": 3.026,
+      "step": 641
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994783558225382,
+      "loss": 3.1124,
+      "step": 642
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029947347211349833,
+      "loss": 3.1018,
+      "step": 643
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994685656900437,
+      "loss": 3.0331,
+      "step": 644
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029946363655291994,
+      "loss": 3.3144,
+      "step": 645
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029945868470287616,
+      "loss": 2.7959,
+      "step": 646
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029945371014066473,
+      "loss": 2.978,
+      "step": 647
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994487128670416,
+      "loss": 2.6586,
+      "step": 648
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994436928827663,
+      "loss": 2.8908,
+      "step": 649
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994386501886016,
+      "loss": 2.6173,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029943358478531375,
+      "loss": 2.9721,
+      "step": 651
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029942849667367253,
+      "loss": 2.4379,
+      "step": 652
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029942338585445125,
+      "loss": 2.5643,
+      "step": 653
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029941825232842646,
+      "loss": 3.517,
+      "step": 654
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029941309609637825,
+      "loss": 3.178,
+      "step": 655
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994079171590902,
+      "loss": 2.8111,
+      "step": 656
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994027155173494,
+      "loss": 2.8516,
+      "step": 657
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993974911719462,
+      "loss": 3.4062,
+      "step": 658
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993922441236746,
+      "loss": 2.7298,
+      "step": 659
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993869743733318,
+      "loss": 3.1501,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993816819217188,
+      "loss": 2.574,
+      "step": 661
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993763667696397,
+      "loss": 3.0352,
+      "step": 662
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993710289179023,
+      "loss": 3.0637,
+      "step": 663
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993656683673177,
+      "loss": 2.8974,
+      "step": 664
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029936028511870055,
+      "loss": 2.9443,
+      "step": 665
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993548791728689,
+      "loss": 2.9226,
+      "step": 666
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993494505306442,
+      "loss": 2.5878,
+      "step": 667
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029934399919285143,
+      "loss": 3.0882,
+      "step": 668
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299338525160319,
+      "loss": 2.9846,
+      "step": 669
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993330284338788,
+      "loss": 2.2053,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029932750901436603,
+      "loss": 3.0056,
+      "step": 671
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993219669026195,
+      "loss": 2.7543,
+      "step": 672
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993164020994814,
+      "loss": 3.0386,
+      "step": 673
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993108146057974,
+      "loss": 3.7333,
+      "step": 674
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029930520442241653,
+      "loss": 3.0871,
+      "step": 675
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992995715501914,
+      "loss": 2.7087,
+      "step": 676
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029929391598997794,
+      "loss": 3.0163,
+      "step": 677
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029928823774263557,
+      "loss": 3.0944,
+      "step": 678
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029928253680902724,
+      "loss": 2.9192,
+      "step": 679
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029927681319001924,
+      "loss": 2.8262,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029927106688648137,
+      "loss": 2.9751,
+      "step": 681
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992652978992868,
+      "loss": 2.8797,
+      "step": 682
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029925950622931223,
+      "loss": 2.5401,
+      "step": 683
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029925369187743786,
+      "loss": 3.2561,
+      "step": 684
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029924785484454714,
+      "loss": 3.8023,
+      "step": 685
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992419951315271,
+      "loss": 3.0819,
+      "step": 686
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992361127392683,
+      "loss": 3.361,
+      "step": 687
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029923020766866454,
+      "loss": 3.3824,
+      "step": 688
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029922427992061317,
+      "loss": 3.2346,
+      "step": 689
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299218329496015,
+      "loss": 2.5669,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029921235639577436,
+      "loss": 2.429,
+      "step": 691
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029920636062079884,
+      "loss": 1.3914,
+      "step": 692
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992003421719996,
+      "loss": 3.152,
+      "step": 693
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029919430105029127,
+      "loss": 3.3003,
+      "step": 694
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029918823725659174,
+      "loss": 2.7584,
+      "step": 695
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029918215079182266,
+      "loss": 2.7744,
+      "step": 696
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029917604165690884,
+      "loss": 2.7507,
+      "step": 697
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029916990985277863,
+      "loss": 2.9829,
+      "step": 698
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002991637553803639,
+      "loss": 2.7207,
+      "step": 699
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029915757824059986,
+      "loss": 3.102,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029915137843442523,
+      "loss": 3.2801,
+      "step": 701
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029914515596278214,
+      "loss": 3.1457,
+      "step": 702
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002991389108266162,
+      "loss": 2.8097,
+      "step": 703
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029913264302687634,
+      "loss": 3.223,
+      "step": 704
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002991263525645151,
+      "loss": 3.1088,
+      "step": 705
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029912003944048847,
+      "loss": 3.0951,
+      "step": 706
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029911370365575567,
+      "loss": 3.1041,
+      "step": 707
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002991073452112796,
+      "loss": 3.0318,
+      "step": 708
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029910096410802655,
+      "loss": 2.7978,
+      "step": 709
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029909456034696606,
+      "loss": 2.7719,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029908813392907137,
+      "loss": 2.6674,
+      "step": 711
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299081684855319,
+      "loss": 2.7086,
+      "step": 712
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029907521312668903,
+      "loss": 2.761,
+      "step": 713
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990687187441649,
+      "loss": 3.1065,
+      "step": 714
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990622017087335,
+      "loss": 3.1261,
+      "step": 715
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029905566202138514,
+      "loss": 3.4489,
+      "step": 716
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990490996831137,
+      "loss": 2.6831,
+      "step": 717
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990425146949164,
+      "loss": 2.9912,
+      "step": 718
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029903590705779385,
+      "loss": 3.1566,
+      "step": 719
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990292767727502,
+      "loss": 2.599,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029902262384079296,
+      "loss": 3.4727,
+      "step": 721
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990159482629332,
+      "loss": 3.3478,
+      "step": 722
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029900925004018537,
+      "loss": 2.8178,
+      "step": 723
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002990025291735673,
+      "loss": 2.9025,
+      "step": 724
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029899578566410027,
+      "loss": 2.7618,
+      "step": 725
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029898901951280917,
+      "loss": 2.958,
+      "step": 726
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989822307207221,
+      "loss": 2.528,
+      "step": 727
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029897541928887076,
+      "loss": 2.4659,
+      "step": 728
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029896858521829016,
+      "loss": 2.5906,
+      "step": 729
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989617285100189,
+      "loss": 2.859,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989548491650989,
+      "loss": 2.9349,
+      "step": 731
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989479471845756,
+      "loss": 2.2265,
+      "step": 732
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989410225694978,
+      "loss": 2.9328,
+      "step": 733
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029893407532091785,
+      "loss": 2.3103,
+      "step": 734
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989271054398914,
+      "loss": 3.4947,
+      "step": 735
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029892011292747774,
+      "loss": 3.2235,
+      "step": 736
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989130977847393,
+      "loss": 3.0896,
+      "step": 737
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002989060600127422,
+      "loss": 3.2275,
+      "step": 738
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029889899961255595,
+      "loss": 2.0682,
+      "step": 739
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029889191658525347,
+      "loss": 3.159,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298884810931911,
+      "loss": 3.1403,
+      "step": 741
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988776826536085,
+      "loss": 2.549,
+      "step": 742
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029887053175142915,
+      "loss": 3.2373,
+      "step": 743
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988633582264595,
+      "loss": 3.3267,
+      "step": 744
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988561620797898,
+      "loss": 3.4588,
+      "step": 745
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988489433125136,
+      "loss": 3.2145,
+      "step": 746
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029884170192572777,
+      "loss": 3.5712,
+      "step": 747
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988344379205329,
+      "loss": 2.8628,
+      "step": 748
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029882715129803263,
+      "loss": 2.7226,
+      "step": 749
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988198420593345,
+      "loss": 2.3771,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298812510205549,
+      "loss": 3.556,
+      "step": 751
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988051557377905,
+      "loss": 3.2049,
+      "step": 752
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002987977786571765,
+      "loss": 3.4572,
+      "step": 753
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029879037896482806,
+      "loss": 3.1216,
+      "step": 754
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029878295666186966,
+      "loss": 2.4722,
+      "step": 755
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002987755117494292,
+      "loss": 3.2164,
+      "step": 756
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002987680442286381,
+      "loss": 3.4579,
+      "step": 757
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029876055410063107,
+      "loss": 3.0082,
+      "step": 758
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002987530413665464,
+      "loss": 3.0107,
+      "step": 759
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002987455060275256,
+      "loss": 3.3934,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029873794808471396,
+      "loss": 3.0204,
+      "step": 761
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029873036753925986,
+      "loss": 2.8658,
+      "step": 762
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002987227643923153,
+      "loss": 3.0444,
+      "step": 763
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029871513864503573,
+      "loss": 2.6328,
+      "step": 764
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029870749029857995,
+      "loss": 3.2415,
+      "step": 765
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029869981935411015,
+      "loss": 2.8063,
+      "step": 766
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029869212581279213,
+      "loss": 2.6973,
+      "step": 767
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298684409675795,
+      "loss": 3.1122,
+      "step": 768
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002986766709442913,
+      "loss": 2.6004,
+      "step": 769
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298668909619457,
+      "loss": 2.1256,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002986611257024716,
+      "loss": 2.8005,
+      "step": 771
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029865331919451795,
+      "loss": 3.1928,
+      "step": 772
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029864549009678233,
+      "loss": 3.3437,
+      "step": 773
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029863763841045447,
+      "loss": 3.2958,
+      "step": 774
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029862976413672754,
+      "loss": 2.7465,
+      "step": 775
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002986218672767981,
+      "loss": 3.027,
+      "step": 776
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029861394783186636,
+      "loss": 2.7511,
+      "step": 777
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002986060058031355,
+      "loss": 3.1115,
+      "step": 778
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985980411918126,
+      "loss": 2.5661,
+      "step": 779
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029859005399910797,
+      "loss": 3.0549,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029858204422623534,
+      "loss": 3.0619,
+      "step": 781
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985740118744119,
+      "loss": 2.3935,
+      "step": 782
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985659569448582,
+      "loss": 2.9419,
+      "step": 783
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985578794387984,
+      "loss": 1.8298,
+      "step": 784
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985497793574599,
+      "loss": 2.3011,
+      "step": 785
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985416567020737,
+      "loss": 3.2418,
+      "step": 786
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985335114738741,
+      "loss": 3.6587,
+      "step": 787
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985253436740988,
+      "loss": 3.1429,
+      "step": 788
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029851715330398907,
+      "loss": 2.9019,
+      "step": 789
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029850894036478957,
+      "loss": 3.0446,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029850070485774833,
+      "loss": 3.1691,
+      "step": 791
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984924467841168,
+      "loss": 3.1699,
+      "step": 792
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029848416614514995,
+      "loss": 2.5596,
+      "step": 793
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984758629421061,
+      "loss": 3.1104,
+      "step": 794
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984675371762471,
+      "loss": 2.8417,
+      "step": 795
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029845918884883804,
+      "loss": 1.5395,
+      "step": 796
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984508179611476,
+      "loss": 2.5654,
+      "step": 797
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029844242451444795,
+      "loss": 2.9995,
+      "step": 798
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029843400851001445,
+      "loss": 2.6699,
+      "step": 799
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984255699491261,
+      "loss": 2.4264,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029841710883306516,
+      "loss": 2.9096,
+      "step": 801
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029840862516311746,
+      "loss": 2.6981,
+      "step": 802
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984001189405722,
+      "loss": 3.2415,
+      "step": 803
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029839159016672203,
+      "loss": 2.9084,
+      "step": 804
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298383038842863,
+      "loss": 2.4474,
+      "step": 805
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002983744649702946,
+      "loss": 2.7087,
+      "step": 806
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029836586855031965,
+      "loss": 1.6068,
+      "step": 807
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029835724958424456,
+      "loss": 3.1705,
+      "step": 808
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002983486080733791,
+      "loss": 2.4808,
+      "step": 809
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002983399440190365,
+      "loss": 2.9721,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002983312574225333,
+      "loss": 2.5585,
+      "step": 811
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029832254828518956,
+      "loss": 2.9357,
+      "step": 812
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029831381660832873,
+      "loss": 2.8964,
+      "step": 813
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029830506239327773,
+      "loss": 3.2306,
+      "step": 814
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982962856413669,
+      "loss": 2.6883,
+      "step": 815
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982874863539299,
+      "loss": 3.1008,
+      "step": 816
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029827866453230393,
+      "loss": 3.2519,
+      "step": 817
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982698201778296,
+      "loss": 2.8043,
+      "step": 818
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982609532918509,
+      "loss": 2.7864,
+      "step": 819
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982520638757153,
+      "loss": 2.9014,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029824315193077364,
+      "loss": 2.8542,
+      "step": 821
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982342174583802,
+      "loss": 1.421,
+      "step": 822
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029822526045989264,
+      "loss": 2.5092,
+      "step": 823
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982162809366722,
+      "loss": 2.3613,
+      "step": 824
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002982072788900834,
+      "loss": 2.8558,
+      "step": 825
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029819825432149416,
+      "loss": 2.7217,
+      "step": 826
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002981892072322759,
+      "loss": 3.4121,
+      "step": 827
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029818013762380346,
+      "loss": 4.5607,
+      "step": 828
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002981710454974551,
+      "loss": 3.0698,
+      "step": 829
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029816193085461243,
+      "loss": 2.7713,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029815279369666063,
+      "loss": 3.0308,
+      "step": 831
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002981436340249881,
+      "loss": 3.1924,
+      "step": 832
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029813445184098686,
+      "loss": 3.4382,
+      "step": 833
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029812524714605215,
+      "loss": 2.925,
+      "step": 834
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029811601994158285,
+      "loss": 2.8631,
+      "step": 835
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002981067702289811,
+      "loss": 2.5966,
+      "step": 836
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029809749800965255,
+      "loss": 3.0971,
+      "step": 837
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002980882032850062,
+      "loss": 2.8269,
+      "step": 838
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002980788860564545,
+      "loss": 2.9129,
+      "step": 839
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029806954632541333,
+      "loss": 3.1911,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029806018409330197,
+      "loss": 2.6419,
+      "step": 841
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002980507993615432,
+      "loss": 2.9138,
+      "step": 842
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002980413921315631,
+      "loss": 3.1013,
+      "step": 843
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002980319624047912,
+      "loss": 3.3109,
+      "step": 844
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029802251018266044,
+      "loss": 2.724,
+      "step": 845
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002980130354666073,
+      "loss": 3.2303,
+      "step": 846
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029800353825807155,
+      "loss": 2.9583,
+      "step": 847
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029799401855849637,
+      "loss": 3.0308,
+      "step": 848
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002979844763693284,
+      "loss": 2.512,
+      "step": 849
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002979749116920178,
+      "loss": 2.7699,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000297965324528018,
+      "loss": 2.2254,
+      "step": 851
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029795571487878576,
+      "loss": 2.9468,
+      "step": 852
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002979460827457816,
+      "loss": 3.3028,
+      "step": 853
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029793642813046907,
+      "loss": 2.8566,
+      "step": 854
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029792675103431537,
+      "loss": 2.5915,
+      "step": 855
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029791705145879114,
+      "loss": 2.6238,
+      "step": 856
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029790732940537026,
+      "loss": 1.4443,
+      "step": 857
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002978975848755302,
+      "loss": 3.2194,
+      "step": 858
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029788781787075164,
+      "loss": 3.0195,
+      "step": 859
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002978780283925189,
+      "loss": 3.0866,
+      "step": 860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029786821644231965,
+      "loss": 3.4054,
+      "step": 861
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029785838202164486,
+      "loss": 2.4338,
+      "step": 862
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000297848525131989,
+      "loss": 3.2552,
+      "step": 863
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029783864577484997,
+      "loss": 3.0145,
+      "step": 864
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029782874395172914,
+      "loss": 2.947,
+      "step": 865
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002978188196641311,
+      "loss": 3.0004,
+      "step": 866
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000297808872913564,
+      "loss": 3.2112,
+      "step": 867
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002977989037015395,
+      "loss": 2.374,
+      "step": 868
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029778891202957236,
+      "loss": 3.2266,
+      "step": 869
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029777889789918107,
+      "loss": 3.1306,
+      "step": 870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029776886131188733,
+      "loss": 3.1248,
+      "step": 871
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002977588022692164,
+      "loss": 2.8406,
+      "step": 872
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029774872077269687,
+      "loss": 2.8333,
+      "step": 873
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002977386168238607,
+      "loss": 3.2521,
+      "step": 874
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029772849042424335,
+      "loss": 3.1621,
+      "step": 875
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029771834157538363,
+      "loss": 3.1371,
+      "step": 876
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029770817027882385,
+      "loss": 2.8705,
+      "step": 877
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029769797653610966,
+      "loss": 2.764,
+      "step": 878
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029768776034879004,
+      "loss": 3.1855,
+      "step": 879
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029767752171841753,
+      "loss": 1.3198,
+      "step": 880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029766726064654807,
+      "loss": 3.0734,
+      "step": 881
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002976569771347409,
+      "loss": 2.9256,
+      "step": 882
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002976466711845587,
+      "loss": 2.2611,
+      "step": 883
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029763634279756767,
+      "loss": 3.1894,
+      "step": 884
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029762599197533736,
+      "loss": 3.0589,
+      "step": 885
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002976156187194406,
+      "loss": 3.0696,
+      "step": 886
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029760522303145376,
+      "loss": 3.1615,
+      "step": 887
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029759480491295675,
+      "loss": 2.5052,
+      "step": 888
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002975843643655325,
+      "loss": 3.1018,
+      "step": 889
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029757390139076783,
+      "loss": 3.6512,
+      "step": 890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002975634159902525,
+      "loss": 3.0801,
+      "step": 891
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029755290816558004,
+      "loss": 2.7738,
+      "step": 892
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029754237791834726,
+      "loss": 3.0349,
+      "step": 893
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029753182525015426,
+      "loss": 2.8232,
+      "step": 894
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029752125016260474,
+      "loss": 3.2597,
+      "step": 895
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002975106526573057,
+      "loss": 2.8879,
+      "step": 896
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029750003273586753,
+      "loss": 2.2516,
+      "step": 897
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002974893903999041,
+      "loss": 2.987,
+      "step": 898
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002974787256510327,
+      "loss": 3.7526,
+      "step": 899
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029746803849087385,
+      "loss": 3.4865,
+      "step": 900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002974573289210517,
+      "loss": 3.2011,
+      "step": 901
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002974465969431937,
+      "loss": 3.8636,
+      "step": 902
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029743584255893064,
+      "loss": 3.0094,
+      "step": 903
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002974250657698969,
+      "loss": 2.8718,
+      "step": 904
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029741426657773,
+      "loss": 2.2779,
+      "step": 905
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002974034449840712,
+      "loss": 3.083,
+      "step": 906
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002973926009905648,
+      "loss": 2.8087,
+      "step": 907
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029738173459885884,
+      "loss": 3.6549,
+      "step": 908
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002973708458106045,
+      "loss": 2.7247,
+      "step": 909
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002973599346274565,
+      "loss": 2.7473,
+      "step": 910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029734900105107294,
+      "loss": 2.4239,
+      "step": 911
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002973380450831154,
+      "loss": 3.1531,
+      "step": 912
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002973270667252486,
+      "loss": 2.8241,
+      "step": 913
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000297316065979141,
+      "loss": 2.8566,
+      "step": 914
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029730504284646416,
+      "loss": 2.5846,
+      "step": 915
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029729399732889337,
+      "loss": 2.7887,
+      "step": 916
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029728292942810696,
+      "loss": 3.1729,
+      "step": 917
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029727183914578703,
+      "loss": 2.9951,
+      "step": 918
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002972607264836187,
+      "loss": 1.9227,
+      "step": 919
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029724959144329075,
+      "loss": 2.9137,
+      "step": 920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002972384340264953,
+      "loss": 3.3913,
+      "step": 921
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000297227254234928,
+      "loss": 2.1551,
+      "step": 922
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029721605207028753,
+      "loss": 3.0074,
+      "step": 923
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002972048275342763,
+      "loss": 2.8015,
+      "step": 924
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029719358062860006,
+      "loss": 3.095,
+      "step": 925
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029718231135496785,
+      "loss": 2.7467,
+      "step": 926
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029717101971509223,
+      "loss": 2.8729,
+      "step": 927
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002971597057106892,
+      "loss": 2.8806,
+      "step": 928
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029714836934347785,
+      "loss": 3.2879,
+      "step": 929
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029713701061518104,
+      "loss": 2.7978,
+      "step": 930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002971256295275248,
+      "loss": 2.5107,
+      "step": 931
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002971142260822388,
+      "loss": 3.2869,
+      "step": 932
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002971028002810557,
+      "loss": 3.0126,
+      "step": 933
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002970913521257119,
+      "loss": 2.955,
+      "step": 934
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029707988161794716,
+      "loss": 2.6878,
+      "step": 935
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002970683887595045,
+      "loss": 2.9298,
+      "step": 936
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029705687355213045,
+      "loss": 3.0588,
+      "step": 937
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029704533599757483,
+      "loss": 3.0412,
+      "step": 938
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029703377609759097,
+      "loss": 2.7513,
+      "step": 939
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002970221938539355,
+      "loss": 2.8019,
+      "step": 940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029701058926836853,
+      "loss": 2.3827,
+      "step": 941
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029699896234265357,
+      "loss": 2.7091,
+      "step": 942
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029698731307855735,
+      "loss": 2.9786,
+      "step": 943
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002969756414778503,
+      "loss": 2.684,
+      "step": 944
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002969639475423059,
+      "loss": 3.1119,
+      "step": 945
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002969522312737013,
+      "loss": 3.011,
+      "step": 946
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029694049267381693,
+      "loss": 2.6288,
+      "step": 947
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029692873174443654,
+      "loss": 3.4253,
+      "step": 948
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002969169484873474,
+      "loss": 3.0893,
+      "step": 949
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002969051429043402,
+      "loss": 3.2267,
+      "step": 950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002968933149972089,
+      "loss": 3.5378,
+      "step": 951
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002968814647677508,
+      "loss": 1.9134,
+      "step": 952
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029686959221776686,
+      "loss": 2.9571,
+      "step": 953
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002968576973490612,
+      "loss": 1.1832,
+      "step": 954
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029684578016344134,
+      "loss": 2.7346,
+      "step": 955
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002968338406627183,
+      "loss": 2.8717,
+      "step": 956
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002968218788487065,
+      "loss": 3.2523,
+      "step": 957
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029680989472322355,
+      "loss": 2.775,
+      "step": 958
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967978882880907,
+      "loss": 3.217,
+      "step": 959
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967858595451325,
+      "loss": 2.8854,
+      "step": 960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967738084961768,
+      "loss": 1.8798,
+      "step": 961
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967617351430549,
+      "loss": 2.7253,
+      "step": 962
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967496394876016,
+      "loss": 2.7885,
+      "step": 963
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029673752153165493,
+      "loss": 2.409,
+      "step": 964
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967253812770563,
+      "loss": 3.1987,
+      "step": 965
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002967132187256507,
+      "loss": 3.3483,
+      "step": 966
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029670103387928633,
+      "loss": 2.9219,
+      "step": 967
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002966888267398148,
+      "loss": 2.3708,
+      "step": 968
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029667659730909124,
+      "loss": 2.4083,
+      "step": 969
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029666434558897403,
+      "loss": 2.7104,
+      "step": 970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002966520715813249,
+      "loss": 2.888,
+      "step": 971
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029663977528800905,
+      "loss": 3.0051,
+      "step": 972
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029662745671089517,
+      "loss": 2.584,
+      "step": 973
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029661511585185515,
+      "loss": 2.7916,
+      "step": 974
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002966027527127644,
+      "loss": 2.9478,
+      "step": 975
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002965903672955016,
+      "loss": 3.2768,
+      "step": 976
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002965779596019488,
+      "loss": 3.3101,
+      "step": 977
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029656552963399166,
+      "loss": 2.2758,
+      "step": 978
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000296553077393519,
+      "loss": 3.075,
+      "step": 979
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002965406028824231,
+      "loss": 2.9187,
+      "step": 980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029652810610259964,
+      "loss": 2.7516,
+      "step": 981
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029651558705594765,
+      "loss": 3.0762,
+      "step": 982
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002965030457443696,
+      "loss": 3.046,
+      "step": 983
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002964904821697712,
+      "loss": 2.4406,
+      "step": 984
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002964778963340618,
+      "loss": 3.2784,
+      "step": 985
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029646528823915383,
+      "loss": 2.5152,
+      "step": 986
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029645265788696333,
+      "loss": 2.9226,
+      "step": 987
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002964400052794096,
+      "loss": 2.5813,
+      "step": 988
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029642733041841543,
+      "loss": 2.9573,
+      "step": 989
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002964146333059069,
+      "loss": 2.8713,
+      "step": 990
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002964019139438135,
+      "loss": 3.2241,
+      "step": 991
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002963891723340681,
+      "loss": 2.919,
+      "step": 992
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002963764084786069,
+      "loss": 2.3101,
+      "step": 993
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002963636223793696,
+      "loss": 2.4686,
+      "step": 994
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002963508140382992,
+      "loss": 2.9191,
+      "step": 995
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002963379834573421,
+      "loss": 2.5754,
+      "step": 996
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002963251306384481,
+      "loss": 2.3529,
+      "step": 997
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029631225558357017,
+      "loss": 3.2797,
+      "step": 998
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029629935829466507,
+      "loss": 2.6987,
+      "step": 999
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002962864387736926,
+      "loss": 2.8838,
+      "step": 1000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029627349702261604,
+      "loss": 2.6708,
+      "step": 1001
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002962605330434021,
+      "loss": 2.9412,
+      "step": 1002
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029624754683802075,
+      "loss": 3.4685,
+      "step": 1003
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002962345384084455,
+      "loss": 2.6186,
+      "step": 1004
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002962215077566531,
+      "loss": 2.8178,
+      "step": 1005
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029620845488462374,
+      "loss": 3.1369,
+      "step": 1006
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029619537979434096,
+      "loss": 2.8163,
+      "step": 1007
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002961822824877917,
+      "loss": 2.7798,
+      "step": 1008
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029616916296696625,
+      "loss": 2.5794,
+      "step": 1009
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002961560212338583,
+      "loss": 2.7866,
+      "step": 1010
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002961428572904649,
+      "loss": 3.0286,
+      "step": 1011
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002961296711387865,
+      "loss": 3.3362,
+      "step": 1012
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029611646278082687,
+      "loss": 2.7021,
+      "step": 1013
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002961032322185932,
+      "loss": 2.9957,
+      "step": 1014
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029608997945409604,
+      "loss": 3.0112,
+      "step": 1015
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002960767044893494,
+      "loss": 3.1267,
+      "step": 1016
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029606340732637044,
+      "loss": 3.1228,
+      "step": 1017
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029605008796717994,
+      "loss": 2.5103,
+      "step": 1018
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029603674641380187,
+      "loss": 3.488,
+      "step": 1019
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002960233826682637,
+      "loss": 3.0732,
+      "step": 1020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029600999673259627,
+      "loss": 2.9122,
+      "step": 1021
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029599658860883363,
+      "loss": 2.476,
+      "step": 1022
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002959831582990134,
+      "loss": 3.0117,
+      "step": 1023
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002959697058051765,
+      "loss": 2.8609,
+      "step": 1024
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002959562311293671,
+      "loss": 2.7152,
+      "step": 1025
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000295942734273633,
+      "loss": 2.961,
+      "step": 1026
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029592921524002514,
+      "loss": 2.8291,
+      "step": 1027
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002959156740305979,
+      "loss": 2.9342,
+      "step": 1028
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000295902110647409,
+      "loss": 3.0214,
+      "step": 1029
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029588852509251966,
+      "loss": 2.7744,
+      "step": 1030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002958749173679944,
+      "loss": 2.8776,
+      "step": 1031
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000295861287475901,
+      "loss": 2.4675,
+      "step": 1032
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029584763541831076,
+      "loss": 3.0355,
+      "step": 1033
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002958339611972982,
+      "loss": 2.5111,
+      "step": 1034
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029582026481494143,
+      "loss": 2.1782,
+      "step": 1035
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029580654627332163,
+      "loss": 3.0379,
+      "step": 1036
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029579280557452366,
+      "loss": 3.232,
+      "step": 1037
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029577904272063554,
+      "loss": 3.5956,
+      "step": 1038
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029576525771374864,
+      "loss": 2.3572,
+      "step": 1039
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029575145055595785,
+      "loss": 2.9649,
+      "step": 1040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002957376212493613,
+      "loss": 2.9153,
+      "step": 1041
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002957237697960606,
+      "loss": 2.7081,
+      "step": 1042
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002957098961981606,
+      "loss": 2.6037,
+      "step": 1043
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002956960004577696,
+      "loss": 3.2755,
+      "step": 1044
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002956820825769992,
+      "loss": 2.6583,
+      "step": 1045
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002956681425579644,
+      "loss": 2.6299,
+      "step": 1046
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002956541804027836,
+      "loss": 3.1256,
+      "step": 1047
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002956401961135785,
+      "loss": 2.3781,
+      "step": 1048
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029562618969247424,
+      "loss": 2.8674,
+      "step": 1049
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029561216114159925,
+      "loss": 3.3315,
+      "step": 1050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029559811046308535,
+      "loss": 3.2973,
+      "step": 1051
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002955840376590677,
+      "loss": 2.5142,
+      "step": 1052
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002955699427316848,
+      "loss": 2.4799,
+      "step": 1053
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002955558256830787,
+      "loss": 1.5244,
+      "step": 1054
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029554168651539447,
+      "loss": 2.9863,
+      "step": 1055
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002955275252307809,
+      "loss": 2.8178,
+      "step": 1056
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002955133418313899,
+      "loss": 2.8306,
+      "step": 1057
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002954991363193769,
+      "loss": 3.1616,
+      "step": 1058
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029548490869690054,
+      "loss": 2.7554,
+      "step": 1059
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002954706589661229,
+      "loss": 3.4022,
+      "step": 1060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002954563871292094,
+      "loss": 2.9035,
+      "step": 1061
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002954420931883289,
+      "loss": 3.0368,
+      "step": 1062
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002954277771456534,
+      "loss": 3.2285,
+      "step": 1063
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002954134390033586,
+      "loss": 2.6334,
+      "step": 1064
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002953990787636232,
+      "loss": 3.0845,
+      "step": 1065
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002953846964286295,
+      "loss": 2.5978,
+      "step": 1066
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029537029200056304,
+      "loss": 3.2898,
+      "step": 1067
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002953558654816128,
+      "loss": 2.8531,
+      "step": 1068
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029534141687397107,
+      "loss": 2.6745,
+      "step": 1069
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002953269461798335,
+      "loss": 2.0595,
+      "step": 1070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029531245340139905,
+      "loss": 3.412,
+      "step": 1071
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029529793854087016,
+      "loss": 2.8889,
+      "step": 1072
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002952834016004525,
+      "loss": 3.2055,
+      "step": 1073
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002952688425823552,
+      "loss": 3.3295,
+      "step": 1074
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002952542614887906,
+      "loss": 3.1964,
+      "step": 1075
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002952396583219745,
+      "loss": 2.9293,
+      "step": 1076
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029522503308412613,
+      "loss": 2.877,
+      "step": 1077
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029521038577746793,
+      "loss": 2.9084,
+      "step": 1078
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002951957164042257,
+      "loss": 3.3334,
+      "step": 1079
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002951810249666287,
+      "loss": 3.265,
+      "step": 1080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002951663114669095,
+      "loss": 2.6798,
+      "step": 1081
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002951515759073039,
+      "loss": 2.9713,
+      "step": 1082
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029513681829005127,
+      "loss": 3.1355,
+      "step": 1083
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029512203861739415,
+      "loss": 2.6306,
+      "step": 1084
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029510723689157853,
+      "loss": 2.6789,
+      "step": 1085
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029509241311485374,
+      "loss": 2.8752,
+      "step": 1086
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002950775672894724,
+      "loss": 2.7474,
+      "step": 1087
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029506269941769055,
+      "loss": 2.9859,
+      "step": 1088
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029504780950176753,
+      "loss": 3.0777,
+      "step": 1089
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002950328975439661,
+      "loss": 2.7593,
+      "step": 1090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002950179635465523,
+      "loss": 2.402,
+      "step": 1091
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029500300751179556,
+      "loss": 2.9708,
+      "step": 1092
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002949880294419686,
+      "loss": 2.4485,
+      "step": 1093
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029497302933934754,
+      "loss": 2.7718,
+      "step": 1094
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029495800720621187,
+      "loss": 2.8448,
+      "step": 1095
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029494296304484436,
+      "loss": 3.3419,
+      "step": 1096
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002949278968575312,
+      "loss": 3.2136,
+      "step": 1097
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029491280864656186,
+      "loss": 3.4249,
+      "step": 1098
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029489769841422925,
+      "loss": 2.4805,
+      "step": 1099
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029488256616282945,
+      "loss": 2.9071,
+      "step": 1100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000294867411894662,
+      "loss": 2.9571,
+      "step": 1101
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029485223561202996,
+      "loss": 2.8129,
+      "step": 1102
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002948370373172394,
+      "loss": 2.9191,
+      "step": 1103
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002948218170126,
+      "loss": 2.955,
+      "step": 1104
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029480657470042456,
+      "loss": 3.0128,
+      "step": 1105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002947913103830295,
+      "loss": 3.0047,
+      "step": 1106
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029477602406273424,
+      "loss": 3.3874,
+      "step": 1107
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002947607157418619,
+      "loss": 2.8657,
+      "step": 1108
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002947453854227387,
+      "loss": 3.0077,
+      "step": 1109
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029473003310769426,
+      "loss": 3.3921,
+      "step": 1110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029471465879906165,
+      "loss": 3.1904,
+      "step": 1111
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029469926249917706,
+      "loss": 3.0288,
+      "step": 1112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029468384421038027,
+      "loss": 2.9237,
+      "step": 1113
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002946684039350142,
+      "loss": 3.2123,
+      "step": 1114
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002946529416754253,
+      "loss": 3.1607,
+      "step": 1115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002946374574339632,
+      "loss": 2.9707,
+      "step": 1116
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029462195121298095,
+      "loss": 2.6848,
+      "step": 1117
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029460642301483485,
+      "loss": 3.0611,
+      "step": 1118
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029459087284188466,
+      "loss": 2.7796,
+      "step": 1119
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002945753006964935,
+      "loss": 2.9245,
+      "step": 1120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029455970658102764,
+      "loss": 2.8831,
+      "step": 1121
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029454409049785687,
+      "loss": 2.7472,
+      "step": 1122
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029452845244935425,
+      "loss": 3.4551,
+      "step": 1123
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029451279243789615,
+      "loss": 3.5008,
+      "step": 1124
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002944971104658624,
+      "loss": 2.2932,
+      "step": 1125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029448140653563586,
+      "loss": 2.842,
+      "step": 1126
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029446568064960324,
+      "loss": 2.8776,
+      "step": 1127
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029444993281015406,
+      "loss": 2.7391,
+      "step": 1128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002944341630196815,
+      "loss": 4.0073,
+      "step": 1129
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000294418371280582,
+      "loss": 2.5543,
+      "step": 1130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029440255759525526,
+      "loss": 2.9611,
+      "step": 1131
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002943867219661044,
+      "loss": 2.8149,
+      "step": 1132
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029437086439553585,
+      "loss": 2.9891,
+      "step": 1133
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029435498488595944,
+      "loss": 2.5911,
+      "step": 1134
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029433908343978813,
+      "loss": 3.2614,
+      "step": 1135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029432316005943845,
+      "loss": 3.3296,
+      "step": 1136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002943072147473301,
+      "loss": 2.7847,
+      "step": 1137
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029429124750588625,
+      "loss": 3.0227,
+      "step": 1138
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029427525833753326,
+      "loss": 3.1984,
+      "step": 1139
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002942592472447009,
+      "loss": 2.8106,
+      "step": 1140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002942432142298222,
+      "loss": 2.9631,
+      "step": 1141
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002942271592953338,
+      "loss": 2.7436,
+      "step": 1142
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002942110824436752,
+      "loss": 2.9765,
+      "step": 1143
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029419498367728964,
+      "loss": 2.9824,
+      "step": 1144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002941788629986235,
+      "loss": 3.1378,
+      "step": 1145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029416272041012646,
+      "loss": 3.1912,
+      "step": 1146
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029414655591425166,
+      "loss": 2.1156,
+      "step": 1147
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002941303695134555,
+      "loss": 2.9338,
+      "step": 1148
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002941141612101977,
+      "loss": 2.7573,
+      "step": 1149
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002940979310069413,
+      "loss": 2.3467,
+      "step": 1150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002940816789061527,
+      "loss": 2.9857,
+      "step": 1151
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029406540491030167,
+      "loss": 2.9344,
+      "step": 1152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029404910902186115,
+      "loss": 2.2701,
+      "step": 1153
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029403279124330764,
+      "loss": 3.3017,
+      "step": 1154
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002940164515771207,
+      "loss": 2.5294,
+      "step": 1155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002940000900257834,
+      "loss": 3.0422,
+      "step": 1156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029398370659178215,
+      "loss": 3.0149,
+      "step": 1157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002939673012776066,
+      "loss": 2.6666,
+      "step": 1158
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029395087408574965,
+      "loss": 2.8776,
+      "step": 1159
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002939344250187077,
+      "loss": 2.8413,
+      "step": 1160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029391795407898046,
+      "loss": 3.1641,
+      "step": 1161
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002939014612690708,
+      "loss": 2.6497,
+      "step": 1162
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000293884946591485,
+      "loss": 3.1089,
+      "step": 1163
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002938684100487328,
+      "loss": 2.4899,
+      "step": 1164
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000293851851643327,
+      "loss": 3.1306,
+      "step": 1165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000293835271377784,
+      "loss": 2.5747,
+      "step": 1166
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029381866925462324,
+      "loss": 2.8234,
+      "step": 1167
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029380204527636774,
+      "loss": 2.9755,
+      "step": 1168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029378539944554367,
+      "loss": 3.1736,
+      "step": 1169
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002937687317646806,
+      "loss": 2.2903,
+      "step": 1170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002937520422363114,
+      "loss": 3.1042,
+      "step": 1171
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029373533086297227,
+      "loss": 1.5341,
+      "step": 1172
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029371859764720266,
+      "loss": 2.5523,
+      "step": 1173
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029370184259154547,
+      "loss": 2.3864,
+      "step": 1174
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029368506569854677,
+      "loss": 3.145,
+      "step": 1175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029366826697075607,
+      "loss": 2.662,
+      "step": 1176
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029365144641072617,
+      "loss": 3.5962,
+      "step": 1177
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029363460402101314,
+      "loss": 2.9739,
+      "step": 1178
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002936177398041764,
+      "loss": 2.7451,
+      "step": 1179
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029360085376277866,
+      "loss": 3.0254,
+      "step": 1180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002935839458993861,
+      "loss": 2.8235,
+      "step": 1181
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002935670162165679,
+      "loss": 2.8801,
+      "step": 1182
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029355006471689683,
+      "loss": 2.9124,
+      "step": 1183
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029353309140294897,
+      "loss": 2.7737,
+      "step": 1184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029351609627730347,
+      "loss": 2.6971,
+      "step": 1185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002934990793425431,
+      "loss": 3.4732,
+      "step": 1186
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002934820406012537,
+      "loss": 3.0331,
+      "step": 1187
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029346498005602457,
+      "loss": 1.6057,
+      "step": 1188
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029344789770944835,
+      "loss": 3.0182,
+      "step": 1189
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029343079356412073,
+      "loss": 1.8399,
+      "step": 1190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002934136676226411,
+      "loss": 2.6961,
+      "step": 1191
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002933965198876119,
+      "loss": 2.6309,
+      "step": 1192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002933793503616389,
+      "loss": 3.254,
+      "step": 1193
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029336215904733127,
+      "loss": 3.277,
+      "step": 1194
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002933449459473015,
+      "loss": 3.1915,
+      "step": 1195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029332771106416525,
+      "loss": 3.1417,
+      "step": 1196
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029331045440054163,
+      "loss": 3.0646,
+      "step": 1197
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000293293175959053,
+      "loss": 2.918,
+      "step": 1198
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000293275875742325,
+      "loss": 2.757,
+      "step": 1199
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002932585537529868,
+      "loss": 3.166,
+      "step": 1200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002932412099936704,
+      "loss": 3.3116,
+      "step": 1201
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029322384446701173,
+      "loss": 3.1528,
+      "step": 1202
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029320645717564945,
+      "loss": 3.0018,
+      "step": 1203
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029318904812222595,
+      "loss": 2.1797,
+      "step": 1204
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002931716173093866,
+      "loss": 2.8093,
+      "step": 1205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002931541647397804,
+      "loss": 2.8529,
+      "step": 1206
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002931366904160594,
+      "loss": 3.4573,
+      "step": 1207
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029311919434087905,
+      "loss": 3.4231,
+      "step": 1208
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029310167651689815,
+      "loss": 2.6195,
+      "step": 1209
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029308413694677867,
+      "loss": 3.3962,
+      "step": 1210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002930665756331861,
+      "loss": 3.3184,
+      "step": 1211
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029304899257878903,
+      "loss": 2.795,
+      "step": 1212
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002930313877862594,
+      "loss": 2.8848,
+      "step": 1213
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002930137612582725,
+      "loss": 2.7534,
+      "step": 1214
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000292996112997507,
+      "loss": 2.4473,
+      "step": 1215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002929784430066447,
+      "loss": 2.8129,
+      "step": 1216
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029296075128837075,
+      "loss": 2.8564,
+      "step": 1217
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029294303784537374,
+      "loss": 3.2594,
+      "step": 1218
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002929253026803454,
+      "loss": 3.436,
+      "step": 1219
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002929075457959807,
+      "loss": 3.0415,
+      "step": 1220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029288976719497824,
+      "loss": 2.8136,
+      "step": 1221
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002928719668800395,
+      "loss": 3.0087,
+      "step": 1222
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002928541448538697,
+      "loss": 2.6431,
+      "step": 1223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029283630111917694,
+      "loss": 2.5268,
+      "step": 1224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002928184356786728,
+      "loss": 2.8961,
+      "step": 1225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002928005485350723,
+      "loss": 2.8229,
+      "step": 1226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002927826396910935,
+      "loss": 2.967,
+      "step": 1227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029276470914945795,
+      "loss": 2.9473,
+      "step": 1228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029274675691289036,
+      "loss": 2.8857,
+      "step": 1229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002927287829841189,
+      "loss": 2.6655,
+      "step": 1230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029271078736587487,
+      "loss": 2.9007,
+      "step": 1231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002926927700608929,
+      "loss": 3.3886,
+      "step": 1232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029267473107191105,
+      "loss": 2.5608,
+      "step": 1233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002926566704016705,
+      "loss": 2.6742,
+      "step": 1234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029263858805291584,
+      "loss": 2.5785,
+      "step": 1235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002926204840283949,
+      "loss": 1.5877,
+      "step": 1236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002926023583308588,
+      "loss": 2.8513,
+      "step": 1237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000292584210963062,
+      "loss": 3.1261,
+      "step": 1238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002925660419277622,
+      "loss": 2.8159,
+      "step": 1239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029254785122772046,
+      "loss": 2.1713,
+      "step": 1240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002925296388657011,
+      "loss": 3.3518,
+      "step": 1241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029251140484447166,
+      "loss": 2.661,
+      "step": 1242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029249314916680305,
+      "loss": 3.2175,
+      "step": 1243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029247487183546946,
+      "loss": 2.9561,
+      "step": 1244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002924565728532484,
+      "loss": 2.3205,
+      "step": 1245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002924382522229206,
+      "loss": 2.916,
+      "step": 1246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029241990994727013,
+      "loss": 2.9484,
+      "step": 1247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002924015460290843,
+      "loss": 2.771,
+      "step": 1248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002923831604711538,
+      "loss": 2.8333,
+      "step": 1249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029236475327627256,
+      "loss": 2.1146,
+      "step": 1250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029234632444723765,
+      "loss": 2.8153,
+      "step": 1251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029232787398684974,
+      "loss": 2.5763,
+      "step": 1252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029230940189791256,
+      "loss": 2.8523,
+      "step": 1253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029229090818323316,
+      "loss": 3.1512,
+      "step": 1254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029227239284562186,
+      "loss": 3.1677,
+      "step": 1255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029225385588789243,
+      "loss": 2.713,
+      "step": 1256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002922352973128617,
+      "loss": 2.4122,
+      "step": 1257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029221671712334983,
+      "loss": 3.2627,
+      "step": 1258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002921981153221804,
+      "loss": 2.7293,
+      "step": 1259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002921794919121802,
+      "loss": 3.0195,
+      "step": 1260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002921608468961793,
+      "loss": 3.0319,
+      "step": 1261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000292142180277011,
+      "loss": 3.1447,
+      "step": 1262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000292123492057512,
+      "loss": 3.1118,
+      "step": 1263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029210478224052215,
+      "loss": 2.9213,
+      "step": 1264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029208605082888466,
+      "loss": 3.05,
+      "step": 1265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000292067297825446,
+      "loss": 3.1351,
+      "step": 1266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000292048523233056,
+      "loss": 1.2668,
+      "step": 1267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029202972705456767,
+      "loss": 3.059,
+      "step": 1268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002920109092928373,
+      "loss": 2.9862,
+      "step": 1269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029199206995072444,
+      "loss": 2.8356,
+      "step": 1270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002919732090310921,
+      "loss": 3.2456,
+      "step": 1271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029195432653680636,
+      "loss": 2.0247,
+      "step": 1272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002919354224707366,
+      "loss": 2.651,
+      "step": 1273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002919164968357557,
+      "loss": 3.2466,
+      "step": 1274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002918975496347395,
+      "loss": 2.5936,
+      "step": 1275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002918785808705674,
+      "loss": 3.8831,
+      "step": 1276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002918595905461218,
+      "loss": 3.0311,
+      "step": 1277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029184057866428866,
+      "loss": 3.0483,
+      "step": 1278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000291821545227957,
+      "loss": 2.8901,
+      "step": 1279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029180249024001923,
+      "loss": 3.0802,
+      "step": 1280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000291783413703371,
+      "loss": 3.3638,
+      "step": 1281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029176431562091115,
+      "loss": 2.6086,
+      "step": 1282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000291745195995542,
+      "loss": 2.3782,
+      "step": 1283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029172605483016896,
+      "loss": 3.1542,
+      "step": 1284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002917068921277008,
+      "loss": 2.6161,
+      "step": 1285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002916877078910495,
+      "loss": 2.9369,
+      "step": 1286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029166850212313043,
+      "loss": 2.4181,
+      "step": 1287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002916492748268621,
+      "loss": 3.0555,
+      "step": 1288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002916300260051663,
+      "loss": 2.6677,
+      "step": 1289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029161075566096824,
+      "loss": 2.9752,
+      "step": 1290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002915914637971962,
+      "loss": 2.9449,
+      "step": 1291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002915721504167819,
+      "loss": 2.8575,
+      "step": 1292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029155281552266013,
+      "loss": 2.5486,
+      "step": 1293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029153345911776926,
+      "loss": 2.5191,
+      "step": 1294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029151408120505065,
+      "loss": 3.2772,
+      "step": 1295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000291494681787449,
+      "loss": 2.5917,
+      "step": 1296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002914752608679123,
+      "loss": 2.5799,
+      "step": 1297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002914558184493919,
+      "loss": 2.4759,
+      "step": 1298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002914363545348422,
+      "loss": 2.9397,
+      "step": 1299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002914168691272211,
+      "loss": 2.9965,
+      "step": 1300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029139736222948956,
+      "loss": 3.8665,
+      "step": 1301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000291377833844612,
+      "loss": 2.7673,
+      "step": 1302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029135828397555593,
+      "loss": 2.507,
+      "step": 1303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002913387126252922,
+      "loss": 3.0463,
+      "step": 1304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029131911979679505,
+      "loss": 2.9436,
+      "step": 1305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029129950549304176,
+      "loss": 2.9717,
+      "step": 1306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029127986971701295,
+      "loss": 2.4895,
+      "step": 1307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002912602124716926,
+      "loss": 2.5054,
+      "step": 1308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029124053376006784,
+      "loss": 2.8732,
+      "step": 1309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002912208335851291,
+      "loss": 3.0007,
+      "step": 1310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002912011119498701,
+      "loss": 2.9969,
+      "step": 1311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002911813688572878,
+      "loss": 3.0209,
+      "step": 1312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002911616043103824,
+      "loss": 2.5019,
+      "step": 1313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029114181831215737,
+      "loss": 3.2543,
+      "step": 1314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002911220108656195,
+      "loss": 2.9928,
+      "step": 1315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029110218197377866,
+      "loss": 2.7887,
+      "step": 1316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029108233163964827,
+      "loss": 2.8165,
+      "step": 1317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002910624598662447,
+      "loss": 2.7713,
+      "step": 1318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002910425666565878,
+      "loss": 2.2886,
+      "step": 1319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029102265201370056,
+      "loss": 2.753,
+      "step": 1320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002910027159406093,
+      "loss": 2.5046,
+      "step": 1321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002909827584403436,
+      "loss": 2.6269,
+      "step": 1322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002909627795159361,
+      "loss": 1.7705,
+      "step": 1323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029094277917042304,
+      "loss": 2.2294,
+      "step": 1324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002909227574068436,
+      "loss": 3.2198,
+      "step": 1325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029090271422824046,
+      "loss": 2.7019,
+      "step": 1326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002908826496376594,
+      "loss": 3.289,
+      "step": 1327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029086256363814936,
+      "loss": 2.7276,
+      "step": 1328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002908424562327628,
+      "loss": 2.7087,
+      "step": 1329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002908223274245553,
+      "loss": 3.0056,
+      "step": 1330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029080217721658564,
+      "loss": 2.2261,
+      "step": 1331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029078200561191594,
+      "loss": 2.8277,
+      "step": 1332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002907618126136115,
+      "loss": 3.1967,
+      "step": 1333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029074159822474097,
+      "loss": 2.7227,
+      "step": 1334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000290721362448376,
+      "loss": 3.1715,
+      "step": 1335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000290701105287592,
+      "loss": 2.4188,
+      "step": 1336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029068082674546704,
+      "loss": 3.1733,
+      "step": 1337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002906605268250828,
+      "loss": 2.8485,
+      "step": 1338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029064020552952406,
+      "loss": 2.6811,
+      "step": 1339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000290619862861879,
+      "loss": 2.7628,
+      "step": 1340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002905994988252389,
+      "loss": 2.9196,
+      "step": 1341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002905791134226982,
+      "loss": 2.8751,
+      "step": 1342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029055870665735503,
+      "loss": 2.6547,
+      "step": 1343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002905382785323102,
+      "loss": 2.9415,
+      "step": 1344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029051782905066813,
+      "loss": 2.7883,
+      "step": 1345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029049735821553636,
+      "loss": 2.8978,
+      "step": 1346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029047686603002565,
+      "loss": 2.5401,
+      "step": 1347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002904563524972501,
+      "loss": 3.4326,
+      "step": 1348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002904358176203271,
+      "loss": 2.8307,
+      "step": 1349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000290415261402377,
+      "loss": 3.0873,
+      "step": 1350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029039468384652376,
+      "loss": 2.8704,
+      "step": 1351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002903740849558942,
+      "loss": 2.7649,
+      "step": 1352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002903534647336188,
+      "loss": 2.9992,
+      "step": 1353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002903328231828309,
+      "loss": 2.8322,
+      "step": 1354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002903121603066674,
+      "loss": 3.2072,
+      "step": 1355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029029147610826813,
+      "loss": 1.6833,
+      "step": 1356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029027077059077643,
+      "loss": 2.6856,
+      "step": 1357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002902500437573388,
+      "loss": 2.6856,
+      "step": 1358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029022929561110485,
+      "loss": 2.8422,
+      "step": 1359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002902085261552275,
+      "loss": 1.7255,
+      "step": 1360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029018773539286305,
+      "loss": 2.8643,
+      "step": 1361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029016692332717086,
+      "loss": 3.1167,
+      "step": 1362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029014608996131365,
+      "loss": 3.3238,
+      "step": 1363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002901252352984572,
+      "loss": 2.9994,
+      "step": 1364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002901043593417708,
+      "loss": 2.6468,
+      "step": 1365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002900834620944267,
+      "loss": 2.6698,
+      "step": 1366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029006254355960054,
+      "loss": 3.2307,
+      "step": 1367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002900416037404712,
+      "loss": 2.4091,
+      "step": 1368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029002064264022065,
+      "loss": 2.823,
+      "step": 1369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002899996602620343,
+      "loss": 3.2216,
+      "step": 1370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002899786566091007,
+      "loss": 3.8017,
+      "step": 1371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002899576316846115,
+      "loss": 2.6455,
+      "step": 1372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002899365854917618,
+      "loss": 2.6638,
+      "step": 1373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028991551803374987,
+      "loss": 3.0439,
+      "step": 1374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002898944293137771,
+      "loss": 2.5071,
+      "step": 1375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028987331933504827,
+      "loss": 2.9293,
+      "step": 1376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002898521881007713,
+      "loss": 2.7319,
+      "step": 1377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028983103561415724,
+      "loss": 3.0585,
+      "step": 1378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002898098618784206,
+      "loss": 2.858,
+      "step": 1379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028978866689677904,
+      "loss": 2.9132,
+      "step": 1380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028976745067245326,
+      "loss": 3.4595,
+      "step": 1381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002897462132086674,
+      "loss": 2.7366,
+      "step": 1382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028972495450864887,
+      "loss": 2.972,
+      "step": 1383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002897036745756281,
+      "loss": 2.795,
+      "step": 1384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002896823734128388,
+      "loss": 3.4003,
+      "step": 1385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028966105102351807,
+      "loss": 2.7974,
+      "step": 1386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028963970741090607,
+      "loss": 3.3983,
+      "step": 1387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002896183425782462,
+      "loss": 2.7854,
+      "step": 1388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002895969565287852,
+      "loss": 3.2509,
+      "step": 1389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002895755492657729,
+      "loss": 2.8307,
+      "step": 1390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028955412079246243,
+      "loss": 2.946,
+      "step": 1391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002895326711121101,
+      "loss": 3.1435,
+      "step": 1392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028951120022797553,
+      "loss": 2.5032,
+      "step": 1393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028948970814332136,
+      "loss": 2.7005,
+      "step": 1394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002894681948614137,
+      "loss": 2.6485,
+      "step": 1395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002894466603855218,
+      "loss": 3.0109,
+      "step": 1396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028942510471891805,
+      "loss": 3.1593,
+      "step": 1397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028940352786487804,
+      "loss": 2.6735,
+      "step": 1398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002893819298266807,
+      "loss": 3.212,
+      "step": 1399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002893603106076082,
+      "loss": 2.3761,
+      "step": 1400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002893386702109458,
+      "loss": 2.6149,
+      "step": 1401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028931700863998205,
+      "loss": 2.4885,
+      "step": 1402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002892953258980086,
+      "loss": 2.4714,
+      "step": 1403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028927362198832065,
+      "loss": 2.9325,
+      "step": 1404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028925189691421616,
+      "loss": 2.6671,
+      "step": 1405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028923015067899664,
+      "loss": 2.445,
+      "step": 1406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002892083832859667,
+      "loss": 2.8811,
+      "step": 1407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028918659473843415,
+      "loss": 2.722,
+      "step": 1408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002891647850397101,
+      "loss": 2.6473,
+      "step": 1409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002891429541931088,
+      "loss": 3.6401,
+      "step": 1410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002891211022019477,
+      "loss": 2.4258,
+      "step": 1411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002890992290695474,
+      "loss": 2.5296,
+      "step": 1412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028907733479923195,
+      "loss": 2.2091,
+      "step": 1413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028905541939432843,
+      "loss": 3.4065,
+      "step": 1414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028903348285816714,
+      "loss": 2.9714,
+      "step": 1415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002890115251940816,
+      "loss": 3.0483,
+      "step": 1416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002889895464054086,
+      "loss": 2.7581,
+      "step": 1417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002889675464954882,
+      "loss": 2.301,
+      "step": 1418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028894552546766334,
+      "loss": 2.7758,
+      "step": 1419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002889234833252805,
+      "loss": 1.6002,
+      "step": 1420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002889014200716893,
+      "loss": 3.1518,
+      "step": 1421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002888793357102425,
+      "loss": 2.7203,
+      "step": 1422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028885723024429623,
+      "loss": 3.4153,
+      "step": 1423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028883510367720943,
+      "loss": 2.9781,
+      "step": 1424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028881295601234475,
+      "loss": 2.7877,
+      "step": 1425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028879078725306783,
+      "loss": 2.5964,
+      "step": 1426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002887685974027473,
+      "loss": 2.5939,
+      "step": 1427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028874638646475526,
+      "loss": 3.2478,
+      "step": 1428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002887241544424671,
+      "loss": 3.3171,
+      "step": 1429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000288701901339261,
+      "loss": 2.9002,
+      "step": 1430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002886796271585189,
+      "loss": 3.2985,
+      "step": 1431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002886573319036254,
+      "loss": 3.1968,
+      "step": 1432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028863501557796863,
+      "loss": 2.6879,
+      "step": 1433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028861267818493995,
+      "loss": 2.6507,
+      "step": 1434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002885903197279337,
+      "loss": 2.8532,
+      "step": 1435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002885679402103475,
+      "loss": 2.4604,
+      "step": 1436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028854553963558234,
+      "loss": 2.4265,
+      "step": 1437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002885231180070422,
+      "loss": 3.1334,
+      "step": 1438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002885006753281343,
+      "loss": 3.2514,
+      "step": 1439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002884782116022691,
+      "loss": 2.8261,
+      "step": 1440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028845572683286035,
+      "loss": 2.8133,
+      "step": 1441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002884332210233248,
+      "loss": 2.8182,
+      "step": 1442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028841069417708253,
+      "loss": 2.7518,
+      "step": 1443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002883881462975568,
+      "loss": 2.7415,
+      "step": 1444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028836557738817396,
+      "loss": 2.3338,
+      "step": 1445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028834298745236377,
+      "loss": 2.983,
+      "step": 1446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028832037649355904,
+      "loss": 3.1829,
+      "step": 1447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002882977445151957,
+      "loss": 2.5463,
+      "step": 1448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002882750915207131,
+      "loss": 3.2177,
+      "step": 1449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028825241751355353,
+      "loss": 2.7496,
+      "step": 1450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028822972249716266,
+      "loss": 2.6318,
+      "step": 1451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002882070064749893,
+      "loss": 3.0836,
+      "step": 1452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028818426945048545,
+      "loss": 1.2672,
+      "step": 1453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002881615114271062,
+      "loss": 2.4442,
+      "step": 1454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028813873240831,
+      "loss": 3.0772,
+      "step": 1455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002881159323975585,
+      "loss": 1.5351,
+      "step": 1456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028809311139831625,
+      "loss": 2.0601,
+      "step": 1457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028807026941405137,
+      "loss": 2.7945,
+      "step": 1458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002880474064482349,
+      "loss": 2.9915,
+      "step": 1459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028802452250434123,
+      "loss": 3.1639,
+      "step": 1460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002880016175858478,
+      "loss": 2.9438,
+      "step": 1461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002879786916962353,
+      "loss": 2.4873,
+      "step": 1462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028795574483898775,
+      "loss": 2.9444,
+      "step": 1463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000287932777017592,
+      "loss": 2.9066,
+      "step": 1464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002879097882355385,
+      "loss": 2.5198,
+      "step": 1465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002878867784963206,
+      "loss": 3.3607,
+      "step": 1466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002878637478034349,
+      "loss": 2.9574,
+      "step": 1467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028784069616038133,
+      "loss": 2.8139,
+      "step": 1468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002878176235706627,
+      "loss": 2.7998,
+      "step": 1469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028779453003778543,
+      "loss": 3.167,
+      "step": 1470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002877714155652586,
+      "loss": 2.4707,
+      "step": 1471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028774828015659497,
+      "loss": 2.4239,
+      "step": 1472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002877251238153102,
+      "loss": 3.0156,
+      "step": 1473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028770194654492313,
+      "loss": 3.4245,
+      "step": 1474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028767874834895586,
+      "loss": 1.6901,
+      "step": 1475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028765552923093376,
+      "loss": 2.7008,
+      "step": 1476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028763228919438514,
+      "loss": 2.8593,
+      "step": 1477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028760902824284174,
+      "loss": 3.2717,
+      "step": 1478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028758574637983823,
+      "loss": 3.1732,
+      "step": 1479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002875624436089127,
+      "loss": 2.7367,
+      "step": 1480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028753911993360624,
+      "loss": 2.8167,
+      "step": 1481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028751577535746325,
+      "loss": 2.4312,
+      "step": 1482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028749240988403115,
+      "loss": 3.0547,
+      "step": 1483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002874690235168607,
+      "loss": 2.9509,
+      "step": 1484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002874456162595057,
+      "loss": 2.9227,
+      "step": 1485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002874221881155233,
+      "loss": 1.324,
+      "step": 1486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002873987390884735,
+      "loss": 2.719,
+      "step": 1487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028737526918191984,
+      "loss": 3.3642,
+      "step": 1488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002873517783994289,
+      "loss": 2.4117,
+      "step": 1489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002873282667445703,
+      "loss": 3.0139,
+      "step": 1490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002873047342209169,
+      "loss": 2.7026,
+      "step": 1491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028728118083204494,
+      "loss": 3.3125,
+      "step": 1492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002872576065815335,
+      "loss": 2.4574,
+      "step": 1493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002872340114729651,
+      "loss": 2.886,
+      "step": 1494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028721039550992525,
+      "loss": 3.3351,
+      "step": 1495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028718675869600276,
+      "loss": 2.618,
+      "step": 1496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002871631010347895,
+      "loss": 2.6006,
+      "step": 1497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028713942252988055,
+      "loss": 3.5669,
+      "step": 1498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028711572318487424,
+      "loss": 2.3894,
+      "step": 1499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002870920030033719,
+      "loss": 3.1315,
+      "step": 1500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002870682619889781,
+      "loss": 3.0599,
+      "step": 1501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028704450014530074,
+      "loss": 2.5158,
+      "step": 1502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002870207174759506,
+      "loss": 3.3198,
+      "step": 1503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002869969139845418,
+      "loss": 3.9574,
+      "step": 1504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028697308967469155,
+      "loss": 2.969,
+      "step": 1505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002869492445500203,
+      "loss": 3.1622,
+      "step": 1506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028692537861415164,
+      "loss": 2.4012,
+      "step": 1507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028690149187071234,
+      "loss": 2.8111,
+      "step": 1508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002868775843233322,
+      "loss": 1.4157,
+      "step": 1509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002868536559756443,
+      "loss": 2.6418,
+      "step": 1510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028682970683128486,
+      "loss": 2.9318,
+      "step": 1511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028680573689389335,
+      "loss": 3.8134,
+      "step": 1512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002867817461671122,
+      "loss": 2.5078,
+      "step": 1513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028675773465458717,
+      "loss": 2.4321,
+      "step": 1514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002867337023599671,
+      "loss": 2.9863,
+      "step": 1515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000286709649286904,
+      "loss": 3.1361,
+      "step": 1516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002866855754390531,
+      "loss": 2.713,
+      "step": 1517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002866614808200726,
+      "loss": 2.9623,
+      "step": 1518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028663736543362413,
+      "loss": 3.8676,
+      "step": 1519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028661322928337225,
+      "loss": 2.7506,
+      "step": 1520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028658907237298474,
+      "loss": 2.7961,
+      "step": 1521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028656489470613264,
+      "loss": 3.2506,
+      "step": 1522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028654069628649,
+      "loss": 3.2202,
+      "step": 1523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002865164771177341,
+      "loss": 2.7636,
+      "step": 1524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002864922372035453,
+      "loss": 2.2585,
+      "step": 1525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002864679765476072,
+      "loss": 2.4861,
+      "step": 1526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028644369515360655,
+      "loss": 2.7023,
+      "step": 1527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028641939302523323,
+      "loss": 2.7692,
+      "step": 1528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028639507016618026,
+      "loss": 2.6201,
+      "step": 1529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028637072658014364,
+      "loss": 2.1704,
+      "step": 1530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028634636227082297,
+      "loss": 2.4193,
+      "step": 1531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002863219772419205,
+      "loss": 2.7625,
+      "step": 1532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028629757149714196,
+      "loss": 2.5334,
+      "step": 1533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002862731450401961,
+      "loss": 2.9834,
+      "step": 1534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028624869787479475,
+      "loss": 2.2152,
+      "step": 1535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002862242300046531,
+      "loss": 2.4714,
+      "step": 1536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002861997414334893,
+      "loss": 2.8994,
+      "step": 1537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002861752321650247,
+      "loss": 2.5196,
+      "step": 1538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002861507022029838,
+      "loss": 3.1982,
+      "step": 1539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002861261515510942,
+      "loss": 3.007,
+      "step": 1540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028610158021308675,
+      "loss": 2.7804,
+      "step": 1541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028607698819269533,
+      "loss": 2.9038,
+      "step": 1542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002860523754936571,
+      "loss": 2.9135,
+      "step": 1543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002860277421197122,
+      "loss": 2.4818,
+      "step": 1544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028600308807460404,
+      "loss": 3.0328,
+      "step": 1545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000285978413362079,
+      "loss": 2.6264,
+      "step": 1546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002859537179858868,
+      "loss": 2.7555,
+      "step": 1547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002859290019497803,
+      "loss": 2.0829,
+      "step": 1548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002859042652575153,
+      "loss": 3.1349,
+      "step": 1549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028587950791285094,
+      "loss": 2.5093,
+      "step": 1550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002858547299195494,
+      "loss": 3.2456,
+      "step": 1551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002858299312813759,
+      "loss": 2.7806,
+      "step": 1552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028580511200209903,
+      "loss": 2.4545,
+      "step": 1553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002857802720854904,
+      "loss": 3.1691,
+      "step": 1554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028575541153532476,
+      "loss": 2.9165,
+      "step": 1555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002857305303553799,
+      "loss": 2.7683,
+      "step": 1556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002857056285494369,
+      "loss": 3.0418,
+      "step": 1557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028568070612127995,
+      "loss": 2.4001,
+      "step": 1558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002856557630746962,
+      "loss": 2.6167,
+      "step": 1559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028563079941347627,
+      "loss": 2.7649,
+      "step": 1560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002856058151414135,
+      "loss": 3.3088,
+      "step": 1561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028558081026230464,
+      "loss": 2.4691,
+      "step": 1562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002855557847799496,
+      "loss": 2.8715,
+      "step": 1563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028553073869815124,
+      "loss": 2.9668,
+      "step": 1564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002855056720207155,
+      "loss": 2.9269,
+      "step": 1565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002854805847514519,
+      "loss": 2.872,
+      "step": 1566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002854554768941726,
+      "loss": 2.52,
+      "step": 1567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028543034845269295,
+      "loss": 2.387,
+      "step": 1568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028540519943083173,
+      "loss": 2.5692,
+      "step": 1569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028538002983241053,
+      "loss": 1.045,
+      "step": 1570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002853548396612543,
+      "loss": 2.7071,
+      "step": 1571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028532962892119086,
+      "loss": 2.9538,
+      "step": 1572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002853043976160514,
+      "loss": 3.8731,
+      "step": 1573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028527914574967024,
+      "loss": 2.5824,
+      "step": 1574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028525387332588455,
+      "loss": 3.0118,
+      "step": 1575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002852285803485349,
+      "loss": 3.4665,
+      "step": 1576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028520326682146487,
+      "loss": 3.4535,
+      "step": 1577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002851779327485211,
+      "loss": 2.9645,
+      "step": 1578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002851525781335535,
+      "loss": 3.4764,
+      "step": 1579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028512720298041506,
+      "loss": 2.9746,
+      "step": 1580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002851018072929618,
+      "loss": 2.9481,
+      "step": 1581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000285076391075053,
+      "loss": 3.4264,
+      "step": 1582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028505095433055085,
+      "loss": 3.1196,
+      "step": 1583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002850254970633209,
+      "loss": 2.5002,
+      "step": 1584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028500001927723166,
+      "loss": 3.1492,
+      "step": 1585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028497452097615475,
+      "loss": 2.7449,
+      "step": 1586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002849490021639651,
+      "loss": 3.4662,
+      "step": 1587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028492346284454055,
+      "loss": 2.1841,
+      "step": 1588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002848979030217621,
+      "loss": 2.6459,
+      "step": 1589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028487232269951395,
+      "loss": 2.859,
+      "step": 1590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028484672188168334,
+      "loss": 1.2763,
+      "step": 1591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028482110057216067,
+      "loss": 2.873,
+      "step": 1592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002847954587748393,
+      "loss": 2.4206,
+      "step": 1593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000284769796493616,
+      "loss": 2.6499,
+      "step": 1594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002847441137323904,
+      "loss": 3.0372,
+      "step": 1595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002847184104950653,
+      "loss": 2.9812,
+      "step": 1596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002846926867855467,
+      "loss": 2.7677,
+      "step": 1597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002846669426077436,
+      "loss": 3.3403,
+      "step": 1598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028464117796556815,
+      "loss": 2.7404,
+      "step": 1599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002846153928629357,
+      "loss": 2.7577,
+      "step": 1600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028458958730376456,
+      "loss": 2.8717,
+      "step": 1601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028456376129197623,
+      "loss": 2.5266,
+      "step": 1602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002845379148314953,
+      "loss": 2.9704,
+      "step": 1603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028451204792624945,
+      "loss": 2.7067,
+      "step": 1604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002844861605801695,
+      "loss": 3.0089,
+      "step": 1605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028446025279718943,
+      "loss": 3.1371,
+      "step": 1606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002844343245812462,
+      "loss": 3.2279,
+      "step": 1607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028440837593627993,
+      "loss": 2.6791,
+      "step": 1608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028438240686623386,
+      "loss": 3.0833,
+      "step": 1609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028435641737505437,
+      "loss": 3.3163,
+      "step": 1610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002843304074666908,
+      "loss": 3.0748,
+      "step": 1611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028430437714509584,
+      "loss": 1.5057,
+      "step": 1612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002842783264142249,
+      "loss": 2.8021,
+      "step": 1613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028425225527803697,
+      "loss": 2.9624,
+      "step": 1614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028422616374049376,
+      "loss": 2.6968,
+      "step": 1615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028420005180556024,
+      "loss": 2.4925,
+      "step": 1616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002841739194772044,
+      "loss": 3.2939,
+      "step": 1617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002841477667593975,
+      "loss": 2.4063,
+      "step": 1618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002841215936561137,
+      "loss": 2.879,
+      "step": 1619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028409540017133037,
+      "loss": 2.7971,
+      "step": 1620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002840691863090279,
+      "loss": 2.6298,
+      "step": 1621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002840429520731899,
+      "loss": 3.0564,
+      "step": 1622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028401669746780297,
+      "loss": 2.9783,
+      "step": 1623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028399042249685675,
+      "loss": 3.5067,
+      "step": 1624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028396412716434415,
+      "loss": 2.9629,
+      "step": 1625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002839378114742611,
+      "loss": 3.0511,
+      "step": 1626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002839114754306066,
+      "loss": 2.3002,
+      "step": 1627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028388511903738265,
+      "loss": 2.7856,
+      "step": 1628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028385874229859456,
+      "loss": 2.5709,
+      "step": 1629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002838323452182505,
+      "loss": 3.2633,
+      "step": 1630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000283805927800362,
+      "loss": 2.5417,
+      "step": 1631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002837794900489434,
+      "loss": 2.9476,
+      "step": 1632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002837530319680123,
+      "loss": 3.4276,
+      "step": 1633
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028372655356158935,
+      "loss": 2.9859,
+      "step": 1634
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002837000548336983,
+      "loss": 2.5053,
+      "step": 1635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002836735357883659,
+      "loss": 3.0151,
+      "step": 1636
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002836469964296221,
+      "loss": 2.839,
+      "step": 1637
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028362043676149996,
+      "loss": 2.464,
+      "step": 1638
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028359385678803544,
+      "loss": 2.8062,
+      "step": 1639
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028356725651326785,
+      "loss": 2.9828,
+      "step": 1640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002835406359412393,
+      "loss": 2.7528,
+      "step": 1641
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002835139950759953,
+      "loss": 2.4277,
+      "step": 1642
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002834873339215841,
+      "loss": 2.6994,
+      "step": 1643
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028346065248205723,
+      "loss": 2.9084,
+      "step": 1644
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002834339507614694,
+      "loss": 2.624,
+      "step": 1645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002834072287638782,
+      "loss": 2.8712,
+      "step": 1646
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002833804864933444,
+      "loss": 3.5638,
+      "step": 1647
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002833537239539317,
+      "loss": 2.2204,
+      "step": 1648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002833269411497072,
+      "loss": 3.1321,
+      "step": 1649
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002833001380847408,
+      "loss": 3.061,
+      "step": 1650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002832733147631056,
+      "loss": 3.2409,
+      "step": 1651
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002832464711888777,
+      "loss": 2.8929,
+      "step": 1652
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002832196073661364,
+      "loss": 3.2157,
+      "step": 1653
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028319272329896394,
+      "loss": 2.5924,
+      "step": 1654
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002831658189914458,
+      "loss": 3.2585,
+      "step": 1655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002831388944476703,
+      "loss": 3.2425,
+      "step": 1656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000283111949671729,
+      "loss": 3.569,
+      "step": 1657
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002830849846677166,
+      "loss": 2.7605,
+      "step": 1658
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028305799943973065,
+      "loss": 2.6186,
+      "step": 1659
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000283030993991872,
+      "loss": 2.2171,
+      "step": 1660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028300396832824445,
+      "loss": 3.2659,
+      "step": 1661
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002829769224529549,
+      "loss": 3.1139,
+      "step": 1662
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002829498563701133,
+      "loss": 2.82,
+      "step": 1663
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028292277008383265,
+      "loss": 2.1501,
+      "step": 1664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002828956635982291,
+      "loss": 2.5171,
+      "step": 1665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028286853691742185,
+      "loss": 2.5955,
+      "step": 1666
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028284139004553313,
+      "loss": 3.5619,
+      "step": 1667
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002828142229866883,
+      "loss": 2.7044,
+      "step": 1668
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028278703574501567,
+      "loss": 3.019,
+      "step": 1669
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002827598283246467,
+      "loss": 2.2031,
+      "step": 1670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000282732600729716,
+      "loss": 3.4973,
+      "step": 1671
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000282705352964361,
+      "loss": 2.9985,
+      "step": 1672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002826780850327224,
+      "loss": 2.6119,
+      "step": 1673
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028265079693894405,
+      "loss": 2.625,
+      "step": 1674
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028262348868717256,
+      "loss": 2.9428,
+      "step": 1675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002825961602815578,
+      "loss": 3.4969,
+      "step": 1676
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028256881172625275,
+      "loss": 2.5652,
+      "step": 1677
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002825414430254133,
+      "loss": 2.866,
+      "step": 1678
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028251405418319855,
+      "loss": 3.3434,
+      "step": 1679
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002824866452037705,
+      "loss": 3.0887,
+      "step": 1680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002824592160912943,
+      "loss": 3.0237,
+      "step": 1681
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002824317668499382,
+      "loss": 2.8199,
+      "step": 1682
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028240429748387346,
+      "loss": 2.6235,
+      "step": 1683
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028237680799727443,
+      "loss": 2.574,
+      "step": 1684
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002823492983943184,
+      "loss": 2.7497,
+      "step": 1685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028232176867918593,
+      "loss": 2.7217,
+      "step": 1686
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028229421885606044,
+      "loss": 2.837,
+      "step": 1687
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028226664892912846,
+      "loss": 3.1218,
+      "step": 1688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028223905890257964,
+      "loss": 2.791,
+      "step": 1689
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028221144878060663,
+      "loss": 2.9351,
+      "step": 1690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028218381856740513,
+      "loss": 2.6271,
+      "step": 1691
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002821561682671739,
+      "loss": 2.7702,
+      "step": 1692
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002821284978841148,
+      "loss": 2.4552,
+      "step": 1693
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002821008074224326,
+      "loss": 2.7071,
+      "step": 1694
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028207309688633536,
+      "loss": 2.4783,
+      "step": 1695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00028204536628003395,
+      "loss": 2.9919,
+      "step": 1696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028201761560774243,
+      "loss": 2.1138,
+      "step": 1697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002819898448736779,
+      "loss": 2.3207,
+      "step": 1698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002819620540820604,
+      "loss": 2.5343,
+      "step": 1699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002819342432371132,
+      "loss": 2.8091,
+      "step": 1700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002819064123430624,
+      "loss": 3.1918,
+      "step": 1701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028187856140413737,
+      "loss": 2.3774,
+      "step": 1702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002818506904245704,
+      "loss": 2.2052,
+      "step": 1703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002818227994085967,
+      "loss": 2.8676,
+      "step": 1704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002817948883604549,
+      "loss": 1.5792,
+      "step": 1705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002817669572843862,
+      "loss": 2.4788,
+      "step": 1706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002817390061846353,
+      "loss": 2.6976,
+      "step": 1707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002817110350654496,
+      "loss": 3.2051,
+      "step": 1708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028168304393107977,
+      "loss": 2.8256,
+      "step": 1709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002816550327857793,
+      "loss": 2.3063,
+      "step": 1710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002816270016338049,
+      "loss": 2.504,
+      "step": 1711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028159895047941635,
+      "loss": 3.124,
+      "step": 1712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028157087932687626,
+      "loss": 2.988,
+      "step": 1713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028154278818045045,
+      "loss": 2.7108,
+      "step": 1714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002815146770444077,
+      "loss": 3.0714,
+      "step": 1715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002814865459230199,
+      "loss": 2.8676,
+      "step": 1716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028145839482056193,
+      "loss": 3.2876,
+      "step": 1717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028143022374131166,
+      "loss": 2.6438,
+      "step": 1718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002814020326895501,
+      "loss": 2.5327,
+      "step": 1719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002813738216695613,
+      "loss": 3.163,
+      "step": 1720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028134559068563215,
+      "loss": 3.0143,
+      "step": 1721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028131733974205275,
+      "loss": 3.0996,
+      "step": 1722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002812890688431163,
+      "loss": 2.9542,
+      "step": 1723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028126077799311885,
+      "loss": 3.0219,
+      "step": 1724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028123246719635953,
+      "loss": 2.9243,
+      "step": 1725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002812041364571405,
+      "loss": 2.7651,
+      "step": 1726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028117578577976715,
+      "loss": 3.3881,
+      "step": 1727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028114741516854753,
+      "loss": 3.3661,
+      "step": 1728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000281119024627793,
+      "loss": 2.8772,
+      "step": 1729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002810906141618179,
+      "loss": 2.4745,
+      "step": 1730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002810621837749396,
+      "loss": 2.7704,
+      "step": 1731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002810337334714783,
+      "loss": 2.8591,
+      "step": 1732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028100526325575755,
+      "loss": 2.8053,
+      "step": 1733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002809767731321037,
+      "loss": 3.2892,
+      "step": 1734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028094826310484617,
+      "loss": 2.9794,
+      "step": 1735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028091973317831745,
+      "loss": 2.5736,
+      "step": 1736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028089118335685305,
+      "loss": 3.1642,
+      "step": 1737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002808626136447915,
+      "loss": 2.8923,
+      "step": 1738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028083402404647427,
+      "loss": 2.8627,
+      "step": 1739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028080541456624593,
+      "loss": 2.6921,
+      "step": 1740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002807767852084541,
+      "loss": 2.1573,
+      "step": 1741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028074813597744923,
+      "loss": 2.8498,
+      "step": 1742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028071946687758517,
+      "loss": 3.4125,
+      "step": 1743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028069077791321843,
+      "loss": 3.041,
+      "step": 1744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028066206908870865,
+      "loss": 1.9306,
+      "step": 1745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002806333404084186,
+      "loss": 2.6174,
+      "step": 1746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002806045918767139,
+      "loss": 2.933,
+      "step": 1747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002805758234979632,
+      "loss": 2.6326,
+      "step": 1748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002805470352765384,
+      "loss": 3.413,
+      "step": 1749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028051822721681405,
+      "loss": 2.3072,
+      "step": 1750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028048939932316804,
+      "loss": 2.9727,
+      "step": 1751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002804605515999811,
+      "loss": 2.6653,
+      "step": 1752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028043168405163693,
+      "loss": 3.0107,
+      "step": 1753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028040279668252243,
+      "loss": 2.6387,
+      "step": 1754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002803738894970274,
+      "loss": 1.9069,
+      "step": 1755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028034496249954455,
+      "loss": 3.1286,
+      "step": 1756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002803160156944699,
+      "loss": 2.3876,
+      "step": 1757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002802870490862021,
+      "loss": 3.15,
+      "step": 1758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002802580626791431,
+      "loss": 2.5482,
+      "step": 1759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028022905647769774,
+      "loss": 3.0949,
+      "step": 1760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028020003048627386,
+      "loss": 2.841,
+      "step": 1761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002801709847092824,
+      "loss": 2.4442,
+      "step": 1762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002801419191511371,
+      "loss": 2.3872,
+      "step": 1763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028011283381625503,
+      "loss": 2.5769,
+      "step": 1764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028008372870905597,
+      "loss": 3.1118,
+      "step": 1765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00028005460383396277,
+      "loss": 3.1104,
+      "step": 1766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002800254591954015,
+      "loss": 3.0876,
+      "step": 1767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027999629479780083,
+      "loss": 2.2485,
+      "step": 1768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027996711064559284,
+      "loss": 2.7303,
+      "step": 1769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002799379067432124,
+      "loss": 2.5195,
+      "step": 1770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027990868309509735,
+      "loss": 2.3152,
+      "step": 1771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002798794397056887,
+      "loss": 2.6641,
+      "step": 1772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002798501765794303,
+      "loss": 3.1556,
+      "step": 1773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027982089372076906,
+      "loss": 2.8871,
+      "step": 1774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002797915911341549,
+      "loss": 2.3635,
+      "step": 1775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002797622688240407,
+      "loss": 2.8916,
+      "step": 1776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002797329267948824,
+      "loss": 2.7969,
+      "step": 1777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027970356505113887,
+      "loss": 3.1197,
+      "step": 1778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027967418359727204,
+      "loss": 2.9563,
+      "step": 1779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027964478243774676,
+      "loss": 2.5167,
+      "step": 1780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002796153615770309,
+      "loss": 2.433,
+      "step": 1781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002795859210195954,
+      "loss": 2.956,
+      "step": 1782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027955646076991406,
+      "loss": 1.0836,
+      "step": 1783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002795269808324638,
+      "loss": 2.4257,
+      "step": 1784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027949748121172446,
+      "loss": 2.8556,
+      "step": 1785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027946796191217886,
+      "loss": 3.1397,
+      "step": 1786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027943842293831286,
+      "loss": 3.0536,
+      "step": 1787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002794088642946153,
+      "loss": 2.5367,
+      "step": 1788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000279379285985578,
+      "loss": 3.0281,
+      "step": 1789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002793496880156957,
+      "loss": 2.6895,
+      "step": 1790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002793200703894663,
+      "loss": 2.1203,
+      "step": 1791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027929043311139047,
+      "loss": 2.5684,
+      "step": 1792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027926077618597203,
+      "loss": 3.2601,
+      "step": 1793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002792310996177177,
+      "loss": 2.5907,
+      "step": 1794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027920140341113726,
+      "loss": 2.638,
+      "step": 1795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027917168757074333,
+      "loss": 2.2548,
+      "step": 1796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002791419521010518,
+      "loss": 3.3028,
+      "step": 1797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002791121970065812,
+      "loss": 2.6874,
+      "step": 1798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027908242229185323,
+      "loss": 2.8076,
+      "step": 1799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027905262796139264,
+      "loss": 3.083,
+      "step": 1800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027902281401972687,
+      "loss": 3.0804,
+      "step": 1801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027899298047138674,
+      "loss": 2.617,
+      "step": 1802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002789631273209057,
+      "loss": 2.5662,
+      "step": 1803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002789332545728203,
+      "loss": 3.3715,
+      "step": 1804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002789033622316702,
+      "loss": 2.4726,
+      "step": 1805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002788734503019979,
+      "loss": 2.5937,
+      "step": 1806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027884351878834884,
+      "loss": 2.3896,
+      "step": 1807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027881356769527153,
+      "loss": 2.7244,
+      "step": 1808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027878359702731745,
+      "loss": 2.4711,
+      "step": 1809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000278753606789041,
+      "loss": 2.979,
+      "step": 1810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027872359698499956,
+      "loss": 2.807,
+      "step": 1811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002786935676197536,
+      "loss": 3.4997,
+      "step": 1812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027866351869786635,
+      "loss": 2.5537,
+      "step": 1813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002786334502239042,
+      "loss": 3.0109,
+      "step": 1814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002786033622024364,
+      "loss": 2.696,
+      "step": 1815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027857325463803527,
+      "loss": 2.3994,
+      "step": 1816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000278543127535276,
+      "loss": 3.1983,
+      "step": 1817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002785129808987367,
+      "loss": 2.3572,
+      "step": 1818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027848281473299877,
+      "loss": 2.8653,
+      "step": 1819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027845262904264615,
+      "loss": 1.9555,
+      "step": 1820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000278422423832266,
+      "loss": 2.5846,
+      "step": 1821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027839219910644837,
+      "loss": 2.8598,
+      "step": 1822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027836195486978635,
+      "loss": 2.708,
+      "step": 1823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002783316911268759,
+      "loss": 2.163,
+      "step": 1824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027830140788231603,
+      "loss": 2.5978,
+      "step": 1825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002782711051407086,
+      "loss": 3.1485,
+      "step": 1826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002782407829066585,
+      "loss": 2.7813,
+      "step": 1827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002782104411847737,
+      "loss": 2.812,
+      "step": 1828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027818007997966487,
+      "loss": 2.7889,
+      "step": 1829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027814969929594585,
+      "loss": 3.1338,
+      "step": 1830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027811929913823335,
+      "loss": 2.7927,
+      "step": 1831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027808887951114714,
+      "loss": 2.8339,
+      "step": 1832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002780584404193098,
+      "loss": 2.5714,
+      "step": 1833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027802798186734695,
+      "loss": 2.2736,
+      "step": 1834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002779975038598872,
+      "loss": 2.5484,
+      "step": 1835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000277967006401562,
+      "loss": 3.1381,
+      "step": 1836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002779364894970059,
+      "loss": 2.7421,
+      "step": 1837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027790595315085626,
+      "loss": 2.8789,
+      "step": 1838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002778753973677536,
+      "loss": 2.6736,
+      "step": 1839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002778448221523411,
+      "loss": 2.4173,
+      "step": 1840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002778142275092652,
+      "loss": 3.0227,
+      "step": 1841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002777836134431751,
+      "loss": 3.1676,
+      "step": 1842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002777529799587229,
+      "loss": 2.5677,
+      "step": 1843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027772232706056396,
+      "loss": 3.3944,
+      "step": 1844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027769165475335624,
+      "loss": 2.6227,
+      "step": 1845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027766096304176075,
+      "loss": 2.9607,
+      "step": 1846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027763025193044163,
+      "loss": 2.3934,
+      "step": 1847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002775995214240657,
+      "loss": 1.9574,
+      "step": 1848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027756877152730294,
+      "loss": 2.7789,
+      "step": 1849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002775380022448261,
+      "loss": 2.7093,
+      "step": 1850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002775072135813111,
+      "loss": 2.4504,
+      "step": 1851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002774764055414366,
+      "loss": 2.9028,
+      "step": 1852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002774455781298842,
+      "loss": 3.3297,
+      "step": 1853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027741473135133867,
+      "loss": 2.728,
+      "step": 1854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002773838652104875,
+      "loss": 2.8007,
+      "step": 1855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027735297971202117,
+      "loss": 2.6511,
+      "step": 1856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002773220748606332,
+      "loss": 2.7912,
+      "step": 1857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002772911506610199,
+      "loss": 2.7278,
+      "step": 1858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027726020711788067,
+      "loss": 2.7221,
+      "step": 1859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002772292442359177,
+      "loss": 3.4392,
+      "step": 1860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002771982620198363,
+      "loss": 3.2679,
+      "step": 1861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027716726047434455,
+      "loss": 2.8555,
+      "step": 1862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002771362396041535,
+      "loss": 2.4852,
+      "step": 1863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002771051994139773,
+      "loss": 2.8298,
+      "step": 1864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027707413990853277,
+      "loss": 3.0852,
+      "step": 1865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002770430610925398,
+      "loss": 2.9334,
+      "step": 1866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027701196297072136,
+      "loss": 2.2346,
+      "step": 1867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000276980845547803,
+      "loss": 3.471,
+      "step": 1868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027694970882851365,
+      "loss": 2.8738,
+      "step": 1869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027691855281758475,
+      "loss": 2.4752,
+      "step": 1870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002768873775197509,
+      "loss": 3.037,
+      "step": 1871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027685618293974963,
+      "loss": 3.037,
+      "step": 1872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027682496908232134,
+      "loss": 2.5633,
+      "step": 1873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027679373595220933,
+      "loss": 2.4331,
+      "step": 1874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027676248355416,
+      "loss": 2.8821,
+      "step": 1875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002767312118929224,
+      "loss": 2.4716,
+      "step": 1876
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002766999209732487,
+      "loss": 3.1376,
+      "step": 1877
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027666861079989405,
+      "loss": 2.9328,
+      "step": 1878
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027663728137761634,
+      "loss": 2.9382,
+      "step": 1879
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027660593271117655,
+      "loss": 2.3653,
+      "step": 1880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027657456480533843,
+      "loss": 3.3081,
+      "step": 1881
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002765431776648688,
+      "loss": 2.6942,
+      "step": 1882
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027651177129453725,
+      "loss": 2.7526,
+      "step": 1883
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027648034569911653,
+      "loss": 3.2326,
+      "step": 1884
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027644890088338205,
+      "loss": 2.4458,
+      "step": 1885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002764174368521123,
+      "loss": 2.7188,
+      "step": 1886
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027638595361008855,
+      "loss": 3.3392,
+      "step": 1887
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002763544511620953,
+      "loss": 3.1855,
+      "step": 1888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002763229295129195,
+      "loss": 3.0585,
+      "step": 1889
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002762913886673514,
+      "loss": 2.7083,
+      "step": 1890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027625982863018405,
+      "loss": 3.2628,
+      "step": 1891
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027622824940621335,
+      "loss": 2.886,
+      "step": 1892
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027619665100023824,
+      "loss": 2.7576,
+      "step": 1893
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027616503341706044,
+      "loss": 2.797,
+      "step": 1894
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002761333966614846,
+      "loss": 2.3262,
+      "step": 1895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027610174073831845,
+      "loss": 1.4434,
+      "step": 1896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027607006565237243,
+      "loss": 2.8222,
+      "step": 1897
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027603837140846,
+      "loss": 2.4679,
+      "step": 1898
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027600665801139757,
+      "loss": 2.7907,
+      "step": 1899
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002759749254660043,
+      "loss": 3.2495,
+      "step": 1900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002759431737771024,
+      "loss": 3.939,
+      "step": 1901
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027591140294951696,
+      "loss": 2.8491,
+      "step": 1902
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002758796129880759,
+      "loss": 2.9337,
+      "step": 1903
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002758478038976102,
+      "loss": 2.7966,
+      "step": 1904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002758159756829537,
+      "loss": 2.9349,
+      "step": 1905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027578412834894297,
+      "loss": 2.7336,
+      "step": 1906
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002757522619004176,
+      "loss": 2.5295,
+      "step": 1907
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002757203763422203,
+      "loss": 2.8026,
+      "step": 1908
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027568847167919636,
+      "loss": 3.1405,
+      "step": 1909
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002756565479161941,
+      "loss": 2.7807,
+      "step": 1910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002756246050580648,
+      "loss": 3.0137,
+      "step": 1911
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027559264310966254,
+      "loss": 2.5007,
+      "step": 1912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027556066207584434,
+      "loss": 2.4823,
+      "step": 1913
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027552866196147016,
+      "loss": 3.1915,
+      "step": 1914
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002754966427714029,
+      "loss": 2.5463,
+      "step": 1915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002754646045105081,
+      "loss": 3.4208,
+      "step": 1916
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027543254718365455,
+      "loss": 2.8551,
+      "step": 1917
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002754004707957137,
+      "loss": 2.6444,
+      "step": 1918
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027536837535156007,
+      "loss": 2.9684,
+      "step": 1919
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002753362608560708,
+      "loss": 3.2062,
+      "step": 1920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002753041273141262,
+      "loss": 2.1716,
+      "step": 1921
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002752719747306094,
+      "loss": 3.3195,
+      "step": 1922
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027523980311040634,
+      "loss": 2.8434,
+      "step": 1923
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027520761245840595,
+      "loss": 2.8135,
+      "step": 1924
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027517540277949997,
+      "loss": 2.9507,
+      "step": 1925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027514317407858314,
+      "loss": 2.3539,
+      "step": 1926
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000275110926360553,
+      "loss": 2.51,
+      "step": 1927
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027507865963031,
+      "loss": 2.8821,
+      "step": 1928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027504637389275745,
+      "loss": 2.6807,
+      "step": 1929
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002750140691528016,
+      "loss": 2.258,
+      "step": 1930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002749817454153516,
+      "loss": 3.2914,
+      "step": 1931
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027494940268531947,
+      "loss": 2.6674,
+      "step": 1932
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027491704096762004,
+      "loss": 2.9512,
+      "step": 1933
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027488466026717116,
+      "loss": 2.9533,
+      "step": 1934
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002748522605888934,
+      "loss": 2.3834,
+      "step": 1935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002748198419377104,
+      "loss": 2.7788,
+      "step": 1936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027478740431854855,
+      "loss": 2.2699,
+      "step": 1937
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002747549477363371,
+      "loss": 2.7472,
+      "step": 1938
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002747224721960084,
+      "loss": 3.1555,
+      "step": 1939
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002746899777024974,
+      "loss": 2.8283,
+      "step": 1940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027465746426074214,
+      "loss": 2.9641,
+      "step": 1941
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027462493187568334,
+      "loss": 3.4759,
+      "step": 1942
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002745923805522648,
+      "loss": 3.5289,
+      "step": 1943
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027455981029543314,
+      "loss": 3.2446,
+      "step": 1944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027452722111013774,
+      "loss": 2.9145,
+      "step": 1945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000274494613001331,
+      "loss": 2.7698,
+      "step": 1946
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002744619859739681,
+      "loss": 2.7201,
+      "step": 1947
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002744293400330072,
+      "loss": 2.6265,
+      "step": 1948
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027439667518340923,
+      "loss": 2.6574,
+      "step": 1949
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027436399143013803,
+      "loss": 1.6757,
+      "step": 1950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027433128877816035,
+      "loss": 2.7713,
+      "step": 1951
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027429856723244577,
+      "loss": 3.57,
+      "step": 1952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027426582679796667,
+      "loss": 1.8158,
+      "step": 1953
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027423306747969845,
+      "loss": 3.042,
+      "step": 1954
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027420028928261934,
+      "loss": 2.5467,
+      "step": 1955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002741674922117103,
+      "loss": 3.6457,
+      "step": 1956
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002741346762719554,
+      "loss": 3.0943,
+      "step": 1957
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002741018414683414,
+      "loss": 3.2386,
+      "step": 1958
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002740689878058579,
+      "loss": 2.9595,
+      "step": 1959
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002740361152894976,
+      "loss": 2.6428,
+      "step": 1960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027400322392425564,
+      "loss": 2.7287,
+      "step": 1961
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002739703137151305,
+      "loss": 2.7793,
+      "step": 1962
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002739373846671232,
+      "loss": 3.4226,
+      "step": 1963
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027390443678523786,
+      "loss": 2.68,
+      "step": 1964
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027387147007448115,
+      "loss": 2.5998,
+      "step": 1965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002738384845398629,
+      "loss": 2.6664,
+      "step": 1966
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027380548018639576,
+      "loss": 2.9146,
+      "step": 1967
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000273772457019095,
+      "loss": 2.7133,
+      "step": 1968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027373941504297905,
+      "loss": 3.2795,
+      "step": 1969
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027370635426306894,
+      "loss": 2.5326,
+      "step": 1970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002736732746843888,
+      "loss": 3.1605,
+      "step": 1971
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027364017631196537,
+      "loss": 3.3058,
+      "step": 1972
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002736070591508285,
+      "loss": 2.686,
+      "step": 1973
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002735739232060107,
+      "loss": 2.85,
+      "step": 1974
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002735407684825474,
+      "loss": 3.3735,
+      "step": 1975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002735075949854769,
+      "loss": 2.8494,
+      "step": 1976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027347440271984034,
+      "loss": 3.0624,
+      "step": 1977
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002734411916906817,
+      "loss": 2.9708,
+      "step": 1978
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027340796190304786,
+      "loss": 2.6892,
+      "step": 1979
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027337471336198846,
+      "loss": 2.8335,
+      "step": 1980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002733414460725561,
+      "loss": 3.1105,
+      "step": 1981
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027330816003980616,
+      "loss": 3.0509,
+      "step": 1982
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002732748552687968,
+      "loss": 3.0136,
+      "step": 1983
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027324153176458925,
+      "loss": 3.0983,
+      "step": 1984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027320818953224725,
+      "loss": 2.5905,
+      "step": 1985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002731748285768378,
+      "loss": 2.5994,
+      "step": 1986
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002731414489034304,
+      "loss": 2.2726,
+      "step": 1987
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002731080505170975,
+      "loss": 2.898,
+      "step": 1988
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002730746334229145,
+      "loss": 2.5874,
+      "step": 1989
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002730411976259595,
+      "loss": 2.2972,
+      "step": 1990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002730077431313135,
+      "loss": 2.4029,
+      "step": 1991
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027297426994406037,
+      "loss": 1.4491,
+      "step": 1992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027294077806928673,
+      "loss": 2.7817,
+      "step": 1993
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027290726751208217,
+      "loss": 2.6873,
+      "step": 1994
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000272873738277539,
+      "loss": 3.1013,
+      "step": 1995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002728401903707525,
+      "loss": 3.0514,
+      "step": 1996
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002728066237968206,
+      "loss": 2.8791,
+      "step": 1997
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027277303856084427,
+      "loss": 2.7998,
+      "step": 1998
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002727394346679271,
+      "loss": 3.2467,
+      "step": 1999
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002727058121231757,
+      "loss": 2.7819,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002726721709316995,
+      "loss": 2.7148,
+      "step": 2001
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027263851109861066,
+      "loss": 2.5684,
+      "step": 2002
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002726048326290242,
+      "loss": 3.1405,
+      "step": 2003
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000272571135528058,
+      "loss": 2.7921,
+      "step": 2004
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002725374198008328,
+      "loss": 2.9145,
+      "step": 2005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027250368545247213,
+      "loss": 2.7286,
+      "step": 2006
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027246993248810237,
+      "loss": 2.6942,
+      "step": 2007
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002724361609128527,
+      "loss": 2.841,
+      "step": 2008
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027240237073185514,
+      "loss": 3.0027,
+      "step": 2009
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002723685619502445,
+      "loss": 2.8789,
+      "step": 2010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027233473457315854,
+      "loss": 2.6241,
+      "step": 2011
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027230088860573777,
+      "loss": 2.7615,
+      "step": 2012
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027226702405312546,
+      "loss": 2.4858,
+      "step": 2013
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027223314092046777,
+      "loss": 2.5982,
+      "step": 2014
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002721992392129136,
+      "loss": 2.6343,
+      "step": 2015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027216531893561495,
+      "loss": 3.7947,
+      "step": 2016
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027213138009372635,
+      "loss": 2.669,
+      "step": 2017
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002720974226924052,
+      "loss": 2.9321,
+      "step": 2018
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027206344673681173,
+      "loss": 3.0496,
+      "step": 2019
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027202945223210915,
+      "loss": 3.0528,
+      "step": 2020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002719954391834633,
+      "loss": 2.8952,
+      "step": 2021
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027196140759604283,
+      "loss": 2.6347,
+      "step": 2022
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027192735747501935,
+      "loss": 3.049,
+      "step": 2023
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027189328882556724,
+      "loss": 2.5933,
+      "step": 2024
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002718592016528636,
+      "loss": 2.7608,
+      "step": 2025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027182509596208854,
+      "loss": 2.8189,
+      "step": 2026
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002717909717584247,
+      "loss": 2.958,
+      "step": 2027
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027175682904705773,
+      "loss": 2.8066,
+      "step": 2028
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002717226678331761,
+      "loss": 1.5485,
+      "step": 2029
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027168848812197103,
+      "loss": 2.7311,
+      "step": 2030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027165428991863654,
+      "loss": 2.975,
+      "step": 2031
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002716200732283695,
+      "loss": 2.9399,
+      "step": 2032
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027158583805636966,
+      "loss": 2.5494,
+      "step": 2033
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027155158440783937,
+      "loss": 2.5064,
+      "step": 2034
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000271517312287984,
+      "loss": 2.9986,
+      "step": 2035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027148302170201154,
+      "loss": 2.6069,
+      "step": 2036
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027144871265513297,
+      "loss": 3.6084,
+      "step": 2037
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027141438515256196,
+      "loss": 2.9777,
+      "step": 2038
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027138003919951505,
+      "loss": 2.5584,
+      "step": 2039
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002713456748012115,
+      "loss": 2.5256,
+      "step": 2040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002713112919628734,
+      "loss": 2.7466,
+      "step": 2041
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027127689068972573,
+      "loss": 2.7735,
+      "step": 2042
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002712424709869962,
+      "loss": 2.6621,
+      "step": 2043
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002712080328599153,
+      "loss": 3.0674,
+      "step": 2044
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027117357631371635,
+      "loss": 2.8577,
+      "step": 2045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002711391013536355,
+      "loss": 2.5811,
+      "step": 2046
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002711046079849116,
+      "loss": 2.7118,
+      "step": 2047
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002710700962127864,
+      "loss": 2.8767,
+      "step": 2048
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027103556604250435,
+      "loss": 2.5837,
+      "step": 2049
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027100101747931285,
+      "loss": 2.6762,
+      "step": 2050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027096645052846196,
+      "loss": 3.0506,
+      "step": 2051
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002709318651952045,
+      "loss": 2.0672,
+      "step": 2052
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002708972614847963,
+      "loss": 3.0633,
+      "step": 2053
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002708626394024957,
+      "loss": 2.9092,
+      "step": 2054
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027082799895356403,
+      "loss": 2.7803,
+      "step": 2055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027079334014326535,
+      "loss": 3.1153,
+      "step": 2056
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002707586629768665,
+      "loss": 2.5328,
+      "step": 2057
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002707239674596372,
+      "loss": 3.2654,
+      "step": 2058
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027068925359684975,
+      "loss": 2.8566,
+      "step": 2059
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027065452139377944,
+      "loss": 2.5088,
+      "step": 2060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002706197708557043,
+      "loss": 2.5104,
+      "step": 2061
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000270585001987905,
+      "loss": 3.0381,
+      "step": 2062
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002705502147956653,
+      "loss": 2.8486,
+      "step": 2063
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002705154092842714,
+      "loss": 2.2672,
+      "step": 2064
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027048058545901246,
+      "loss": 2.7899,
+      "step": 2065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002704457433251805,
+      "loss": 2.827,
+      "step": 2066
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027041088288807017,
+      "loss": 3.1636,
+      "step": 2067
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027037600415297903,
+      "loss": 2.5588,
+      "step": 2068
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002703411071252072,
+      "loss": 2.7721,
+      "step": 2069
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002703061918100579,
+      "loss": 3.2445,
+      "step": 2070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002702712582128369,
+      "loss": 2.7531,
+      "step": 2071
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027023630633885276,
+      "loss": 2.2277,
+      "step": 2072
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002702013361934169,
+      "loss": 2.8458,
+      "step": 2073
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002701663477818435,
+      "loss": 2.6339,
+      "step": 2074
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002701313411094495,
+      "loss": 2.2801,
+      "step": 2075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027009631618155457,
+      "loss": 2.0429,
+      "step": 2076
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027006127300348125,
+      "loss": 2.3466,
+      "step": 2077
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00027002621158055476,
+      "loss": 2.6504,
+      "step": 2078
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026999113191810315,
+      "loss": 2.3735,
+      "step": 2079
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026995603402145727,
+      "loss": 2.3521,
+      "step": 2080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002699209178959506,
+      "loss": 3.1794,
+      "step": 2081
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026988578354691955,
+      "loss": 2.7896,
+      "step": 2082
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026985063097970325,
+      "loss": 2.9561,
+      "step": 2083
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026981546019964355,
+      "loss": 2.6144,
+      "step": 2084
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002697802712120851,
+      "loss": 2.7933,
+      "step": 2085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026974506402237535,
+      "loss": 2.782,
+      "step": 2086
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026970983863586445,
+      "loss": 2.6553,
+      "step": 2087
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026967459505790537,
+      "loss": 3.0134,
+      "step": 2088
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026963933329385384,
+      "loss": 2.6935,
+      "step": 2089
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002696040533490683,
+      "loss": 2.7287,
+      "step": 2090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026956875522891,
+      "loss": 3.0563,
+      "step": 2091
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026953343893874296,
+      "loss": 2.727,
+      "step": 2092
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026949810448393394,
+      "loss": 3.5139,
+      "step": 2093
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026946275186985245,
+      "loss": 2.7188,
+      "step": 2094
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002694273811018708,
+      "loss": 3.1612,
+      "step": 2095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026939199218536404,
+      "loss": 2.6707,
+      "step": 2096
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002693565851257099,
+      "loss": 2.7089,
+      "step": 2097
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026932115992828903,
+      "loss": 2.7893,
+      "step": 2098
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002692857165984847,
+      "loss": 2.9867,
+      "step": 2099
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002692502551416829,
+      "loss": 2.506,
+      "step": 2100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002692147755632726,
+      "loss": 2.8482,
+      "step": 2101
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026917927786864534,
+      "loss": 2.5351,
+      "step": 2102
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002691437620631953,
+      "loss": 3.0935,
+      "step": 2103
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002691082281523198,
+      "loss": 2.8415,
+      "step": 2104
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026907267614141854,
+      "loss": 1.6575,
+      "step": 2105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002690371060358941,
+      "loss": 2.887,
+      "step": 2106
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002690015178411519,
+      "loss": 2.1566,
+      "step": 2107
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002689659115625999,
+      "loss": 2.5045,
+      "step": 2108
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000268930287205649,
+      "loss": 2.42,
+      "step": 2109
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002688946447757128,
+      "loss": 2.2795,
+      "step": 2110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002688589842782075,
+      "loss": 3.0712,
+      "step": 2111
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026882330571855237,
+      "loss": 2.9018,
+      "step": 2112
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002687876091021691,
+      "loss": 2.5033,
+      "step": 2113
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026875189443448233,
+      "loss": 2.3258,
+      "step": 2114
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002687161617209192,
+      "loss": 2.7765,
+      "step": 2115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026868041096690995,
+      "loss": 3.252,
+      "step": 2116
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026864464217788727,
+      "loss": 3.1052,
+      "step": 2117
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026860885535928664,
+      "loss": 2.819,
+      "step": 2118
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002685730505165464,
+      "loss": 2.9251,
+      "step": 2119
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002685372276551075,
+      "loss": 1.56,
+      "step": 2120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026850138678041383,
+      "loss": 2.4561,
+      "step": 2121
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026846552789791165,
+      "loss": 3.1009,
+      "step": 2122
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026842965101305037,
+      "loss": 2.6584,
+      "step": 2123
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026839375613128184,
+      "loss": 3.2898,
+      "step": 2124
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026835784325806076,
+      "loss": 3.3363,
+      "step": 2125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002683219123988446,
+      "loss": 2.7847,
+      "step": 2126
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002682859635590934,
+      "loss": 1.8625,
+      "step": 2127
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002682499967442702,
+      "loss": 2.8448,
+      "step": 2128
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002682140119598406,
+      "loss": 2.5566,
+      "step": 2129
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002681780092112728,
+      "loss": 3.1852,
+      "step": 2130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000268141988504038,
+      "loss": 2.4957,
+      "step": 2131
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026810594984361,
+      "loss": 3.2194,
+      "step": 2132
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026806989323546537,
+      "loss": 3.211,
+      "step": 2133
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026803381868508325,
+      "loss": 3.3177,
+      "step": 2134
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002679977261979458,
+      "loss": 2.6353,
+      "step": 2135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002679616157795376,
+      "loss": 2.6999,
+      "step": 2136
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002679254874353462,
+      "loss": 3.1522,
+      "step": 2137
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002678893411708616,
+      "loss": 2.7676,
+      "step": 2138
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026785317699157686,
+      "loss": 2.8721,
+      "step": 2139
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026781699490298756,
+      "loss": 3.1041,
+      "step": 2140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026778079491059195,
+      "loss": 2.6539,
+      "step": 2141
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026774457701989117,
+      "loss": 2.3565,
+      "step": 2142
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026770834123638894,
+      "loss": 2.6721,
+      "step": 2143
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002676720875655918,
+      "loss": 3.0128,
+      "step": 2144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002676358160130089,
+      "loss": 2.9945,
+      "step": 2145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026759952658415224,
+      "loss": 2.8895,
+      "step": 2146
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002675632192845364,
+      "loss": 2.9412,
+      "step": 2147
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026752689411967877,
+      "loss": 2.7166,
+      "step": 2148
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026749055109509946,
+      "loss": 2.3463,
+      "step": 2149
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002674541902163212,
+      "loss": 3.0708,
+      "step": 2150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026741781148886953,
+      "loss": 2.4134,
+      "step": 2151
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002673814149182726,
+      "loss": 3.0489,
+      "step": 2152
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026734500051006144,
+      "loss": 2.378,
+      "step": 2153
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026730856826976963,
+      "loss": 2.5569,
+      "step": 2154
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026727211820293354,
+      "loss": 3.1679,
+      "step": 2155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002672356503150922,
+      "loss": 2.7396,
+      "step": 2156
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026719916461178734,
+      "loss": 2.6139,
+      "step": 2157
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002671626610985635,
+      "loss": 3.3925,
+      "step": 2158
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026712613978096785,
+      "loss": 2.7114,
+      "step": 2159
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026708960066455025,
+      "loss": 2.837,
+      "step": 2160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002670530437548633,
+      "loss": 2.5898,
+      "step": 2161
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026701646905746227,
+      "loss": 2.7305,
+      "step": 2162
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026697987657790513,
+      "loss": 2.6439,
+      "step": 2163
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002669432663217527,
+      "loss": 2.9081,
+      "step": 2164
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026690663829456826,
+      "loss": 2.1738,
+      "step": 2165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026686999250191793,
+      "loss": 2.7464,
+      "step": 2166
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002668333289493706,
+      "loss": 2.9429,
+      "step": 2167
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002667966476424976,
+      "loss": 2.9152,
+      "step": 2168
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026675994858687326,
+      "loss": 2.6224,
+      "step": 2169
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002667232317880745,
+      "loss": 2.7228,
+      "step": 2170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002666864972516807,
+      "loss": 2.7524,
+      "step": 2171
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002666497449832744,
+      "loss": 3.202,
+      "step": 2172
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002666129749884404,
+      "loss": 2.7151,
+      "step": 2173
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002665761872727665,
+      "loss": 2.7551,
+      "step": 2174
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026653938184184297,
+      "loss": 2.6682,
+      "step": 2175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026650255870126294,
+      "loss": 3.1781,
+      "step": 2176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026646571785662204,
+      "loss": 3.2647,
+      "step": 2177
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026642885931351887,
+      "loss": 2.7793,
+      "step": 2178
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026639198307755447,
+      "loss": 2.5356,
+      "step": 2179
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026635508915433265,
+      "loss": 2.3809,
+      "step": 2180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002663181775494599,
+      "loss": 2.6339,
+      "step": 2181
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002662812482685456,
+      "loss": 2.4229,
+      "step": 2182
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026624430131720134,
+      "loss": 2.1464,
+      "step": 2183
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026620733670104185,
+      "loss": 2.9486,
+      "step": 2184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026617035442568437,
+      "loss": 3.0422,
+      "step": 2185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026613335449674884,
+      "loss": 3.0444,
+      "step": 2186
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026609633691985774,
+      "loss": 3.1775,
+      "step": 2187
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002660593017006366,
+      "loss": 2.9686,
+      "step": 2188
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026602224884471315,
+      "loss": 2.89,
+      "step": 2189
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002659851783577183,
+      "loss": 2.7793,
+      "step": 2190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026594809024528516,
+      "loss": 2.9258,
+      "step": 2191
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026591098451304987,
+      "loss": 2.7894,
+      "step": 2192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002658738611666511,
+      "loss": 3.1413,
+      "step": 2193
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002658367202117302,
+      "loss": 3.6967,
+      "step": 2194
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002657995616539312,
+      "loss": 2.77,
+      "step": 2195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002657623854989009,
+      "loss": 2.6676,
+      "step": 2196
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002657251917522886,
+      "loss": 2.821,
+      "step": 2197
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026568798041974644,
+      "loss": 2.8998,
+      "step": 2198
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002656507515069291,
+      "loss": 2.384,
+      "step": 2199
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002656135050194941,
+      "loss": 2.8975,
+      "step": 2200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026557624096310137,
+      "loss": 2.6181,
+      "step": 2201
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026553895934341374,
+      "loss": 2.5275,
+      "step": 2202
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026550166016609667,
+      "loss": 2.585,
+      "step": 2203
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026546434343681813,
+      "loss": 2.6398,
+      "step": 2204
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000265427009161249,
+      "loss": 3.3042,
+      "step": 2205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002653896573450627,
+      "loss": 2.6254,
+      "step": 2206
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026535228799393523,
+      "loss": 3.055,
+      "step": 2207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002653149011135454,
+      "loss": 0.8582,
+      "step": 2208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002652774967095746,
+      "loss": 2.3675,
+      "step": 2209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002652400747877069,
+      "loss": 3.0375,
+      "step": 2210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002652026353536291,
+      "loss": 2.8784,
+      "step": 2211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002651651784130306,
+      "loss": 2.9782,
+      "step": 2212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026512770397160343,
+      "loss": 2.8513,
+      "step": 2213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026509021203504225,
+      "loss": 2.5725,
+      "step": 2214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002650527026090445,
+      "loss": 2.6585,
+      "step": 2215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002650151756993103,
+      "loss": 3.0785,
+      "step": 2216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002649776313115423,
+      "loss": 2.4201,
+      "step": 2217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002649400694514457,
+      "loss": 3.2461,
+      "step": 2218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026490249012472875,
+      "loss": 2.7091,
+      "step": 2219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026486489333710196,
+      "loss": 3.0495,
+      "step": 2220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026482727909427864,
+      "loss": 2.8735,
+      "step": 2221
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026478964740197483,
+      "loss": 2.8144,
+      "step": 2222
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002647519982659091,
+      "loss": 2.7561,
+      "step": 2223
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002647143316918027,
+      "loss": 2.7701,
+      "step": 2224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026467664768537965,
+      "loss": 2.7071,
+      "step": 2225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026463894625236643,
+      "loss": 2.7116,
+      "step": 2226
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026460122739849234,
+      "loss": 2.8939,
+      "step": 2227
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002645634911294891,
+      "loss": 3.1739,
+      "step": 2228
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026452573745109134,
+      "loss": 2.1659,
+      "step": 2229
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002644879663690362,
+      "loss": 2.3508,
+      "step": 2230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026445017788906346,
+      "loss": 2.7127,
+      "step": 2231
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026441237201691554,
+      "loss": 3.002,
+      "step": 2232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026437454875833754,
+      "loss": 2.95,
+      "step": 2233
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026433670811907726,
+      "loss": 3.1688,
+      "step": 2234
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000264298850104885,
+      "loss": 2.6564,
+      "step": 2235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026426097472151384,
+      "loss": 2.7327,
+      "step": 2236
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026422308197471934,
+      "loss": 2.6584,
+      "step": 2237
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002641851718702598,
+      "loss": 2.3295,
+      "step": 2238
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026414724441389624,
+      "loss": 1.9844,
+      "step": 2239
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026410929961139207,
+      "loss": 2.991,
+      "step": 2240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002640713374685137,
+      "loss": 2.9538,
+      "step": 2241
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002640333579910298,
+      "loss": 2.8979,
+      "step": 2242
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026399536118471185,
+      "loss": 2.7309,
+      "step": 2243
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026395734705533406,
+      "loss": 2.9351,
+      "step": 2244
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002639193156086731,
+      "loss": 2.943,
+      "step": 2245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026388126685050826,
+      "loss": 3.5657,
+      "step": 2246
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026384320078662164,
+      "loss": 2.5064,
+      "step": 2247
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026380511742279787,
+      "loss": 3.3528,
+      "step": 2248
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026376701676482416,
+      "loss": 2.7964,
+      "step": 2249
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002637288988184904,
+      "loss": 2.4671,
+      "step": 2250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026369076358958915,
+      "loss": 3.192,
+      "step": 2251
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002636526110839155,
+      "loss": 2.6807,
+      "step": 2252
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002636144413072672,
+      "loss": 2.8498,
+      "step": 2253
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026357625426544465,
+      "loss": 2.914,
+      "step": 2254
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002635380499642509,
+      "loss": 2.5694,
+      "step": 2255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002634998284094916,
+      "loss": 2.9059,
+      "step": 2256
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026346158960697494,
+      "loss": 3.0062,
+      "step": 2257
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026342333356251175,
+      "loss": 2.331,
+      "step": 2258
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026338506028191573,
+      "loss": 2.3254,
+      "step": 2259
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002633467697710027,
+      "loss": 1.6289,
+      "step": 2260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002633084620355917,
+      "loss": 2.5373,
+      "step": 2261
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002632701370815039,
+      "loss": 2.7235,
+      "step": 2262
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002632317949145633,
+      "loss": 3.283,
+      "step": 2263
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026319343554059653,
+      "loss": 2.7184,
+      "step": 2264
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026315505896543274,
+      "loss": 2.555,
+      "step": 2265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026311666519490375,
+      "loss": 2.4084,
+      "step": 2266
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000263078254234844,
+      "loss": 2.5591,
+      "step": 2267
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026303982609109057,
+      "loss": 2.8448,
+      "step": 2268
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000263001380769483,
+      "loss": 1.8714,
+      "step": 2269
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002629629182758637,
+      "loss": 3.0967,
+      "step": 2270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026292443861607737,
+      "loss": 2.6573,
+      "step": 2271
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026288594179597164,
+      "loss": 2.8441,
+      "step": 2272
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002628474278213965,
+      "loss": 3.4417,
+      "step": 2273
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026280889669820474,
+      "loss": 2.8656,
+      "step": 2274
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026277034843225154,
+      "loss": 2.5561,
+      "step": 2275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026273178302939484,
+      "loss": 3.5309,
+      "step": 2276
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002626932004954952,
+      "loss": 1.9343,
+      "step": 2277
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002626546008364157,
+      "loss": 2.6925,
+      "step": 2278
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002626159840580221,
+      "loss": 2.377,
+      "step": 2279
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002625773501661826,
+      "loss": 3.0639,
+      "step": 2280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002625386991667682,
+      "loss": 3.362,
+      "step": 2281
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002625000310656524,
+      "loss": 2.2016,
+      "step": 2282
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026246134586871135,
+      "loss": 3.1123,
+      "step": 2283
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002624226435818237,
+      "loss": 3.5924,
+      "step": 2284
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026238392421087085,
+      "loss": 2.6698,
+      "step": 2285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002623451877617366,
+      "loss": 3.0593,
+      "step": 2286
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026230643424030747,
+      "loss": 2.8331,
+      "step": 2287
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002622676636524726,
+      "loss": 2.0805,
+      "step": 2288
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002622288760041237,
+      "loss": 2.5743,
+      "step": 2289
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000262190071301155,
+      "loss": 2.5906,
+      "step": 2290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026215124954946336,
+      "loss": 2.2077,
+      "step": 2291
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026211241075494834,
+      "loss": 2.8309,
+      "step": 2292
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026207355492351186,
+      "loss": 3.0436,
+      "step": 2293
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026203468206105864,
+      "loss": 2.6381,
+      "step": 2294
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002619957921734959,
+      "loss": 2.4193,
+      "step": 2295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026195688526673344,
+      "loss": 3.2163,
+      "step": 2296
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026191796134668376,
+      "loss": 2.5481,
+      "step": 2297
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002618790204192617,
+      "loss": 3.0774,
+      "step": 2298
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026184006249038494,
+      "loss": 2.6151,
+      "step": 2299
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026180108756597355,
+      "loss": 2.5496,
+      "step": 2300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026176209565195036,
+      "loss": 3.3535,
+      "step": 2301
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002617230867542406,
+      "loss": 3.0444,
+      "step": 2302
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026168406087877233,
+      "loss": 2.7997,
+      "step": 2303
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002616450180314759,
+      "loss": 2.153,
+      "step": 2304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002616059582182844,
+      "loss": 2.1542,
+      "step": 2305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026156688144513343,
+      "loss": 1.5743,
+      "step": 2306
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002615277877179613,
+      "loss": 2.8212,
+      "step": 2307
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002614886770427087,
+      "loss": 2.7934,
+      "step": 2308
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002614495494253191,
+      "loss": 2.3364,
+      "step": 2309
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002614104048717384,
+      "loss": 2.5766,
+      "step": 2310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026137124338791517,
+      "loss": 2.8633,
+      "step": 2311
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002613320649798004,
+      "loss": 2.6447,
+      "step": 2312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002612928696533478,
+      "loss": 2.2579,
+      "step": 2313
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026125365741451364,
+      "loss": 2.5759,
+      "step": 2314
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026121442826925666,
+      "loss": 2.6758,
+      "step": 2315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002611751822235383,
+      "loss": 3.113,
+      "step": 2316
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002611359192833224,
+      "loss": 3.0955,
+      "step": 2317
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026109663945457555,
+      "loss": 2.892,
+      "step": 2318
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002610573427432668,
+      "loss": 2.8517,
+      "step": 2319
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026101802915536786,
+      "loss": 2.5578,
+      "step": 2320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002609786986968528,
+      "loss": 2.3432,
+      "step": 2321
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002609393513736985,
+      "loss": 2.6762,
+      "step": 2322
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002608999871918842,
+      "loss": 2.7935,
+      "step": 2323
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002608606061573919,
+      "loss": 1.8682,
+      "step": 2324
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026082120827620596,
+      "loss": 1.77,
+      "step": 2325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002607817935543135,
+      "loss": 2.5314,
+      "step": 2326
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026074236199770395,
+      "loss": 2.3046,
+      "step": 2327
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026070291361236954,
+      "loss": 2.6624,
+      "step": 2328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026066344840430496,
+      "loss": 2.6729,
+      "step": 2329
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002606239663795074,
+      "loss": 2.9882,
+      "step": 2330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002605844675439767,
+      "loss": 2.5537,
+      "step": 2331
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002605449519037152,
+      "loss": 3.277,
+      "step": 2332
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002605054194647279,
+      "loss": 1.9762,
+      "step": 2333
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002604658702330221,
+      "loss": 1.049,
+      "step": 2334
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026042630421460795,
+      "loss": 3.0271,
+      "step": 2335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002603867214154979,
+      "loss": 2.8188,
+      "step": 2336
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026034712184170715,
+      "loss": 2.8525,
+      "step": 2337
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026030750549925334,
+      "loss": 2.8084,
+      "step": 2338
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026026787239415665,
+      "loss": 2.685,
+      "step": 2339
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002602282225324399,
+      "loss": 2.6078,
+      "step": 2340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026018855592012835,
+      "loss": 2.158,
+      "step": 2341
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026014887256324993,
+      "loss": 2.7306,
+      "step": 2342
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002601091724678349,
+      "loss": 2.6546,
+      "step": 2343
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002600694556399163,
+      "loss": 2.8498,
+      "step": 2344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00026002972208552955,
+      "loss": 2.6734,
+      "step": 2345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025998997181071276,
+      "loss": 2.5403,
+      "step": 2346
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025995020482150644,
+      "loss": 2.4253,
+      "step": 2347
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002599104211239537,
+      "loss": 3.1357,
+      "step": 2348
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002598706207241002,
+      "loss": 3.0977,
+      "step": 2349
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002598308036279941,
+      "loss": 2.8793,
+      "step": 2350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025979096984168605,
+      "loss": 2.6402,
+      "step": 2351
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025975111937122944,
+      "loss": 3.3322,
+      "step": 2352
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025971125222268,
+      "loss": 2.8716,
+      "step": 2353
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025967136840209605,
+      "loss": 1.1909,
+      "step": 2354
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002596314679155384,
+      "loss": 2.7532,
+      "step": 2355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025959155076907064,
+      "loss": 2.9442,
+      "step": 2356
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002595516169687585,
+      "loss": 2.5794,
+      "step": 2357
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002595116665206704,
+      "loss": 3.0412,
+      "step": 2358
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025947169943087747,
+      "loss": 2.9441,
+      "step": 2359
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025943171570545314,
+      "loss": 2.7851,
+      "step": 2360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002593917153504735,
+      "loss": 2.4378,
+      "step": 2361
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025935169837201707,
+      "loss": 2.5787,
+      "step": 2362
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025931166477616497,
+      "loss": 2.8426,
+      "step": 2363
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002592716145690009,
+      "loss": 2.9368,
+      "step": 2364
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002592315477566108,
+      "loss": 2.8098,
+      "step": 2365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002591914643450835,
+      "loss": 2.8705,
+      "step": 2366
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002591513643405101,
+      "loss": 3.0212,
+      "step": 2367
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025911124774898443,
+      "loss": 2.3073,
+      "step": 2368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025907111457660264,
+      "loss": 2.6401,
+      "step": 2369
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025903096482946344,
+      "loss": 3.634,
+      "step": 2370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002589907985136682,
+      "loss": 3.0702,
+      "step": 2371
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025895061563532064,
+      "loss": 2.9599,
+      "step": 2372
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025891041620052706,
+      "loss": 1.6705,
+      "step": 2373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002588702002153963,
+      "loss": 3.1524,
+      "step": 2374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00025882996768603967,
+      "loss": 2.1486,
+      "step": 2375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002587897186185711,
+      "loss": 2.6439,
+      "step": 2376
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002587494530191069,
+      "loss": 2.85,
+      "step": 2377
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002587091708937659,
+      "loss": 2.4027,
+      "step": 2378
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002586688722486696,
+      "loss": 3.048,
+      "step": 2379
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002586285570899418,
+      "loss": 2.9142,
+      "step": 2380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002585882254237089,
+      "loss": 2.7126,
+      "step": 2381
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002585478772560999,
+      "loss": 2.3847,
+      "step": 2382
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002585075125932461,
+      "loss": 2.1415,
+      "step": 2383
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025846713144128154,
+      "loss": 2.6203,
+      "step": 2384
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025842673380634264,
+      "loss": 3.0773,
+      "step": 2385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002583863196945683,
+      "loss": 2.759,
+      "step": 2386
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002583458891121,
+      "loss": 2.6942,
+      "step": 2387
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025830544206508166,
+      "loss": 3.1192,
+      "step": 2388
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002582649785596597,
+      "loss": 3.02,
+      "step": 2389
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002582244986019831,
+      "loss": 2.587,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002581840021982034,
+      "loss": 3.1605,
+      "step": 2391
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002581434893544744,
+      "loss": 2.8121,
+      "step": 2392
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002581029600769526,
+      "loss": 3.1325,
+      "step": 2393
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025806241437179695,
+      "loss": 2.9231,
+      "step": 2394
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002580218522451689,
+      "loss": 2.5691,
+      "step": 2395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025798127370323245,
+      "loss": 2.6403,
+      "step": 2396
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002579406787521539,
+      "loss": 2.899,
+      "step": 2397
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002579000673981022,
+      "loss": 3.0178,
+      "step": 2398
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002578594396472489,
+      "loss": 1.6602,
+      "step": 2399
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025781879550576775,
+      "loss": 2.1582,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002577781349798352,
+      "loss": 2.4533,
+      "step": 2401
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025773745807563023,
+      "loss": 3.1804,
+      "step": 2402
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002576967647993341,
+      "loss": 2.6968,
+      "step": 2403
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002576560551571307,
+      "loss": 3.0052,
+      "step": 2404
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002576153291552064,
+      "loss": 2.5389,
+      "step": 2405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025757458679975,
+      "loss": 2.3649,
+      "step": 2406
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025753382809695294,
+      "loss": 2.9613,
+      "step": 2407
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025749305305300895,
+      "loss": 2.5902,
+      "step": 2408
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025745226167411426,
+      "loss": 1.7052,
+      "step": 2409
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025741145396646775,
+      "loss": 2.6583,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002573706299362706,
+      "loss": 2.2846,
+      "step": 2411
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025732978958972665,
+      "loss": 2.5845,
+      "step": 2412
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000257288932933042,
+      "loss": 2.7103,
+      "step": 2413
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025724805997242537,
+      "loss": 2.8741,
+      "step": 2414
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025720717071408796,
+      "loss": 2.6072,
+      "step": 2415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025716626516424345,
+      "loss": 3.0569,
+      "step": 2416
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002571253433291079,
+      "loss": 2.7937,
+      "step": 2417
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002570844052148999,
+      "loss": 2.7178,
+      "step": 2418
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002570434508278406,
+      "loss": 2.7136,
+      "step": 2419
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025700248017415347,
+      "loss": 2.4555,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025696149326006457,
+      "loss": 2.9399,
+      "step": 2421
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002569204900918024,
+      "loss": 2.9208,
+      "step": 2422
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025687947067559784,
+      "loss": 2.5875,
+      "step": 2423
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002568384350176844,
+      "loss": 2.8961,
+      "step": 2424
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002567973831242979,
+      "loss": 2.6067,
+      "step": 2425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002567563150016768,
+      "loss": 2.6911,
+      "step": 2426
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002567152306560619,
+      "loss": 2.639,
+      "step": 2427
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025667413009369637,
+      "loss": 2.0395,
+      "step": 2428
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002566330133208262,
+      "loss": 2.6729,
+      "step": 2429
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002565918803436994,
+      "loss": 3.1866,
+      "step": 2430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025655073116856675,
+      "loss": 3.0416,
+      "step": 2431
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002565095658016814,
+      "loss": 2.0584,
+      "step": 2432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002564683842492989,
+      "loss": 3.1132,
+      "step": 2433
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025642718651767743,
+      "loss": 2.941,
+      "step": 2434
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002563859726130774,
+      "loss": 2.3165,
+      "step": 2435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025634474254176184,
+      "loss": 2.6237,
+      "step": 2436
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025630349630999613,
+      "loss": 2.9562,
+      "step": 2437
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025626223392404826,
+      "loss": 2.1913,
+      "step": 2438
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025622095539018855,
+      "loss": 3.248,
+      "step": 2439
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025617966071468976,
+      "loss": 3.0835,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025613834990382715,
+      "loss": 1.9732,
+      "step": 2441
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002560970229638785,
+      "loss": 3.1032,
+      "step": 2442
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025605567990112394,
+      "loss": 1.7959,
+      "step": 2443
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025601432072184605,
+      "loss": 2.57,
+      "step": 2444
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025597294543232984,
+      "loss": 3.0082,
+      "step": 2445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025593155403886295,
+      "loss": 2.6085,
+      "step": 2446
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025589014654773526,
+      "loss": 2.461,
+      "step": 2447
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002558487229652392,
+      "loss": 2.8886,
+      "step": 2448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025580728329766954,
+      "loss": 2.8225,
+      "step": 2449
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002557658275513236,
+      "loss": 2.3748,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025572435573250116,
+      "loss": 2.7549,
+      "step": 2451
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002556828678475044,
+      "loss": 2.6679,
+      "step": 2452
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025564136390263787,
+      "loss": 2.4288,
+      "step": 2453
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025559984390420863,
+      "loss": 3.1932,
+      "step": 2454
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002555583078585263,
+      "loss": 1.2981,
+      "step": 2455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025551675577190265,
+      "loss": 2.4434,
+      "step": 2456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002554751876506522,
+      "loss": 2.638,
+      "step": 2457
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002554336035010916,
+      "loss": 2.5394,
+      "step": 2458
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025539200332954027,
+      "loss": 1.9436,
+      "step": 2459
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002553503871423198,
+      "loss": 2.6805,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025530875494575423,
+      "loss": 2.8777,
+      "step": 2461
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002552671067461703,
+      "loss": 2.9566,
+      "step": 2462
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002552254425498967,
+      "loss": 2.8144,
+      "step": 2463
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002551837623632652,
+      "loss": 3.2491,
+      "step": 2464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002551420661926094,
+      "loss": 2.6316,
+      "step": 2465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002551003540442656,
+      "loss": 2.7803,
+      "step": 2466
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025505862592457255,
+      "loss": 3.2118,
+      "step": 2467
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025501688183987136,
+      "loss": 2.7911,
+      "step": 2468
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002549751217965055,
+      "loss": 2.6941,
+      "step": 2469
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002549333458008211,
+      "loss": 2.6524,
+      "step": 2470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002548915538591664,
+      "loss": 2.9061,
+      "step": 2471
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025484974597789233,
+      "loss": 2.7521,
+      "step": 2472
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002548079221633521,
+      "loss": 2.2855,
+      "step": 2473
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025476608242190136,
+      "loss": 2.4071,
+      "step": 2474
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025472422675989825,
+      "loss": 2.5085,
+      "step": 2475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002546823551837032,
+      "loss": 2.8243,
+      "step": 2476
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025464046769967916,
+      "loss": 2.7213,
+      "step": 2477
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025459856431419154,
+      "loss": 3.1006,
+      "step": 2478
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000254556645033608,
+      "loss": 2.8043,
+      "step": 2479
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025451470986429875,
+      "loss": 2.2679,
+      "step": 2480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002544727588126364,
+      "loss": 2.7253,
+      "step": 2481
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025443079188499593,
+      "loss": 2.5694,
+      "step": 2482
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002543888090877547,
+      "loss": 3.0228,
+      "step": 2483
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002543468104272927,
+      "loss": 2.8967,
+      "step": 2484
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025430479590999195,
+      "loss": 3.0091,
+      "step": 2485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002542627655422373,
+      "loss": 3.0498,
+      "step": 2486
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025422071933041564,
+      "loss": 2.8422,
+      "step": 2487
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002541786572809165,
+      "loss": 2.765,
+      "step": 2488
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002541365794001318,
+      "loss": 2.4497,
+      "step": 2489
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002540944856944557,
+      "loss": 2.5879,
+      "step": 2490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025405237617028497,
+      "loss": 2.3327,
+      "step": 2491
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002540102508340187,
+      "loss": 2.5863,
+      "step": 2492
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025396810969205837,
+      "loss": 3.4522,
+      "step": 2493
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002539259527508077,
+      "loss": 3.1443,
+      "step": 2494
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025388378001667326,
+      "loss": 3.1612,
+      "step": 2495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025384159149606354,
+      "loss": 3.1342,
+      "step": 2496
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025379938719538966,
+      "loss": 1.8356,
+      "step": 2497
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025375716712106527,
+      "loss": 2.5769,
+      "step": 2498
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000253714931279506,
+      "loss": 2.833,
+      "step": 2499
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025367267967713026,
+      "loss": 3.1313,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025363041232035865,
+      "loss": 2.6328,
+      "step": 2501
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002535881292156144,
+      "loss": 2.7216,
+      "step": 2502
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002535458303693228,
+      "loss": 2.7587,
+      "step": 2503
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002535035157879118,
+      "loss": 2.8335,
+      "step": 2504
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002534611854778116,
+      "loss": 2.9719,
+      "step": 2505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025341883944545487,
+      "loss": 2.5731,
+      "step": 2506
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025337647769727663,
+      "loss": 2.5579,
+      "step": 2507
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025333410023971423,
+      "loss": 3.1188,
+      "step": 2508
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002532917070792075,
+      "loss": 2.2816,
+      "step": 2509
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002532492982221987,
+      "loss": 1.2249,
+      "step": 2510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002532068736751323,
+      "loss": 2.6556,
+      "step": 2511
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002531644334444553,
+      "loss": 2.2792,
+      "step": 2512
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025312197753661706,
+      "loss": 1.4786,
+      "step": 2513
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025307950595806924,
+      "loss": 2.798,
+      "step": 2514
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025303701871526595,
+      "loss": 2.3698,
+      "step": 2515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025299451581466375,
+      "loss": 2.4992,
+      "step": 2516
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025295199726272147,
+      "loss": 2.1201,
+      "step": 2517
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002529094630659003,
+      "loss": 2.0991,
+      "step": 2518
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002528669132306639,
+      "loss": 3.1477,
+      "step": 2519
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025282434776347824,
+      "loss": 1.8424,
+      "step": 2520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002527817666708117,
+      "loss": 2.4168,
+      "step": 2521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000252739169959135,
+      "loss": 2.5354,
+      "step": 2522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002526965576349213,
+      "loss": 3.0674,
+      "step": 2523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025265392970464615,
+      "loss": 3.2604,
+      "step": 2524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002526112861747873,
+      "loss": 3.1469,
+      "step": 2525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000252568627051825,
+      "loss": 2.44,
+      "step": 2526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002525259523422418,
+      "loss": 2.4718,
+      "step": 2527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025248326205252287,
+      "loss": 3.2363,
+      "step": 2528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002524405561891554,
+      "loss": 2.4712,
+      "step": 2529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025239783475862904,
+      "loss": 2.9572,
+      "step": 2530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002523550977674359,
+      "loss": 2.8152,
+      "step": 2531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002523123452220706,
+      "loss": 2.736,
+      "step": 2532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002522695771290297,
+      "loss": 2.752,
+      "step": 2533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002522267934948124,
+      "loss": 2.7474,
+      "step": 2534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002521839943259204,
+      "loss": 2.3359,
+      "step": 2535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025214117962885736,
+      "loss": 2.7578,
+      "step": 2536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025209834941012965,
+      "loss": 2.8576,
+      "step": 2537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025205550367624584,
+      "loss": 3.071,
+      "step": 2538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002520126424337169,
+      "loss": 2.6386,
+      "step": 2539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002519697656890561,
+      "loss": 3.0702,
+      "step": 2540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025192687344877915,
+      "loss": 2.5998,
+      "step": 2541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025188396571940413,
+      "loss": 2.2878,
+      "step": 2542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002518410425074513,
+      "loss": 2.9215,
+      "step": 2543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002517981038194435,
+      "loss": 3.1181,
+      "step": 2544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025175514966190585,
+      "loss": 2.341,
+      "step": 2545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002517121800413656,
+      "loss": 2.4119,
+      "step": 2546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002516691949643527,
+      "loss": 2.9242,
+      "step": 2547
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025162619443739925,
+      "loss": 3.1976,
+      "step": 2548
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002515831784670397,
+      "loss": 3.3545,
+      "step": 2549
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025154014705981086,
+      "loss": 2.5029,
+      "step": 2550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000251497100222252,
+      "loss": 2.4035,
+      "step": 2551
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002514540379609045,
+      "loss": 2.7395,
+      "step": 2552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002514109602823124,
+      "loss": 2.3997,
+      "step": 2553
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025136786719302177,
+      "loss": 2.9775,
+      "step": 2554
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002513247586995812,
+      "loss": 2.5857,
+      "step": 2555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025128163480854157,
+      "loss": 2.7345,
+      "step": 2556
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002512384955264561,
+      "loss": 2.6212,
+      "step": 2557
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002511953408598804,
+      "loss": 2.7963,
+      "step": 2558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002511521708153724,
+      "loss": 2.9647,
+      "step": 2559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025110898539949225,
+      "loss": 2.4086,
+      "step": 2560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025106578461880257,
+      "loss": 3.2679,
+      "step": 2561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002510225684798683,
+      "loss": 2.8281,
+      "step": 2562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002509793369892567,
+      "loss": 2.617,
+      "step": 2563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025093609015353725,
+      "loss": 3.8695,
+      "step": 2564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025089282797928193,
+      "loss": 2.7672,
+      "step": 2565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000250849550473065,
+      "loss": 2.9589,
+      "step": 2566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000250806257641463,
+      "loss": 3.2315,
+      "step": 2567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002507629494910549,
+      "loss": 2.4957,
+      "step": 2568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025071962602842186,
+      "loss": 2.7675,
+      "step": 2569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025067628726014744,
+      "loss": 2.6243,
+      "step": 2570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002506329331928176,
+      "loss": 2.5769,
+      "step": 2571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025058956383302043,
+      "loss": 2.8746,
+      "step": 2572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002505461791873465,
+      "loss": 2.7007,
+      "step": 2573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002505027792623887,
+      "loss": 3.4741,
+      "step": 2574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002504593640647422,
+      "loss": 2.7732,
+      "step": 2575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025041593360100443,
+      "loss": 2.3565,
+      "step": 2576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002503724878777753,
+      "loss": 2.6681,
+      "step": 2577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025032902690165694,
+      "loss": 3.0431,
+      "step": 2578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025028555067925376,
+      "loss": 2.6568,
+      "step": 2579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025024205921717253,
+      "loss": 2.7635,
+      "step": 2580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002501985525220223,
+      "loss": 3.5272,
+      "step": 2581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025015503060041457,
+      "loss": 2.5763,
+      "step": 2582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000250111493458963,
+      "loss": 3.0519,
+      "step": 2583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002500679411042836,
+      "loss": 2.6919,
+      "step": 2584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00025002437354299477,
+      "loss": 2.5307,
+      "step": 2585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024998079078171714,
+      "loss": 2.4512,
+      "step": 2586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002499371928270736,
+      "loss": 2.2681,
+      "step": 2587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002498935796856895,
+      "loss": 2.8755,
+      "step": 2588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002498499513641924,
+      "loss": 2.2788,
+      "step": 2589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024980630786921215,
+      "loss": 2.2608,
+      "step": 2590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000249762649207381,
+      "loss": 2.8411,
+      "step": 2591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024971897538533333,
+      "loss": 2.7441,
+      "step": 2592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002496752864097061,
+      "loss": 2.6327,
+      "step": 2593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002496315822871383,
+      "loss": 3.169,
+      "step": 2594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024958786302427136,
+      "loss": 2.7099,
+      "step": 2595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000249544128627749,
+      "loss": 2.912,
+      "step": 2596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024950037910421725,
+      "loss": 2.4896,
+      "step": 2597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024945661446032426,
+      "loss": 3.3329,
+      "step": 2598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024941283470272086,
+      "loss": 2.7925,
+      "step": 2599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024936903983805976,
+      "loss": 2.9095,
+      "step": 2600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024932522987299624,
+      "loss": 2.6178,
+      "step": 2601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002492814048141877,
+      "loss": 2.8387,
+      "step": 2602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002492375646682941,
+      "loss": 2.7278,
+      "step": 2603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024919370944197737,
+      "loss": 2.6967,
+      "step": 2604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024914983914190185,
+      "loss": 2.7104,
+      "step": 2605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024910595377473433,
+      "loss": 3.0016,
+      "step": 2606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024906205334714363,
+      "loss": 2.5666,
+      "step": 2607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024901813786580107,
+      "loss": 3.2524,
+      "step": 2608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024897420733738007,
+      "loss": 2.4984,
+      "step": 2609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024893026176855656,
+      "loss": 2.5444,
+      "step": 2610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002488863011660086,
+      "loss": 3.0623,
+      "step": 2611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002488423255364165,
+      "loss": 2.8274,
+      "step": 2612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024879833488646296,
+      "loss": 2.5365,
+      "step": 2613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000248754329222833,
+      "loss": 3.0713,
+      "step": 2614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002487103085522137,
+      "loss": 2.6011,
+      "step": 2615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024866627288129464,
+      "loss": 2.7972,
+      "step": 2616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024862222221676766,
+      "loss": 2.8897,
+      "step": 2617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024857815656532673,
+      "loss": 2.0571,
+      "step": 2618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002485340759336683,
+      "loss": 3.0085,
+      "step": 2619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024848998032849087,
+      "loss": 3.12,
+      "step": 2620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024844586975649537,
+      "loss": 3.5336,
+      "step": 2621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024840174422438506,
+      "loss": 2.7317,
+      "step": 2622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024835760373886526,
+      "loss": 2.2804,
+      "step": 2623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002483134483066437,
+      "loss": 2.2833,
+      "step": 2624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002482692779344305,
+      "loss": 3.2195,
+      "step": 2625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002482250926289377,
+      "loss": 2.9233,
+      "step": 2626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024818089239688,
+      "loss": 1.2115,
+      "step": 2627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024813667724497417,
+      "loss": 2.5488,
+      "step": 2628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002480924471799392,
+      "loss": 2.6301,
+      "step": 2629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024804820220849643,
+      "loss": 2.9684,
+      "step": 2630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002480039423373695,
+      "loss": 2.7271,
+      "step": 2631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024795966757328416,
+      "loss": 2.2181,
+      "step": 2632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024791537792296866,
+      "loss": 2.8356,
+      "step": 2633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024787107339315336,
+      "loss": 2.7503,
+      "step": 2634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002478267539905708,
+      "loss": 2.9062,
+      "step": 2635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000247782419721956,
+      "loss": 1.9916,
+      "step": 2636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002477380705940461,
+      "loss": 2.996,
+      "step": 2637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002476937066135804,
+      "loss": 3.1879,
+      "step": 2638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002476493277873008,
+      "loss": 2.6029,
+      "step": 2639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024760493412195097,
+      "loss": 2.61,
+      "step": 2640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002475605256242773,
+      "loss": 2.8795,
+      "step": 2641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002475161023010282,
+      "loss": 2.9682,
+      "step": 2642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024747166415895426,
+      "loss": 2.2764,
+      "step": 2643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024742721120480857,
+      "loss": 2.9328,
+      "step": 2644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024738274344534623,
+      "loss": 2.4318,
+      "step": 2645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002473382608873247,
+      "loss": 2.1315,
+      "step": 2646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002472937635375036,
+      "loss": 2.8329,
+      "step": 2647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024724925140264507,
+      "loss": 2.4123,
+      "step": 2648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002472047244895132,
+      "loss": 2.9874,
+      "step": 2649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002471601828048744,
+      "loss": 2.4869,
+      "step": 2650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002471156263554973,
+      "loss": 3.1705,
+      "step": 2651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024707105514815294,
+      "loss": 2.8401,
+      "step": 2652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002470264691896145,
+      "loss": 2.5939,
+      "step": 2653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002469818684866572,
+      "loss": 2.4422,
+      "step": 2654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002469372530460589,
+      "loss": 1.5993,
+      "step": 2655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024689262287459947,
+      "loss": 2.6399,
+      "step": 2656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024684797797906087,
+      "loss": 2.8114,
+      "step": 2657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024680331836622767,
+      "loss": 2.9353,
+      "step": 2658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002467586440428863,
+      "loss": 2.6249,
+      "step": 2659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002467139550158257,
+      "loss": 2.9998,
+      "step": 2660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000246669251291837,
+      "loss": 3.3112,
+      "step": 2661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002466245328777133,
+      "loss": 2.0824,
+      "step": 2662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002465797997802503,
+      "loss": 2.2655,
+      "step": 2663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002465350520062457,
+      "loss": 2.6317,
+      "step": 2664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002464902895624996,
+      "loss": 2.8978,
+      "step": 2665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002464455124558141,
+      "loss": 2.7516,
+      "step": 2666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002464007206929937,
+      "loss": 3.3002,
+      "step": 2667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002463559142808452,
+      "loss": 2.9185,
+      "step": 2668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002463110932261773,
+      "loss": 2.6474,
+      "step": 2669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002462662575358012,
+      "loss": 2.9182,
+      "step": 2670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024622140721653043,
+      "loss": 2.5863,
+      "step": 2671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002461765422751804,
+      "loss": 2.6206,
+      "step": 2672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002461316627185689,
+      "loss": 3.2415,
+      "step": 2673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024608676855351604,
+      "loss": 2.8006,
+      "step": 2674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024604185978684405,
+      "loss": 2.6217,
+      "step": 2675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002459969364253774,
+      "loss": 2.42,
+      "step": 2676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002459519984759427,
+      "loss": 2.8706,
+      "step": 2677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024590704594536894,
+      "loss": 3.2522,
+      "step": 2678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002458620788404872,
+      "loss": 3.0098,
+      "step": 2679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024581709716813086,
+      "loss": 2.4937,
+      "step": 2680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024577210093513536,
+      "loss": 2.7306,
+      "step": 2681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002457270901483385,
+      "loss": 2.9099,
+      "step": 2682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024568206481458024,
+      "loss": 2.3554,
+      "step": 2683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024563702494070276,
+      "loss": 3.0532,
+      "step": 2684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002455919705335505,
+      "loss": 2.5909,
+      "step": 2685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024554690159997005,
+      "loss": 2.2098,
+      "step": 2686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002455018181468101,
+      "loss": 2.4728,
+      "step": 2687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002454567201809218,
+      "loss": 2.702,
+      "step": 2688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002454116077091583,
+      "loss": 2.4197,
+      "step": 2689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024536648073837495,
+      "loss": 2.3965,
+      "step": 2690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002453213392754295,
+      "loss": 3.2408,
+      "step": 2691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024527618332718167,
+      "loss": 2.775,
+      "step": 2692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002452310129004936,
+      "loss": 3.3776,
+      "step": 2693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002451858280022294,
+      "loss": 2.792,
+      "step": 2694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024514062863925564,
+      "loss": 3.2583,
+      "step": 2695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002450954148184408,
+      "loss": 2.9263,
+      "step": 2696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024505018654665575,
+      "loss": 2.8524,
+      "step": 2697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002450049438307735,
+      "loss": 2.7875,
+      "step": 2698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024495968667766937,
+      "loss": 3.1292,
+      "step": 2699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002449144150942206,
+      "loss": 2.9415,
+      "step": 2700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002448691290873068,
+      "loss": 2.5017,
+      "step": 2701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002448238286638099,
+      "loss": 2.6025,
+      "step": 2702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002447785138306138,
+      "loss": 2.5172,
+      "step": 2703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002447331845946047,
+      "loss": 2.5544,
+      "step": 2704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024468784096267086,
+      "loss": 2.9043,
+      "step": 2705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002446424829417029,
+      "loss": 2.8141,
+      "step": 2706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002445971105385936,
+      "loss": 1.5055,
+      "step": 2707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024455172376023786,
+      "loss": 2.4553,
+      "step": 2708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002445063226135327,
+      "loss": 2.8373,
+      "step": 2709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002444609071053775,
+      "loss": 2.6676,
+      "step": 2710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002444154772426737,
+      "loss": 2.5125,
+      "step": 2711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002443700330323249,
+      "loss": 3.2233,
+      "step": 2712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002443245744812371,
+      "loss": 3.0682,
+      "step": 2713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000244279101596318,
+      "loss": 2.7879,
+      "step": 2714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002442336143844781,
+      "loss": 3.1705,
+      "step": 2715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002441881128526297,
+      "loss": 2.7674,
+      "step": 2716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002441425970076872,
+      "loss": 2.4948,
+      "step": 2717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002440970668565674,
+      "loss": 2.9374,
+      "step": 2718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024405152240618923,
+      "loss": 2.8806,
+      "step": 2719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024400596366347375,
+      "loss": 2.3638,
+      "step": 2720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002439603906353441,
+      "loss": 2.5756,
+      "step": 2721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002439148033287258,
+      "loss": 3.0677,
+      "step": 2722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024386920175054636,
+      "loss": 2.879,
+      "step": 2723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024382358590773552,
+      "loss": 2.7093,
+      "step": 2724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024377795580722526,
+      "loss": 3.2082,
+      "step": 2725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024373231145594963,
+      "loss": 3.1673,
+      "step": 2726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024368665286084488,
+      "loss": 2.4592,
+      "step": 2727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002436409800288494,
+      "loss": 2.4063,
+      "step": 2728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024359529296690374,
+      "loss": 3.1045,
+      "step": 2729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002435495916819507,
+      "loss": 2.8785,
+      "step": 2730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002435038761809351,
+      "loss": 2.7617,
+      "step": 2731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024345814647080407,
+      "loss": 2.6416,
+      "step": 2732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024341240255850683,
+      "loss": 2.3562,
+      "step": 2733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024336664445099467,
+      "loss": 2.9664,
+      "step": 2734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002433208721552212,
+      "loss": 3.2331,
+      "step": 2735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024327508567814212,
+      "loss": 2.5301,
+      "step": 2736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002432292850267152,
+      "loss": 2.8517,
+      "step": 2737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024318347020790048,
+      "loss": 2.8003,
+      "step": 2738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002431376412286601,
+      "loss": 2.4621,
+      "step": 2739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024309179809595837,
+      "loss": 1.8629,
+      "step": 2740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002430459408167618,
+      "loss": 2.3968,
+      "step": 2741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024300006939803888,
+      "loss": 2.5516,
+      "step": 2742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002429541838467604,
+      "loss": 2.4539,
+      "step": 2743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024290828416989935,
+      "loss": 2.8267,
+      "step": 2744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002428623703744306,
+      "loss": 3.3071,
+      "step": 2745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024281644246733152,
+      "loss": 2.7774,
+      "step": 2746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024277050045558134,
+      "loss": 2.6057,
+      "step": 2747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002427245443461616,
+      "loss": 2.9519,
+      "step": 2748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024267857414605584,
+      "loss": 3.002,
+      "step": 2749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002426325898622499,
+      "loss": 2.4459,
+      "step": 2750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024258659150173167,
+      "loss": 3.2301,
+      "step": 2751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024254057907149114,
+      "loss": 2.4859,
+      "step": 2752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024249455257852054,
+      "loss": 2.5441,
+      "step": 2753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002424485120298142,
+      "loss": 2.7307,
+      "step": 2754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024240245743236853,
+      "loss": 3.2462,
+      "step": 2755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024235638879318208,
+      "loss": 2.702,
+      "step": 2756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024231030611925568,
+      "loss": 2.8012,
+      "step": 2757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024226420941759216,
+      "loss": 3.0712,
+      "step": 2758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024221809869519643,
+      "loss": 2.5293,
+      "step": 2759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024217197395907564,
+      "loss": 2.7975,
+      "step": 2760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002421258352162391,
+      "loss": 3.0153,
+      "step": 2761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002420796824736981,
+      "loss": 2.6766,
+      "step": 2762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024203351573846618,
+      "loss": 1.5174,
+      "step": 2763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024198733501755898,
+      "loss": 2.5845,
+      "step": 2764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002419411403179942,
+      "loss": 2.6194,
+      "step": 2765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002418949316467918,
+      "loss": 1.9036,
+      "step": 2766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002418487090109737,
+      "loss": 2.8133,
+      "step": 2767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002418024724175641,
+      "loss": 2.8394,
+      "step": 2768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002417562218735892,
+      "loss": 3.1027,
+      "step": 2769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024170995738607734,
+      "loss": 2.9028,
+      "step": 2770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024166367896205908,
+      "loss": 2.9354,
+      "step": 2771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024161738660856695,
+      "loss": 2.8226,
+      "step": 2772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002415710803326357,
+      "loss": 2.5885,
+      "step": 2773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024152476014130216,
+      "loss": 2.1625,
+      "step": 2774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024147842604160528,
+      "loss": 2.3919,
+      "step": 2775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024143207804058612,
+      "loss": 1.3306,
+      "step": 2776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002413857161452878,
+      "loss": 5.113,
+      "step": 2777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024133934036275572,
+      "loss": 2.1188,
+      "step": 2778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002412929507000372,
+      "loss": 2.6736,
+      "step": 2779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024124654716418172,
+      "loss": 2.7847,
+      "step": 2780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024120012976224097,
+      "loss": 2.2488,
+      "step": 2781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024115369850126865,
+      "loss": 3.3295,
+      "step": 2782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002411072533883205,
+      "loss": 3.0383,
+      "step": 2783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024106079443045455,
+      "loss": 2.8467,
+      "step": 2784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002410143216347309,
+      "loss": 2.4817,
+      "step": 2785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024096783500821147,
+      "loss": 2.1438,
+      "step": 2786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002409213345579607,
+      "loss": 2.2693,
+      "step": 2787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002408748202910448,
+      "loss": 2.3002,
+      "step": 2788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024082829221453234,
+      "loss": 2.3269,
+      "step": 2789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002407817503354937,
+      "loss": 2.6159,
+      "step": 2790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024073519466100165,
+      "loss": 2.6398,
+      "step": 2791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002406886251981309,
+      "loss": 2.6726,
+      "step": 2792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024064204195395822,
+      "loss": 2.6985,
+      "step": 2793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024059544493556257,
+      "loss": 2.7173,
+      "step": 2794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024054883415002497,
+      "loss": 3.3832,
+      "step": 2795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002405022096044285,
+      "loss": 2.9652,
+      "step": 2796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002404555713058584,
+      "loss": 2.5935,
+      "step": 2797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024040891926140192,
+      "loss": 2.2701,
+      "step": 2798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024036225347814848,
+      "loss": 2.3545,
+      "step": 2799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002403155739631895,
+      "loss": 2.3112,
+      "step": 2800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024026888072361854,
+      "loss": 0.9874,
+      "step": 2801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024022217376653128,
+      "loss": 2.982,
+      "step": 2802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024017545309902542,
+      "loss": 2.4246,
+      "step": 2803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024012871872820072,
+      "loss": 2.3412,
+      "step": 2804
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002400819706611591,
+      "loss": 2.3553,
+      "step": 2805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00024003520890500458,
+      "loss": 2.3994,
+      "step": 2806
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023998843346684313,
+      "loss": 2.5846,
+      "step": 2807
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023994164435378295,
+      "loss": 2.8488,
+      "step": 2808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023989484157293416,
+      "loss": 3.0923,
+      "step": 2809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023984802513140916,
+      "loss": 2.5332,
+      "step": 2810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023980119503632226,
+      "loss": 2.2909,
+      "step": 2811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023975435129478976,
+      "loss": 2.5546,
+      "step": 2812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002397074939139304,
+      "loss": 2.4924,
+      "step": 2813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002396606229008646,
+      "loss": 2.5474,
+      "step": 2814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023961373826271508,
+      "loss": 2.9166,
+      "step": 2815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023956684000660657,
+      "loss": 3.1214,
+      "step": 2816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023951992813966576,
+      "loss": 2.5899,
+      "step": 2817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023947300266902166,
+      "loss": 2.5411,
+      "step": 2818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002394260636018051,
+      "loss": 2.84,
+      "step": 2819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023937911094514913,
+      "loss": 2.6372,
+      "step": 2820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023933214470618871,
+      "loss": 2.783,
+      "step": 2821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002392851648920611,
+      "loss": 2.9539,
+      "step": 2822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023923817150990541,
+      "loss": 2.9005,
+      "step": 2823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023919116456686288,
+      "loss": 2.8024,
+      "step": 2824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023914414407007686,
+      "loss": 2.2426,
+      "step": 2825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002390971100266927,
+      "loss": 2.5834,
+      "step": 2826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002390500624438578,
+      "loss": 2.8239,
+      "step": 2827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023900300132872167,
+      "loss": 2.0533,
+      "step": 2828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023895592668843587,
+      "loss": 2.2351,
+      "step": 2829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023890883853015395,
+      "loss": 2.8577,
+      "step": 2830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023886173686103163,
+      "loss": 3.1788,
+      "step": 2831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023881462168822654,
+      "loss": 2.5999,
+      "step": 2832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002387674930188985,
+      "loss": 2.7144,
+      "step": 2833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023872035086020923,
+      "loss": 2.6882,
+      "step": 2834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023867319521932268,
+      "loss": 3.4996,
+      "step": 2835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023862602610340468,
+      "loss": 2.3019,
+      "step": 2836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023857884351962324,
+      "loss": 1.392,
+      "step": 2837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023853164747514836,
+      "loss": 2.7878,
+      "step": 2838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023848443797715207,
+      "loss": 2.9086,
+      "step": 2839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023843721503280843,
+      "loss": 2.5218,
+      "step": 2840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023838997864929356,
+      "loss": 2.6297,
+      "step": 2841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023834272883378575,
+      "loss": 2.6532,
+      "step": 2842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002382954655934651,
+      "loss": 3.1029,
+      "step": 2843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002382481889355139,
+      "loss": 2.4999,
+      "step": 2844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023820089886711647,
+      "loss": 2.5196,
+      "step": 2845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023815359539545913,
+      "loss": 2.4617,
+      "step": 2846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023810627852773023,
+      "loss": 2.4569,
+      "step": 2847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023805894827112019,
+      "loss": 2.4547,
+      "step": 2848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023801160463282145,
+      "loss": 1.2875,
+      "step": 2849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023796424762002853,
+      "loss": 2.505,
+      "step": 2850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023791687723993783,
+      "loss": 1.9571,
+      "step": 2851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000237869493499748,
+      "loss": 2.551,
+      "step": 2852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023782209640665954,
+      "loss": 3.0589,
+      "step": 2853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023777468596787506,
+      "loss": 3.2028,
+      "step": 2854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002377272621905992,
+      "loss": 2.5335,
+      "step": 2855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002376798250820386,
+      "loss": 2.2462,
+      "step": 2856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023763237464940196,
+      "loss": 2.9056,
+      "step": 2857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023758491089989996,
+      "loss": 2.3989,
+      "step": 2858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002375374338407453,
+      "loss": 2.457,
+      "step": 2859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002374899434791528,
+      "loss": 3.1265,
+      "step": 2860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023744243982233923,
+      "loss": 2.2138,
+      "step": 2861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023739492287752327,
+      "loss": 2.4619,
+      "step": 2862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023734739265192584,
+      "loss": 2.9554,
+      "step": 2863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023729984915276973,
+      "loss": 3.0967,
+      "step": 2864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023725229238727975,
+      "loss": 2.053,
+      "step": 2865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023720472236268285,
+      "loss": 2.471,
+      "step": 2866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023715713908620785,
+      "loss": 2.1949,
+      "step": 2867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023710954256508563,
+      "loss": 2.6434,
+      "step": 2868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002370619328065491,
+      "loss": 2.636,
+      "step": 2869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002370143098178332,
+      "loss": 3.1978,
+      "step": 2870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023696667360617485,
+      "loss": 2.2547,
+      "step": 2871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023691902417881294,
+      "loss": 2.9926,
+      "step": 2872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023687136154298846,
+      "loss": 2.4313,
+      "step": 2873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023682368570594433,
+      "loss": 2.7319,
+      "step": 2874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023677599667492558,
+      "loss": 2.3164,
+      "step": 2875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023672829445717906,
+      "loss": 3.2414,
+      "step": 2876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023668057905995379,
+      "loss": 2.3844,
+      "step": 2877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023663285049050075,
+      "loss": 2.1477,
+      "step": 2878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023658510875607287,
+      "loss": 2.3934,
+      "step": 2879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023653735386392518,
+      "loss": 2.9779,
+      "step": 2880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023648958582131463,
+      "loss": 2.2712,
+      "step": 2881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023644180463550012,
+      "loss": 2.7583,
+      "step": 2882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023639401031374269,
+      "loss": 2.203,
+      "step": 2883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002363462028633053,
+      "loss": 2.9533,
+      "step": 2884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023629838229145288,
+      "loss": 2.9346,
+      "step": 2885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023625054860545235,
+      "loss": 2.9178,
+      "step": 2886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002362027018125727,
+      "loss": 2.51,
+      "step": 2887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002361548419200849,
+      "loss": 2.2944,
+      "step": 2888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023610696893526176,
+      "loss": 2.9827,
+      "step": 2889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023605908286537832,
+      "loss": 1.9514,
+      "step": 2890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023601118371771143,
+      "loss": 2.9215,
+      "step": 2891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023596327149953995,
+      "loss": 2.6133,
+      "step": 2892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023591534621814476,
+      "loss": 2.8023,
+      "step": 2893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023586740788080878,
+      "loss": 2.5288,
+      "step": 2894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023581945649481687,
+      "loss": 2.3756,
+      "step": 2895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023577149206745575,
+      "loss": 2.9708,
+      "step": 2896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002357235146060143,
+      "loss": 2.7388,
+      "step": 2897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023567552411778335,
+      "loss": 2.5513,
+      "step": 2898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023562752061005558,
+      "loss": 2.5708,
+      "step": 2899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023557950409012584,
+      "loss": 2.4383,
+      "step": 2900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023553147456529077,
+      "loss": 2.8557,
+      "step": 2901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002354834320428491,
+      "loss": 2.3295,
+      "step": 2902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023543537653010157,
+      "loss": 2.7162,
+      "step": 2903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002353873080343507,
+      "loss": 2.3141,
+      "step": 2904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023533922656290127,
+      "loss": 2.2812,
+      "step": 2905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023529113212305975,
+      "loss": 2.628,
+      "step": 2906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023524302472213482,
+      "loss": 2.4377,
+      "step": 2907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023519490436743688,
+      "loss": 2.8699,
+      "step": 2908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002351467710662786,
+      "loss": 2.6102,
+      "step": 2909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023509862482597432,
+      "loss": 2.8457,
+      "step": 2910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002350504656538405,
+      "loss": 2.8979,
+      "step": 2911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023500229355719564,
+      "loss": 2.553,
+      "step": 2912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023495410854336,
+      "loss": 3.017,
+      "step": 2913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000234905910619656,
+      "loss": 2.3808,
+      "step": 2914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023485769979340784,
+      "loss": 3.1231,
+      "step": 2915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023480947607194182,
+      "loss": 2.389,
+      "step": 2916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002347612394625862,
+      "loss": 2.5453,
+      "step": 2917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002347129899726711,
+      "loss": 2.9394,
+      "step": 2918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023466472760952863,
+      "loss": 2.8946,
+      "step": 2919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023461645238049292,
+      "loss": 2.694,
+      "step": 2920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023456816429290003,
+      "loss": 2.2785,
+      "step": 2921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023451986335408788,
+      "loss": 2.7896,
+      "step": 2922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023447154957139646,
+      "loss": 2.8055,
+      "step": 2923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023442322295216768,
+      "loss": 2.6999,
+      "step": 2924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002343748835037454,
+      "loss": 2.5287,
+      "step": 2925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002343265312334754,
+      "loss": 2.3587,
+      "step": 2926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002342781661487054,
+      "loss": 2.7997,
+      "step": 2927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002342297882567852,
+      "loss": 2.2389,
+      "step": 2928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023418139756506632,
+      "loss": 2.2206,
+      "step": 2929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023413299408090238,
+      "loss": 2.6412,
+      "step": 2930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023408457781164894,
+      "loss": 2.9647,
+      "step": 2931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023403614876466348,
+      "loss": 2.5516,
+      "step": 2932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002339877069473054,
+      "loss": 2.295,
+      "step": 2933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023393925236693606,
+      "loss": 2.5513,
+      "step": 2934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023389078503091874,
+      "loss": 2.8466,
+      "step": 2935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023384230494661863,
+      "loss": 3.1067,
+      "step": 2936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023379381212140302,
+      "loss": 2.2793,
+      "step": 2937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023374530656264096,
+      "loss": 2.6221,
+      "step": 2938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023369678827770348,
+      "loss": 2.1661,
+      "step": 2939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023364825727396354,
+      "loss": 3.055,
+      "step": 2940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002335997135587961,
+      "loss": 2.9591,
+      "step": 2941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023355115713957796,
+      "loss": 2.4977,
+      "step": 2942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023350258802368794,
+      "loss": 2.7269,
+      "step": 2943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023345400621850666,
+      "loss": 2.4434,
+      "step": 2944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023340541173141682,
+      "loss": 2.8052,
+      "step": 2945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023335680456980295,
+      "loss": 2.6834,
+      "step": 2946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023330818474105158,
+      "loss": 2.7813,
+      "step": 2947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023325955225255102,
+      "loss": 2.628,
+      "step": 2948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023321090711169167,
+      "loss": 2.8415,
+      "step": 2949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023316224932586578,
+      "loss": 3.1962,
+      "step": 2950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002331135789024675,
+      "loss": 3.481,
+      "step": 2951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023306489584889298,
+      "loss": 2.7573,
+      "step": 2952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023301620017254017,
+      "loss": 2.7526,
+      "step": 2953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023296749188080903,
+      "loss": 2.4488,
+      "step": 2954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023291877098110145,
+      "loss": 2.3619,
+      "step": 2955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023287003748082114,
+      "loss": 2.5861,
+      "step": 2956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023282129138737378,
+      "loss": 2.3104,
+      "step": 2957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000232772532708167,
+      "loss": 1.9745,
+      "step": 2958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002327237614506103,
+      "loss": 2.9421,
+      "step": 2959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023267497762211509,
+      "loss": 2.7411,
+      "step": 2960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002326261812300947,
+      "loss": 2.9333,
+      "step": 2961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002325773722819644,
+      "loss": 2.2316,
+      "step": 2962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002325285507851413,
+      "loss": 2.4721,
+      "step": 2963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023247971674704446,
+      "loss": 2.7373,
+      "step": 2964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023243087017509482,
+      "loss": 3.0964,
+      "step": 2965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023238201107671528,
+      "loss": 2.105,
+      "step": 2966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002323331394593306,
+      "loss": 3.2497,
+      "step": 2967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023228425533036745,
+      "loss": 2.762,
+      "step": 2968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023223535869725442,
+      "loss": 3.0527,
+      "step": 2969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023218644956742192,
+      "loss": 2.9975,
+      "step": 2970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002321375279483024,
+      "loss": 3.2973,
+      "step": 2971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023208859384733004,
+      "loss": 3.0777,
+      "step": 2972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023203964727194103,
+      "loss": 2.9094,
+      "step": 2973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023199068822957352,
+      "loss": 3.2144,
+      "step": 2974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023194171672766736,
+      "loss": 3.2043,
+      "step": 2975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023189273277366446,
+      "loss": 2.8019,
+      "step": 2976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023184373637500855,
+      "loss": 2.5313,
+      "step": 2977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023179472753914523,
+      "loss": 2.7666,
+      "step": 2978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023174570627352203,
+      "loss": 2.9094,
+      "step": 2979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023169667258558842,
+      "loss": 2.2028,
+      "step": 2980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002316476264827956,
+      "loss": 2.102,
+      "step": 2981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023159856797259687,
+      "loss": 3.1844,
+      "step": 2982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023154949706244722,
+      "loss": 2.7544,
+      "step": 2983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023150041375980362,
+      "loss": 2.68,
+      "step": 2984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002314513180721249,
+      "loss": 2.1721,
+      "step": 2985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002314022100068718,
+      "loss": 2.1817,
+      "step": 2986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023135308957150696,
+      "loss": 2.7281,
+      "step": 2987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023130395677349477,
+      "loss": 2.7309,
+      "step": 2988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002312548116203017,
+      "loss": 2.7223,
+      "step": 2989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023120565411939584,
+      "loss": 3.1032,
+      "step": 2990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023115648427824743,
+      "loss": 2.3789,
+      "step": 2991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023110730210432843,
+      "loss": 3.0314,
+      "step": 2992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023105810760511264,
+      "loss": 2.2588,
+      "step": 2993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002310089007880759,
+      "loss": 2.9962,
+      "step": 2994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002309596816606958,
+      "loss": 3.3206,
+      "step": 2995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002309104502304517,
+      "loss": 2.8657,
+      "step": 2996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002308612065048251,
+      "loss": 2.5826,
+      "step": 2997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023081195049129915,
+      "loss": 3.2743,
+      "step": 2998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002307626821973589,
+      "loss": 2.9605,
+      "step": 2999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023071340163049135,
+      "loss": 2.2397,
+      "step": 3000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002306641087981853,
+      "loss": 2.5313,
+      "step": 3001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023061480370793145,
+      "loss": 2.7993,
+      "step": 3002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002305654863672223,
+      "loss": 2.2684,
+      "step": 3003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002305161567835523,
+      "loss": 2.5762,
+      "step": 3004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023046681496441767,
+      "loss": 2.641,
+      "step": 3005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002304174609173166,
+      "loss": 2.5587,
+      "step": 3006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023036809464974895,
+      "loss": 2.1724,
+      "step": 3007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023031871616921665,
+      "loss": 2.5902,
+      "step": 3008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002302693254832234,
+      "loss": 2.2897,
+      "step": 3009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023021992259927471,
+      "loss": 3.1139,
+      "step": 3010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000230170507524878,
+      "loss": 2.8079,
+      "step": 3011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023012108026754257,
+      "loss": 2.5521,
+      "step": 3012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023007164083477943,
+      "loss": 2.2725,
+      "step": 3013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00023002218923410162,
+      "loss": 2.0922,
+      "step": 3014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002299727254730239,
+      "loss": 2.4265,
+      "step": 3015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022992324955906294,
+      "loss": 3.4073,
+      "step": 3016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022987376149973728,
+      "loss": 2.4314,
+      "step": 3017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002298242613025672,
+      "loss": 3.0676,
+      "step": 3018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002297747489750749,
+      "loss": 2.6647,
+      "step": 3019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022972522452478447,
+      "loss": 2.4208,
+      "step": 3020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002296756879592217,
+      "loss": 2.5889,
+      "step": 3021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022962613928591437,
+      "loss": 2.7281,
+      "step": 3022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022957657851239204,
+      "loss": 2.4688,
+      "step": 3023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022952700564618611,
+      "loss": 1.1785,
+      "step": 3024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022947742069482976,
+      "loss": 2.2944,
+      "step": 3025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022942782366585813,
+      "loss": 3.0349,
+      "step": 3026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022937821456680806,
+      "loss": 2.0488,
+      "step": 3027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022932859340521833,
+      "loss": 2.7166,
+      "step": 3028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002292789601886295,
+      "loss": 2.2718,
+      "step": 3029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022922931492458398,
+      "loss": 2.9237,
+      "step": 3030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022917965762062601,
+      "loss": 2.5105,
+      "step": 3031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002291299882843017,
+      "loss": 2.7367,
+      "step": 3032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022908030692315884,
+      "loss": 2.5958,
+      "step": 3033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002290306135447472,
+      "loss": 2.1083,
+      "step": 3034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002289809081566184,
+      "loss": 2.6977,
+      "step": 3035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022893119076632568,
+      "loss": 1.929,
+      "step": 3036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022888146138142428,
+      "loss": 2.8158,
+      "step": 3037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022883172000947132,
+      "loss": 2.68,
+      "step": 3038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002287819666580255,
+      "loss": 2.9044,
+      "step": 3039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022873220133464757,
+      "loss": 3.3497,
+      "step": 3040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002286824240469,
+      "loss": 2.5521,
+      "step": 3041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022863263480234702,
+      "loss": 2.7792,
+      "step": 3042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002285828336085548,
+      "loss": 3.185,
+      "step": 3043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002285330204730913,
+      "loss": 2.9155,
+      "step": 3044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022848319540352623,
+      "loss": 2.8056,
+      "step": 3045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022843335840743115,
+      "loss": 3.1478,
+      "step": 3046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002283835094923795,
+      "loss": 2.1546,
+      "step": 3047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002283336486659463,
+      "loss": 2.9182,
+      "step": 3048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022828377593570872,
+      "loss": 2.8488,
+      "step": 3049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022823389130924543,
+      "loss": 2.3215,
+      "step": 3050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022818399479413711,
+      "loss": 2.9288,
+      "step": 3051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022813408639796618,
+      "loss": 2.4475,
+      "step": 3052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002280841661283168,
+      "loss": 1.1541,
+      "step": 3053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00022803423399277506,
+      "loss": 2.3922,
+      "step": 3054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002279842899989288,
+      "loss": 2.8339,
+      "step": 3055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002279343341543676,
+      "loss": 3.0589,
+      "step": 3056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022788436646668288,
+      "loss": 2.8961,
+      "step": 3057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022783438694346794,
+      "loss": 3.3837,
+      "step": 3058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022778439559231775,
+      "loss": 3.1244,
+      "step": 3059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002277343924208292,
+      "loss": 2.4954,
+      "step": 3060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022768437743660088,
+      "loss": 2.8306,
+      "step": 3061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002276343506472332,
+      "loss": 2.5145,
+      "step": 3062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022758431206032836,
+      "loss": 2.6926,
+      "step": 3063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002275342616834904,
+      "loss": 2.9331,
+      "step": 3064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022748419952432514,
+      "loss": 2.7672,
+      "step": 3065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022743412559044006,
+      "loss": 2.3614,
+      "step": 3066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022738403988944468,
+      "loss": 2.5715,
+      "step": 3067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022733394242895006,
+      "loss": 3.1692,
+      "step": 3068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022728383321656923,
+      "loss": 2.7469,
+      "step": 3069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022723371225991684,
+      "loss": 2.4554,
+      "step": 3070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022718357956660946,
+      "loss": 3.0478,
+      "step": 3071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022713343514426545,
+      "loss": 2.9541,
+      "step": 3072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002270832790005048,
+      "loss": 2.4138,
+      "step": 3073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002270331111429495,
+      "loss": 2.3589,
+      "step": 3074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022698293157922308,
+      "loss": 2.4463,
+      "step": 3075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022693274031695103,
+      "loss": 2.8879,
+      "step": 3076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022688253736376053,
+      "loss": 0.9165,
+      "step": 3077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022683232272728058,
+      "loss": 2.2493,
+      "step": 3078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000226782096415142,
+      "loss": 2.9447,
+      "step": 3079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022673185843497724,
+      "loss": 3.0735,
+      "step": 3080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002266816087944206,
+      "loss": 2.6702,
+      "step": 3081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022663134750110816,
+      "loss": 2.8164,
+      "step": 3082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022658107456267784,
+      "loss": 2.1541,
+      "step": 3083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002265307899867692,
+      "loss": 2.3881,
+      "step": 3084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022648049378102365,
+      "loss": 2.6641,
+      "step": 3085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022643018595308435,
+      "loss": 2.9583,
+      "step": 3086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002263798665105962,
+      "loss": 2.6067,
+      "step": 3087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022632953546120591,
+      "loss": 2.7056,
+      "step": 3088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002262791928125619,
+      "loss": 2.767,
+      "step": 3089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022622883857231443,
+      "loss": 2.3383,
+      "step": 3090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022617847274811537,
+      "loss": 2.6289,
+      "step": 3091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022612809534761858,
+      "loss": 3.035,
+      "step": 3092
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002260777063784795,
+      "loss": 2.6334,
+      "step": 3093
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022602730584835539,
+      "loss": 2.5247,
+      "step": 3094
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002259768937649052,
+      "loss": 2.5778,
+      "step": 3095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022592647013578977,
+      "loss": 2.2507,
+      "step": 3096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002258760349686716,
+      "loss": 3.3482,
+      "step": 3097
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022582558827121497,
+      "loss": 2.1749,
+      "step": 3098
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022577513005108584,
+      "loss": 3.0712,
+      "step": 3099
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022572466031595205,
+      "loss": 2.3273,
+      "step": 3100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002256741790734831,
+      "loss": 2.8962,
+      "step": 3101
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022562368633135028,
+      "loss": 2.4135,
+      "step": 3102
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022557318209722664,
+      "loss": 3.3722,
+      "step": 3103
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022552266637878681,
+      "loss": 2.4545,
+      "step": 3104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002254721391837075,
+      "loss": 2.8321,
+      "step": 3105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002254216005196668,
+      "loss": 3.0012,
+      "step": 3106
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022537105039434482,
+      "loss": 2.9046,
+      "step": 3107
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022532048881542324,
+      "loss": 3.1964,
+      "step": 3108
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022526991579058556,
+      "loss": 2.6539,
+      "step": 3109
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022521933132751702,
+      "loss": 2.6221,
+      "step": 3110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002251687354339045,
+      "loss": 3.3889,
+      "step": 3111
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002251181281174368,
+      "loss": 2.7891,
+      "step": 3112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022506750938580428,
+      "loss": 2.5277,
+      "step": 3113
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002250168792466992,
+      "loss": 2.2903,
+      "step": 3114
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022496623770781532,
+      "loss": 2.5865,
+      "step": 3115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022491558477684837,
+      "loss": 2.7749,
+      "step": 3116
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002248649204614957,
+      "loss": 2.1697,
+      "step": 3117
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002248142447694564,
+      "loss": 2.9416,
+      "step": 3118
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022476355770843125,
+      "loss": 2.9494,
+      "step": 3119
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022471285928612287,
+      "loss": 2.6002,
+      "step": 3120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022466214951023553,
+      "loss": 2.9064,
+      "step": 3121
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002246114283884752,
+      "loss": 3.2247,
+      "step": 3122
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022456069592854955,
+      "loss": 2.5875,
+      "step": 3123
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022450995213816813,
+      "loss": 2.7436,
+      "step": 3124
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022445919702504207,
+      "loss": 2.6141,
+      "step": 3125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002244084305968843,
+      "loss": 2.5347,
+      "step": 3126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022435765286140934,
+      "loss": 2.2841,
+      "step": 3127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022430686382633356,
+      "loss": 2.8399,
+      "step": 3128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022425606349937507,
+      "loss": 2.6932,
+      "step": 3129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022420525188825355,
+      "loss": 2.8262,
+      "step": 3130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022415442900069053,
+      "loss": 2.4597,
+      "step": 3131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022410359484440915,
+      "loss": 2.5837,
+      "step": 3132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002240527494271344,
+      "loss": 2.587,
+      "step": 3133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002240018927565928,
+      "loss": 3.0031,
+      "step": 3134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022395102484051274,
+      "loss": 2.9438,
+      "step": 3135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002239001456866242,
+      "loss": 2.3332,
+      "step": 3136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022384925530265898,
+      "loss": 3.2454,
+      "step": 3137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022379835369635046,
+      "loss": 2.285,
+      "step": 3138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022374744087543387,
+      "loss": 2.6899,
+      "step": 3139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000223696516847646,
+      "loss": 2.9539,
+      "step": 3140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022364558162072542,
+      "loss": 2.9336,
+      "step": 3141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022359463520241243,
+      "loss": 2.8078,
+      "step": 3142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000223543677600449,
+      "loss": 2.9051,
+      "step": 3143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002234927088225787,
+      "loss": 2.8755,
+      "step": 3144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022344172887654699,
+      "loss": 3.0225,
+      "step": 3145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022339073777010086,
+      "loss": 2.865,
+      "step": 3146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022333973551098913,
+      "loss": 2.7458,
+      "step": 3147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002232887221069622,
+      "loss": 2.7436,
+      "step": 3148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022323769756577228,
+      "loss": 2.5184,
+      "step": 3149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022318666189517308,
+      "loss": 2.5237,
+      "step": 3150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002231356151029203,
+      "loss": 1.5556,
+      "step": 3151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022308455719677095,
+      "loss": 2.6909,
+      "step": 3152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022303348818448405,
+      "loss": 2.8524,
+      "step": 3153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022298240807382025,
+      "loss": 2.3327,
+      "step": 3154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002229313168725417,
+      "loss": 3.0905,
+      "step": 3155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022288021458841245,
+      "loss": 2.1818,
+      "step": 3156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002228291012291981,
+      "loss": 2.7544,
+      "step": 3157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022277797680266603,
+      "loss": 2.1012,
+      "step": 3158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022272684131658521,
+      "loss": 2.7389,
+      "step": 3159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022267569477872636,
+      "loss": 2.8744,
+      "step": 3160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022262453719686185,
+      "loss": 2.6848,
+      "step": 3161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022257336857876573,
+      "loss": 2.6086,
+      "step": 3162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022252218893221374,
+      "loss": 2.426,
+      "step": 3163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022247099826498324,
+      "loss": 3.0857,
+      "step": 3164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022241979658485338,
+      "loss": 2.6858,
+      "step": 3165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022236858389960485,
+      "loss": 2.4386,
+      "step": 3166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002223173602170201,
+      "loss": 3.3952,
+      "step": 3167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002222661255448833,
+      "loss": 2.5708,
+      "step": 3168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022221487989098005,
+      "loss": 2.2638,
+      "step": 3169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022216362326309792,
+      "loss": 2.8235,
+      "step": 3170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022211235566902595,
+      "loss": 2.7041,
+      "step": 3171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022206107711655496,
+      "loss": 2.7536,
+      "step": 3172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022200978761347735,
+      "loss": 2.567,
+      "step": 3173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022195848716758721,
+      "loss": 2.9607,
+      "step": 3174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022190717578668033,
+      "loss": 2.4987,
+      "step": 3175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022185585347855413,
+      "loss": 3.0234,
+      "step": 3176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002218045202510077,
+      "loss": 2.7782,
+      "step": 3177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022175317611184173,
+      "loss": 2.1999,
+      "step": 3178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002217018210688587,
+      "loss": 3.1225,
+      "step": 3179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002216504551298626,
+      "loss": 2.7544,
+      "step": 3180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002215990783026592,
+      "loss": 2.9307,
+      "step": 3181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022154769059505586,
+      "loss": 3.0873,
+      "step": 3182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022149629201486157,
+      "loss": 2.4797,
+      "step": 3183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000221444882569887,
+      "loss": 2.113,
+      "step": 3184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022139346226794453,
+      "loss": 2.9313,
+      "step": 3185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002213420311168481,
+      "loss": 2.7687,
+      "step": 3186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002212905891244133,
+      "loss": 2.5719,
+      "step": 3187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022123913629845747,
+      "loss": 2.1899,
+      "step": 3188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002211876726467995,
+      "loss": 1.9271,
+      "step": 3189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022113619817725993,
+      "loss": 2.3169,
+      "step": 3190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022108471289766102,
+      "loss": 3.0672,
+      "step": 3191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022103321681582653,
+      "loss": 2.8313,
+      "step": 3192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022098170993958206,
+      "loss": 2.616,
+      "step": 3193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022093019227675467,
+      "loss": 2.4975,
+      "step": 3194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022087866383517315,
+      "loss": 1.9873,
+      "step": 3195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022082712462266792,
+      "loss": 2.6427,
+      "step": 3196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022077557464707104,
+      "loss": 2.6192,
+      "step": 3197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022072401391621613,
+      "loss": 2.9728,
+      "step": 3198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022067244243793854,
+      "loss": 2.753,
+      "step": 3199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022062086022007522,
+      "loss": 2.467,
+      "step": 3200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022056926727046476,
+      "loss": 2.3661,
+      "step": 3201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022051766359694737,
+      "loss": 2.5196,
+      "step": 3202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002204660492073649,
+      "loss": 1.1156,
+      "step": 3203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022041442410956078,
+      "loss": 2.1419,
+      "step": 3204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002203627883113801,
+      "loss": 2.7449,
+      "step": 3205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022031114182066968,
+      "loss": 2.5475,
+      "step": 3206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022025948464527776,
+      "loss": 2.44,
+      "step": 3207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022020781679305435,
+      "loss": 2.8273,
+      "step": 3208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022015613827185108,
+      "loss": 2.6079,
+      "step": 3209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022010444908952112,
+      "loss": 2.556,
+      "step": 3210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002200527492539193,
+      "loss": 1.8118,
+      "step": 3211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00022000103877290215,
+      "loss": 2.6836,
+      "step": 3212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021994931765432764,
+      "loss": 2.8582,
+      "step": 3213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021989758590605548,
+      "loss": 2.3972,
+      "step": 3214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021984584353594705,
+      "loss": 2.4085,
+      "step": 3215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021979409055186518,
+      "loss": 2.9238,
+      "step": 3216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021974232696167443,
+      "loss": 2.9119,
+      "step": 3217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021969055277324098,
+      "loss": 2.7612,
+      "step": 3218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021963876799443253,
+      "loss": 2.3381,
+      "step": 3219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021958697263311846,
+      "loss": 2.4121,
+      "step": 3220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021953516669716974,
+      "loss": 2.9344,
+      "step": 3221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021948335019445902,
+      "loss": 3.1271,
+      "step": 3222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021943152313286032,
+      "loss": 2.6475,
+      "step": 3223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021937968552024963,
+      "loss": 1.7912,
+      "step": 3224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021932783736450416,
+      "loss": 2.5135,
+      "step": 3225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021927597867350303,
+      "loss": 2.459,
+      "step": 3226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002192241094551267,
+      "loss": 2.4327,
+      "step": 3227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002191722297172575,
+      "loss": 2.2155,
+      "step": 3228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021912033946777922,
+      "loss": 2.8127,
+      "step": 3229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021906843871457714,
+      "loss": 2.3901,
+      "step": 3230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021901652746553833,
+      "loss": 2.7793,
+      "step": 3231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021896460572855131,
+      "loss": 2.8681,
+      "step": 3232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021891267351150634,
+      "loss": 3.0698,
+      "step": 3233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002188607308222951,
+      "loss": 3.0342,
+      "step": 3234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000218808777668811,
+      "loss": 2.7925,
+      "step": 3235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021875681405894898,
+      "loss": 3.6619,
+      "step": 3236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002187048400006056,
+      "loss": 2.4754,
+      "step": 3237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021865285550167897,
+      "loss": 1.2231,
+      "step": 3238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021860086057006877,
+      "loss": 2.7179,
+      "step": 3239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021854885521367632,
+      "loss": 2.599,
+      "step": 3240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021849683944040452,
+      "loss": 3.1422,
+      "step": 3241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021844481325815782,
+      "loss": 2.5191,
+      "step": 3242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021839277667484226,
+      "loss": 2.6337,
+      "step": 3243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002183407296983655,
+      "loss": 2.7327,
+      "step": 3244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021828867233663674,
+      "loss": 2.9616,
+      "step": 3245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021823660459756672,
+      "loss": 3.365,
+      "step": 3246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002181845264890678,
+      "loss": 2.6744,
+      "step": 3247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021813243801905398,
+      "loss": 3.0056,
+      "step": 3248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021808033919544072,
+      "loss": 2.5756,
+      "step": 3249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021802823002614512,
+      "loss": 2.7164,
+      "step": 3250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021797611051908587,
+      "loss": 2.9175,
+      "step": 3251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021792398068218319,
+      "loss": 2.3051,
+      "step": 3252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021787184052335883,
+      "loss": 2.4282,
+      "step": 3253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021781969005053617,
+      "loss": 2.8412,
+      "step": 3254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021776752927164022,
+      "loss": 2.7982,
+      "step": 3255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021771535819459735,
+      "loss": 2.5578,
+      "step": 3256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021766317682733575,
+      "loss": 2.1839,
+      "step": 3257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021761098517778503,
+      "loss": 2.9526,
+      "step": 3258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021755878325387627,
+      "loss": 3.2093,
+      "step": 3259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021750657106354233,
+      "loss": 2.988,
+      "step": 3260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021745434861471753,
+      "loss": 2.4834,
+      "step": 3261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021740211591533766,
+      "loss": 3.0247,
+      "step": 3262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021734987297334024,
+      "loss": 2.7402,
+      "step": 3263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021729761979666416,
+      "loss": 2.7745,
+      "step": 3264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021724535639325006,
+      "loss": 2.4049,
+      "step": 3265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021719308277104,
+      "loss": 2.6197,
+      "step": 3266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002171407989379776,
+      "loss": 2.372,
+      "step": 3267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021708850490200805,
+      "loss": 2.9299,
+      "step": 3268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021703620067107818,
+      "loss": 2.3253,
+      "step": 3269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021698388625313622,
+      "loss": 3.1734,
+      "step": 3270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021693156165613202,
+      "loss": 2.5597,
+      "step": 3271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021687922688801704,
+      "loss": 2.7203,
+      "step": 3272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021682688195674417,
+      "loss": 2.7311,
+      "step": 3273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021677452687026785,
+      "loss": 2.6604,
+      "step": 3274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002167221616365442,
+      "loss": 2.3963,
+      "step": 3275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021666978626353075,
+      "loss": 2.7092,
+      "step": 3276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021661740075918658,
+      "loss": 2.5784,
+      "step": 3277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002165650051314724,
+      "loss": 2.6714,
+      "step": 3278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021651259938835034,
+      "loss": 2.9525,
+      "step": 3279
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021646018353778417,
+      "loss": 2.6622,
+      "step": 3280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021640775758773913,
+      "loss": 3.0934,
+      "step": 3281
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000216355321546182,
+      "loss": 2.3162,
+      "step": 3282
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021630287542108114,
+      "loss": 2.6031,
+      "step": 3283
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021625041922040643,
+      "loss": 2.6694,
+      "step": 3284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021619795295212924,
+      "loss": 3.2834,
+      "step": 3285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021614547662422248,
+      "loss": 2.2813,
+      "step": 3286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021609299024466063,
+      "loss": 3.0141,
+      "step": 3287
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021604049382141965,
+      "loss": 2.7831,
+      "step": 3288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021598798736247707,
+      "loss": 2.9504,
+      "step": 3289
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021593547087581186,
+      "loss": 2.4594,
+      "step": 3290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021588294436940462,
+      "loss": 1.5438,
+      "step": 3291
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021583040785123745,
+      "loss": 2.4302,
+      "step": 3292
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002157778613292939,
+      "loss": 2.6126,
+      "step": 3293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021572530481155913,
+      "loss": 2.3584,
+      "step": 3294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021567273830601973,
+      "loss": 3.0297,
+      "step": 3295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021562016182066394,
+      "loss": 2.6901,
+      "step": 3296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002155675753634814,
+      "loss": 2.4227,
+      "step": 3297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002155149789424632,
+      "loss": 2.3469,
+      "step": 3298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021546237256560215,
+      "loss": 3.0195,
+      "step": 3299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021540975624089243,
+      "loss": 2.6315,
+      "step": 3300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021535712997632978,
+      "loss": 2.7303,
+      "step": 3301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021530449377991145,
+      "loss": 2.7848,
+      "step": 3302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002152518476596361,
+      "loss": 2.924,
+      "step": 3303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021519919162350408,
+      "loss": 2.8635,
+      "step": 3304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021514652567951716,
+      "loss": 2.6905,
+      "step": 3305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021509384983567852,
+      "loss": 3.0293,
+      "step": 3306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021504116409999302,
+      "loss": 2.8957,
+      "step": 3307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021498846848046684,
+      "loss": 3.058,
+      "step": 3308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021493576298510785,
+      "loss": 3.066,
+      "step": 3309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021488304762192528,
+      "loss": 1.7008,
+      "step": 3310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002148303223989299,
+      "loss": 1.8775,
+      "step": 3311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021477758732413402,
+      "loss": 2.6124,
+      "step": 3312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021472484240555136,
+      "loss": 2.6322,
+      "step": 3313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021467208765119723,
+      "loss": 1.9699,
+      "step": 3314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002146193230690884,
+      "loss": 2.2201,
+      "step": 3315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002145665486672431,
+      "loss": 3.0072,
+      "step": 3316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021451376445368105,
+      "loss": 2.4954,
+      "step": 3317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021446097043642355,
+      "loss": 2.4182,
+      "step": 3318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021440816662349338,
+      "loss": 2.9674,
+      "step": 3319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002143553530229146,
+      "loss": 2.5143,
+      "step": 3320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000214302529642713,
+      "loss": 2.2519,
+      "step": 3321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002142496964909158,
+      "loss": 2.8587,
+      "step": 3322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021419685357555162,
+      "loss": 2.9746,
+      "step": 3323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021414400090465066,
+      "loss": 2.6023,
+      "step": 3324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021409113848624452,
+      "loss": 2.933,
+      "step": 3325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002140382663283664,
+      "loss": 2.6641,
+      "step": 3326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021398538443905083,
+      "loss": 2.6523,
+      "step": 3327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021393249282633393,
+      "loss": 2.8818,
+      "step": 3328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021387959149825325,
+      "loss": 2.934,
+      "step": 3329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002138266804628478,
+      "loss": 2.857,
+      "step": 3330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021377375972815815,
+      "loss": 2.3372,
+      "step": 3331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002137208293022262,
+      "loss": 3.1193,
+      "step": 3332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021366788919309552,
+      "loss": 2.4741,
+      "step": 3333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000213614939408811,
+      "loss": 2.6217,
+      "step": 3334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021356197995741897,
+      "loss": 2.6148,
+      "step": 3335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021350901084696736,
+      "loss": 2.913,
+      "step": 3336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021345603208550547,
+      "loss": 2.7691,
+      "step": 3337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021340304368108417,
+      "loss": 2.8836,
+      "step": 3338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002133500456417557,
+      "loss": 2.8453,
+      "step": 3339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021329703797557374,
+      "loss": 2.8305,
+      "step": 3340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021324402069059355,
+      "loss": 2.1936,
+      "step": 3341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002131909937948718,
+      "loss": 2.7995,
+      "step": 3342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002131379572964665,
+      "loss": 2.3099,
+      "step": 3343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021308491120343735,
+      "loss": 2.5298,
+      "step": 3344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021303185552384534,
+      "loss": 2.6655,
+      "step": 3345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021297879026575294,
+      "loss": 2.5958,
+      "step": 3346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021292571543722415,
+      "loss": 2.9358,
+      "step": 3347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021287263104632432,
+      "loss": 2.6096,
+      "step": 3348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021281953710112033,
+      "loss": 2.3777,
+      "step": 3349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021276643360968047,
+      "loss": 2.5203,
+      "step": 3350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002127133205800745,
+      "loss": 2.7531,
+      "step": 3351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021266019802037369,
+      "loss": 2.4031,
+      "step": 3352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021260706593865056,
+      "loss": 2.4729,
+      "step": 3353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021255392434297936,
+      "loss": 2.7569,
+      "step": 3354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021250077324143555,
+      "loss": 2.4873,
+      "step": 3355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021244761264209612,
+      "loss": 2.5827,
+      "step": 3356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021239444255303957,
+      "loss": 2.816,
+      "step": 3357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021234126298234576,
+      "loss": 3.1656,
+      "step": 3358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021228807393809595,
+      "loss": 2.276,
+      "step": 3359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021223487542837294,
+      "loss": 2.5818,
+      "step": 3360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021218166746126096,
+      "loss": 2.6818,
+      "step": 3361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002121284500448456,
+      "loss": 2.4758,
+      "step": 3362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021207522318721395,
+      "loss": 2.7278,
+      "step": 3363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021202198689645449,
+      "loss": 2.3482,
+      "step": 3364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021196874118065715,
+      "loss": 2.5348,
+      "step": 3365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021191548604791334,
+      "loss": 2.3181,
+      "step": 3366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021186222150631587,
+      "loss": 3.0017,
+      "step": 3367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021180894756395893,
+      "loss": 2.8823,
+      "step": 3368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002117556642289382,
+      "loss": 2.5303,
+      "step": 3369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002117023715093508,
+      "loss": 2.3042,
+      "step": 3370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021164906941329513,
+      "loss": 2.5292,
+      "step": 3371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021159575794887125,
+      "loss": 1.4729,
+      "step": 3372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021154243712418047,
+      "loss": 2.7451,
+      "step": 3373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002114891069473256,
+      "loss": 3.2712,
+      "step": 3374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002114357674264108,
+      "loss": 2.3655,
+      "step": 3375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021138241856954177,
+      "loss": 2.5398,
+      "step": 3376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021132906038482549,
+      "loss": 2.3934,
+      "step": 3377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002112756928803704,
+      "loss": 2.6124,
+      "step": 3378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021122231606428646,
+      "loss": 2.6338,
+      "step": 3379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021116892994468495,
+      "loss": 2.2757,
+      "step": 3380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021111553452967852,
+      "loss": 2.6699,
+      "step": 3381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002110621298273813,
+      "loss": 2.47,
+      "step": 3382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021100871584590886,
+      "loss": 2.717,
+      "step": 3383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021095529259337808,
+      "loss": 2.4456,
+      "step": 3384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002109018600779074,
+      "loss": 2.1783,
+      "step": 3385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021084841830761652,
+      "loss": 2.2886,
+      "step": 3386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021079496729062657,
+      "loss": 2.865,
+      "step": 3387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002107415070350602,
+      "loss": 2.4971,
+      "step": 3388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021068803754904135,
+      "loss": 2.6562,
+      "step": 3389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021063455884069533,
+      "loss": 1.9623,
+      "step": 3390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021058107091814903,
+      "loss": 2.4576,
+      "step": 3391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021052757378953058,
+      "loss": 2.4625,
+      "step": 3392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021047406746296947,
+      "loss": 2.4842,
+      "step": 3393
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002104205519465968,
+      "loss": 2.9357,
+      "step": 3394
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021036702724854496,
+      "loss": 1.4374,
+      "step": 3395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002103134933769476,
+      "loss": 2.1587,
+      "step": 3396
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021025995033993987,
+      "loss": 2.4603,
+      "step": 3397
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021020639814565842,
+      "loss": 2.7357,
+      "step": 3398
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002101528368022412,
+      "loss": 2.9244,
+      "step": 3399
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021009926631782743,
+      "loss": 1.7695,
+      "step": 3400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00021004568670055792,
+      "loss": 2.8832,
+      "step": 3401
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020999209795857476,
+      "loss": 2.4515,
+      "step": 3402
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020993850010002146,
+      "loss": 2.4605,
+      "step": 3403
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002098848931330429,
+      "loss": 2.1175,
+      "step": 3404
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020983127706578533,
+      "loss": 3.1478,
+      "step": 3405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020977765190639643,
+      "loss": 2.7173,
+      "step": 3406
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002097240176630252,
+      "loss": 2.3493,
+      "step": 3407
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020967037434382204,
+      "loss": 1.3279,
+      "step": 3408
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020961672195693885,
+      "loss": 2.9658,
+      "step": 3409
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020956306051052865,
+      "loss": 2.8658,
+      "step": 3410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020950939001274606,
+      "loss": 2.9057,
+      "step": 3411
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000209455710471747,
+      "loss": 2.704,
+      "step": 3412
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020940202189568877,
+      "loss": 1.9019,
+      "step": 3413
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020934832429273003,
+      "loss": 3.0579,
+      "step": 3414
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020929461767103078,
+      "loss": 2.7203,
+      "step": 3415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020924090203875255,
+      "loss": 2.6228,
+      "step": 3416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020918717740405797,
+      "loss": 2.5345,
+      "step": 3417
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020913344377511132,
+      "loss": 2.3357,
+      "step": 3418
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020907970116007803,
+      "loss": 2.6041,
+      "step": 3419
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020902594956712502,
+      "loss": 3.1432,
+      "step": 3420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020897218900442052,
+      "loss": 2.9058,
+      "step": 3421
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002089184194801341,
+      "loss": 2.3572,
+      "step": 3422
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020886464100243685,
+      "loss": 1.5922,
+      "step": 3423
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020881085357950097,
+      "loss": 2.8169,
+      "step": 3424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020875705721950022,
+      "loss": 2.3821,
+      "step": 3425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020870325193060964,
+      "loss": 2.8277,
+      "step": 3426
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002086494377210056,
+      "loss": 2.3502,
+      "step": 3427
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002085956145988659,
+      "loss": 2.6843,
+      "step": 3428
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002085417825723696,
+      "loss": 2.66,
+      "step": 3429
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020848794164969727,
+      "loss": 2.9627,
+      "step": 3430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020843409183903065,
+      "loss": 2.8535,
+      "step": 3431
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002083802331485529,
+      "loss": 2.4445,
+      "step": 3432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020832636558644862,
+      "loss": 2.3594,
+      "step": 3433
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020827248916090358,
+      "loss": 2.4057,
+      "step": 3434
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020821860388010508,
+      "loss": 2.2756,
+      "step": 3435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020816470975224165,
+      "loss": 2.8009,
+      "step": 3436
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020811080678550314,
+      "loss": 2.5429,
+      "step": 3437
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020805689498808094,
+      "loss": 2.4676,
+      "step": 3438
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002080029743681675,
+      "loss": 2.7178,
+      "step": 3439
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020794904493395675,
+      "loss": 2.6681,
+      "step": 3440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020789510669364406,
+      "loss": 2.848,
+      "step": 3441
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020784115965542604,
+      "loss": 2.0992,
+      "step": 3442
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020778720382750047,
+      "loss": 2.9169,
+      "step": 3443
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020773323921806684,
+      "loss": 2.8029,
+      "step": 3444
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002076792658353256,
+      "loss": 2.6412,
+      "step": 3445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002076252836874788,
+      "loss": 3.0855,
+      "step": 3446
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002075712927827297,
+      "loss": 3.1832,
+      "step": 3447
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020751729312928286,
+      "loss": 2.534,
+      "step": 3448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020746328473534434,
+      "loss": 2.5423,
+      "step": 3449
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020740926760912128,
+      "loss": 2.6196,
+      "step": 3450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020735524175882235,
+      "loss": 2.4714,
+      "step": 3451
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020730120719265744,
+      "loss": 2.487,
+      "step": 3452
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020724716391883782,
+      "loss": 2.2766,
+      "step": 3453
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020719311194557604,
+      "loss": 2.726,
+      "step": 3454
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020713905128108602,
+      "loss": 2.8304,
+      "step": 3455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020708498193358296,
+      "loss": 2.819,
+      "step": 3456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020703090391128336,
+      "loss": 2.8979,
+      "step": 3457
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020697681722240508,
+      "loss": 2.3173,
+      "step": 3458
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020692272187516732,
+      "loss": 2.1561,
+      "step": 3459
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020686861787779057,
+      "loss": 2.7086,
+      "step": 3460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002068145052384966,
+      "loss": 2.4522,
+      "step": 3461
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020676038396550848,
+      "loss": 2.1363,
+      "step": 3462
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002067062540670507,
+      "loss": 2.9836,
+      "step": 3463
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020665211555134902,
+      "loss": 2.5721,
+      "step": 3464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020659796842663038,
+      "loss": 3.0482,
+      "step": 3465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020654381270112317,
+      "loss": 2.7299,
+      "step": 3466
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020648964838305712,
+      "loss": 2.5337,
+      "step": 3467
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002064354754806631,
+      "loss": 2.6848,
+      "step": 3468
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020638129400217344,
+      "loss": 2.4427,
+      "step": 3469
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002063271039558217,
+      "loss": 2.8391,
+      "step": 3470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020627290534984272,
+      "loss": 3.0058,
+      "step": 3471
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020621869819247268,
+      "loss": 2.9365,
+      "step": 3472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020616448249194912,
+      "loss": 2.2203,
+      "step": 3473
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002061102582565107,
+      "loss": 2.5434,
+      "step": 3474
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020605602549439762,
+      "loss": 2.5927,
+      "step": 3475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020600178421385115,
+      "loss": 2.757,
+      "step": 3476
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020594753442311397,
+      "loss": 2.6685,
+      "step": 3477
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020589327613043003,
+      "loss": 2.2512,
+      "step": 3478
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002058390093440446,
+      "loss": 3.02,
+      "step": 3479
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020578473407220418,
+      "loss": 2.9115,
+      "step": 3480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020573045032315664,
+      "loss": 2.7544,
+      "step": 3481
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020567615810515106,
+      "loss": 2.9815,
+      "step": 3482
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020562185742643784,
+      "loss": 3.0222,
+      "step": 3483
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020556754829526873,
+      "loss": 3.1878,
+      "step": 3484
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002055132307198966,
+      "loss": 2.6911,
+      "step": 3485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020545890470857574,
+      "loss": 2.6445,
+      "step": 3486
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002054045702695617,
+      "loss": 3.0946,
+      "step": 3487
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002053502274111113,
+      "loss": 2.8043,
+      "step": 3488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002052958761414826,
+      "loss": 2.6313,
+      "step": 3489
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020524151646893505,
+      "loss": 2.7966,
+      "step": 3490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002051871484017292,
+      "loss": 2.6347,
+      "step": 3491
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020513277194812707,
+      "loss": 2.2523,
+      "step": 3492
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020507838711639176,
+      "loss": 2.6468,
+      "step": 3493
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020502399391478788,
+      "loss": 2.0059,
+      "step": 3494
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020496959235158103,
+      "loss": 2.5379,
+      "step": 3495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002049151824350383,
+      "loss": 2.4734,
+      "step": 3496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000204860764173428,
+      "loss": 2.6399,
+      "step": 3497
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020480633757501965,
+      "loss": 2.7345,
+      "step": 3498
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002047519026480841,
+      "loss": 3.2384,
+      "step": 3499
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020469745940089338,
+      "loss": 2.5079,
+      "step": 3500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020464300784172088,
+      "loss": 2.4053,
+      "step": 3501
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002045885479788412,
+      "loss": 2.7559,
+      "step": 3502
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002045340798205302,
+      "loss": 2.4447,
+      "step": 3503
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020447960337506509,
+      "loss": 2.5493,
+      "step": 3504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002044251186507242,
+      "loss": 2.3906,
+      "step": 3505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020437062565578716,
+      "loss": 2.625,
+      "step": 3506
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020431612439853494,
+      "loss": 3.0947,
+      "step": 3507
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020426161488724972,
+      "loss": 2.4224,
+      "step": 3508
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020420709713021485,
+      "loss": 2.4026,
+      "step": 3509
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020415257113571504,
+      "loss": 2.7141,
+      "step": 3510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020409803691203622,
+      "loss": 2.7057,
+      "step": 3511
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002040434944674656,
+      "loss": 2.2475,
+      "step": 3512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020398894381029153,
+      "loss": 3.3317,
+      "step": 3513
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020393438494880375,
+      "loss": 2.6177,
+      "step": 3514
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020387981789129312,
+      "loss": 2.6974,
+      "step": 3515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020382524264605186,
+      "loss": 2.8524,
+      "step": 3516
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020377065922137335,
+      "loss": 2.2535,
+      "step": 3517
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020371606762555227,
+      "loss": 3.1007,
+      "step": 3518
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020366146786688445,
+      "loss": 2.2164,
+      "step": 3519
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002036068599536671,
+      "loss": 2.9959,
+      "step": 3520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020355224389419853,
+      "loss": 2.9197,
+      "step": 3521
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002034976196967784,
+      "loss": 2.8749,
+      "step": 3522
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020344298736970753,
+      "loss": 2.8591,
+      "step": 3523
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020338834692128804,
+      "loss": 2.6385,
+      "step": 3524
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020333369835982322,
+      "loss": 2.9861,
+      "step": 3525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020327904169361758,
+      "loss": 2.7012,
+      "step": 3526
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000203224376930977,
+      "loss": 2.5133,
+      "step": 3527
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020316970408020845,
+      "loss": 2.3179,
+      "step": 3528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020311502314962016,
+      "loss": 2.8964,
+      "step": 3529
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020306033414752158,
+      "loss": 1.2702,
+      "step": 3530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020300563708222348,
+      "loss": 2.7643,
+      "step": 3531
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020295093196203768,
+      "loss": 2.7764,
+      "step": 3532
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020289621879527743,
+      "loss": 2.5114,
+      "step": 3533
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000202841497590257,
+      "loss": 2.1831,
+      "step": 3534
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020278676835529204,
+      "loss": 3.0565,
+      "step": 3535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020273203109869934,
+      "loss": 2.6864,
+      "step": 3536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020267728582879695,
+      "loss": 2.6027,
+      "step": 3537
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002026225325539041,
+      "loss": 2.7572,
+      "step": 3538
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020256777128234126,
+      "loss": 3.0444,
+      "step": 3539
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002025130020224301,
+      "loss": 2.4864,
+      "step": 3540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020245822478249354,
+      "loss": 3.1203,
+      "step": 3541
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002024034395708557,
+      "loss": 3.0599,
+      "step": 3542
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020234864639584178,
+      "loss": 2.9664,
+      "step": 3543
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020229384526577841,
+      "loss": 2.7785,
+      "step": 3544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020223903618899334,
+      "loss": 2.4181,
+      "step": 3545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002021842191738155,
+      "loss": 2.7367,
+      "step": 3546
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020212939422857497,
+      "loss": 2.3938,
+      "step": 3547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020207456136160317,
+      "loss": 2.0986,
+      "step": 3548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020201972058123265,
+      "loss": 2.6566,
+      "step": 3549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020196487189579713,
+      "loss": 2.1935,
+      "step": 3550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020191001531363163,
+      "loss": 3.0726,
+      "step": 3551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002018551508430723,
+      "loss": 2.8073,
+      "step": 3552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020180027849245647,
+      "loss": 2.2894,
+      "step": 3553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020174539827012272,
+      "loss": 2.3188,
+      "step": 3554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020169051018441083,
+      "loss": 2.8154,
+      "step": 3555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002016356142436617,
+      "loss": 2.7287,
+      "step": 3556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002015807104562175,
+      "loss": 3.1837,
+      "step": 3557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020152579883042158,
+      "loss": 2.6558,
+      "step": 3558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020147087937461843,
+      "loss": 2.8198,
+      "step": 3559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020141595209715384,
+      "loss": 2.2845,
+      "step": 3560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002013610170063746,
+      "loss": 3.0712,
+      "step": 3561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020130607411062888,
+      "loss": 2.4616,
+      "step": 3562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020125112341826595,
+      "loss": 2.4451,
+      "step": 3563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020119616493763625,
+      "loss": 2.8786,
+      "step": 3564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020114119867709147,
+      "loss": 2.394,
+      "step": 3565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002010862246449844,
+      "loss": 2.738,
+      "step": 3566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002010312428496691,
+      "loss": 2.8191,
+      "step": 3567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002009762532995007,
+      "loss": 2.3757,
+      "step": 3568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020092125600283558,
+      "loss": 2.5007,
+      "step": 3569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020086625096803133,
+      "loss": 2.214,
+      "step": 3570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020081123820344662,
+      "loss": 2.5295,
+      "step": 3571
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002007562177174414,
+      "loss": 2.5732,
+      "step": 3572
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002007011895183767,
+      "loss": 1.1175,
+      "step": 3573
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020064615361461478,
+      "loss": 2.6463,
+      "step": 3574
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020059111001451905,
+      "loss": 2.7621,
+      "step": 3575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020053605872645403,
+      "loss": 2.4698,
+      "step": 3576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020048099975878558,
+      "loss": 2.9934,
+      "step": 3577
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020042593311988052,
+      "loss": 2.1244,
+      "step": 3578
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000200370858818107,
+      "loss": 2.7842,
+      "step": 3579
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020031577686183426,
+      "loss": 2.7392,
+      "step": 3580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020026068725943267,
+      "loss": 2.4937,
+      "step": 3581
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020020559001927385,
+      "loss": 2.9048,
+      "step": 3582
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020015048514973052,
+      "loss": 2.7108,
+      "step": 3583
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00020009537265917654,
+      "loss": 2.1783,
+      "step": 3584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000200040252555987,
+      "loss": 3.193,
+      "step": 3585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019998512484853807,
+      "loss": 2.1585,
+      "step": 3586
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019992998954520723,
+      "loss": 3.0789,
+      "step": 3587
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987484665437286,
+      "loss": 1.0348,
+      "step": 3588
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981969618441465,
+      "loss": 2.7002,
+      "step": 3589
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976453814371344,
+      "loss": 2.6501,
+      "step": 3590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970937254065122,
+      "loss": 2.6228,
+      "step": 3591
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019965419938361117,
+      "loss": 2.6785,
+      "step": 3592
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019959901868097743,
+      "loss": 2.5672,
+      "step": 3593
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001995438304411355,
+      "loss": 2.6381,
+      "step": 3594
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019948863467247194,
+      "loss": 2.6528,
+      "step": 3595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019943343138337442,
+      "loss": 2.396,
+      "step": 3596
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019937822058223182,
+      "loss": 2.5191,
+      "step": 3597
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019932300227743408,
+      "loss": 2.8277,
+      "step": 3598
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001992677764773724,
+      "loss": 2.6131,
+      "step": 3599
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019921254319043904,
+      "loss": 3.3718,
+      "step": 3600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019915730242502737,
+      "loss": 2.6287,
+      "step": 3601
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019910205418953195,
+      "loss": 2.8534,
+      "step": 3602
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019904679849234845,
+      "loss": 2.8648,
+      "step": 3603
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019899153534187373,
+      "loss": 2.0757,
+      "step": 3604
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019893626474650568,
+      "loss": 2.5184,
+      "step": 3605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019888098671464344,
+      "loss": 1.4692,
+      "step": 3606
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019882570125468715,
+      "loss": 2.6664,
+      "step": 3607
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019877040837503815,
+      "loss": 2.1026,
+      "step": 3608
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001987151080840989,
+      "loss": 2.67,
+      "step": 3609
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019865980039027309,
+      "loss": 2.7186,
+      "step": 3610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019860448530196532,
+      "loss": 2.5042,
+      "step": 3611
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019854916282758148,
+      "loss": 2.6704,
+      "step": 3612
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001984938329755285,
+      "loss": 2.6392,
+      "step": 3613
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001984384957542145,
+      "loss": 2.1871,
+      "step": 3614
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019838315117204865,
+      "loss": 2.9367,
+      "step": 3615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019832779923744127,
+      "loss": 2.9092,
+      "step": 3616
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019827243995880386,
+      "loss": 2.6316,
+      "step": 3617
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019821707334454888,
+      "loss": 2.8892,
+      "step": 3618
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019816169940309005,
+      "loss": 2.5253,
+      "step": 3619
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019810631814284215,
+      "loss": 2.3843,
+      "step": 3620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001980509295722211,
+      "loss": 3.0017,
+      "step": 3621
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019799553369964388,
+      "loss": 2.3264,
+      "step": 3622
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019794013053352856,
+      "loss": 2.3484,
+      "step": 3623
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001978847200822944,
+      "loss": 2.1227,
+      "step": 3624
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019782930235436171,
+      "loss": 2.772,
+      "step": 3625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000197773877358152,
+      "loss": 2.8942,
+      "step": 3626
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019771844510208772,
+      "loss": 3.1102,
+      "step": 3627
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019766300559459257,
+      "loss": 2.4245,
+      "step": 3628
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001976075588440913,
+      "loss": 2.4299,
+      "step": 3629
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001975521048590097,
+      "loss": 3.2522,
+      "step": 3630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019749664364777472,
+      "loss": 2.2822,
+      "step": 3631
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001974411752188145,
+      "loss": 2.3227,
+      "step": 3632
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019738569958055807,
+      "loss": 2.8353,
+      "step": 3633
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019733021674143574,
+      "loss": 2.4282,
+      "step": 3634
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001972747267098788,
+      "loss": 2.5458,
+      "step": 3635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019721922949431968,
+      "loss": 2.2405,
+      "step": 3636
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019716372510319188,
+      "loss": 2.7017,
+      "step": 3637
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001971082135449301,
+      "loss": 2.7822,
+      "step": 3638
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019705269482796987,
+      "loss": 2.7642,
+      "step": 3639
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001969971689607481,
+      "loss": 2.2852,
+      "step": 3640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019694163595170258,
+      "loss": 2.5272,
+      "step": 3641
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019688609580927234,
+      "loss": 2.4205,
+      "step": 3642
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019683054854189737,
+      "loss": 2.268,
+      "step": 3643
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001967749941580188,
+      "loss": 3.1683,
+      "step": 3644
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019671943266607887,
+      "loss": 2.6919,
+      "step": 3645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019666386407452078,
+      "loss": 2.9089,
+      "step": 3646
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019660828839178894,
+      "loss": 2.0806,
+      "step": 3647
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019655270562632884,
+      "loss": 2.1015,
+      "step": 3648
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019649711578658693,
+      "loss": 3.1434,
+      "step": 3649
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001964415188810108,
+      "loss": 2.7407,
+      "step": 3650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019638591491804913,
+      "loss": 2.6553,
+      "step": 3651
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019633030390615164,
+      "loss": 2.4305,
+      "step": 3652
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019627468585376915,
+      "loss": 2.5672,
+      "step": 3653
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019621906076935355,
+      "loss": 2.5148,
+      "step": 3654
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019616342866135778,
+      "loss": 2.6395,
+      "step": 3655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019610778953823587,
+      "loss": 2.7324,
+      "step": 3656
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019605214340844286,
+      "loss": 2.5742,
+      "step": 3657
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019599649028043494,
+      "loss": 2.6102,
+      "step": 3658
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001959408301626693,
+      "loss": 2.9813,
+      "step": 3659
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019588516306360418,
+      "loss": 2.3207,
+      "step": 3660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019582948899169892,
+      "loss": 2.5048,
+      "step": 3661
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019577380795541394,
+      "loss": 2.8412,
+      "step": 3662
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001957181199632107,
+      "loss": 2.6991,
+      "step": 3663
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019566242502355166,
+      "loss": 1.6828,
+      "step": 3664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019560672314490035,
+      "loss": 2.766,
+      "step": 3665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019555101433572152,
+      "loss": 2.8383,
+      "step": 3666
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001954952986044807,
+      "loss": 2.7207,
+      "step": 3667
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019543957595964472,
+      "loss": 2.5853,
+      "step": 3668
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019538384640968123,
+      "loss": 2.4365,
+      "step": 3669
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019532810996305918,
+      "loss": 3.1532,
+      "step": 3670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019527236662824837,
+      "loss": 2.3417,
+      "step": 3671
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019521661641371972,
+      "loss": 2.4748,
+      "step": 3672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019516085932794517,
+      "loss": 2.355,
+      "step": 3673
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019510509537939776,
+      "loss": 2.709,
+      "step": 3674
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019504932457655157,
+      "loss": 2.6734,
+      "step": 3675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019499354692788164,
+      "loss": 3.0489,
+      "step": 3676
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001949377624418641,
+      "loss": 3.0596,
+      "step": 3677
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001948819711269761,
+      "loss": 2.6125,
+      "step": 3678
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019482617299169594,
+      "loss": 2.1636,
+      "step": 3679
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019477036804450279,
+      "loss": 2.649,
+      "step": 3680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019471455629387695,
+      "loss": 2.5006,
+      "step": 3681
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019465873774829968,
+      "loss": 2.5689,
+      "step": 3682
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019460291241625344,
+      "loss": 2.6296,
+      "step": 3683
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019454708030622146,
+      "loss": 2.1681,
+      "step": 3684
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019449124142668827,
+      "loss": 1.8909,
+      "step": 3685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019443539578613926,
+      "loss": 3.2399,
+      "step": 3686
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019437954339306083,
+      "loss": 2.5029,
+      "step": 3687
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019432368425594058,
+      "loss": 2.6444,
+      "step": 3688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019426781838326697,
+      "loss": 2.9534,
+      "step": 3689
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019421194578352955,
+      "loss": 1.483,
+      "step": 3690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019415606646521884,
+      "loss": 2.6389,
+      "step": 3691
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019410018043682648,
+      "loss": 3.0686,
+      "step": 3692
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019404428770684504,
+      "loss": 2.6026,
+      "step": 3693
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001939883882837681,
+      "loss": 2.456,
+      "step": 3694
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019393248217609035,
+      "loss": 3.0389,
+      "step": 3695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001938765693923075,
+      "loss": 2.8037,
+      "step": 3696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019382064994091604,
+      "loss": 2.6913,
+      "step": 3697
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001937647238304138,
+      "loss": 2.0098,
+      "step": 3698
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019370879106929943,
+      "loss": 2.6738,
+      "step": 3699
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001936528516660726,
+      "loss": 2.694,
+      "step": 3700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019359690562923402,
+      "loss": 1.9736,
+      "step": 3701
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019354095296728548,
+      "loss": 2.1332,
+      "step": 3702
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019348499368872964,
+      "loss": 2.2594,
+      "step": 3703
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019342902780207028,
+      "loss": 2.9803,
+      "step": 3704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001933730553158121,
+      "loss": 2.4253,
+      "step": 3705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001933170762384608,
+      "loss": 2.8917,
+      "step": 3706
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019326109057852316,
+      "loss": 2.4552,
+      "step": 3707
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000193205098344507,
+      "loss": 2.9817,
+      "step": 3708
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019314909954492093,
+      "loss": 2.6164,
+      "step": 3709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001930930941882748,
+      "loss": 2.8946,
+      "step": 3710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019303708228307924,
+      "loss": 2.7348,
+      "step": 3711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019298106383784603,
+      "loss": 3.1581,
+      "step": 3712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001929250388610879,
+      "loss": 2.6489,
+      "step": 3713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019286900736131853,
+      "loss": 2.6765,
+      "step": 3714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019281296934705268,
+      "loss": 2.5386,
+      "step": 3715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000192756924826806,
+      "loss": 2.9336,
+      "step": 3716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019270087380909514,
+      "loss": 3.018,
+      "step": 3717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001926448163024379,
+      "loss": 2.7729,
+      "step": 3718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019258875231535283,
+      "loss": 1.4675,
+      "step": 3719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019253268185635963,
+      "loss": 2.9378,
+      "step": 3720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019247660493397885,
+      "loss": 2.4287,
+      "step": 3721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001924205215567322,
+      "loss": 2.5983,
+      "step": 3722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019236443173314217,
+      "loss": 2.5272,
+      "step": 3723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001923083354717324,
+      "loss": 2.4346,
+      "step": 3724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019225223278102746,
+      "loss": 2.8879,
+      "step": 3725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019219612366955276,
+      "loss": 2.3323,
+      "step": 3726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019214000814583493,
+      "loss": 2.5334,
+      "step": 3727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019208388621840135,
+      "loss": 3.0004,
+      "step": 3728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019202775789578053,
+      "loss": 2.3614,
+      "step": 3729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019197162318650187,
+      "loss": 2.786,
+      "step": 3730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019191548209909572,
+      "loss": 2.7334,
+      "step": 3731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019185933464209353,
+      "loss": 2.34,
+      "step": 3732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019180318082402754,
+      "loss": 2.499,
+      "step": 3733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019174702065343107,
+      "loss": 2.302,
+      "step": 3734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019169085413883838,
+      "loss": 2.7388,
+      "step": 3735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001916346812887847,
+      "loss": 2.5388,
+      "step": 3736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019157850211180626,
+      "loss": 2.5233,
+      "step": 3737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001915223166164401,
+      "loss": 2.5918,
+      "step": 3738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019146612481122445,
+      "loss": 2.2154,
+      "step": 3739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019140992670469825,
+      "loss": 2.3663,
+      "step": 3740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019135372230540167,
+      "loss": 2.7081,
+      "step": 3741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001912975116218756,
+      "loss": 2.6832,
+      "step": 3742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019124129466266198,
+      "loss": 2.6093,
+      "step": 3743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019118507143630367,
+      "loss": 1.6864,
+      "step": 3744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019112884195134457,
+      "loss": 2.731,
+      "step": 3745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019107260621632947,
+      "loss": 2.5556,
+      "step": 3746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019101636423980408,
+      "loss": 2.2453,
+      "step": 3747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019096011603031513,
+      "loss": 2.7884,
+      "step": 3748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001909038615964102,
+      "loss": 2.6124,
+      "step": 3749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019084760094663792,
+      "loss": 2.9284,
+      "step": 3750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019079133408954783,
+      "loss": 3.1409,
+      "step": 3751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019073506103369037,
+      "loss": 2.124,
+      "step": 3752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019067878178761694,
+      "loss": 2.3946,
+      "step": 3753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019062249635987995,
+      "loss": 2.9386,
+      "step": 3754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019056620475903265,
+      "loss": 2.6315,
+      "step": 3755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001905099069936293,
+      "loss": 0.9171,
+      "step": 3756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019045360307222504,
+      "loss": 2.4199,
+      "step": 3757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019039729300337598,
+      "loss": 2.7394,
+      "step": 3758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001903409767956392,
+      "loss": 2.9785,
+      "step": 3759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019028465445757256,
+      "loss": 2.1704,
+      "step": 3760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019022832599773512,
+      "loss": 2.5559,
+      "step": 3761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001901719914246866,
+      "loss": 2.6653,
+      "step": 3762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019011565074698777,
+      "loss": 2.698,
+      "step": 3763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019005930397320036,
+      "loss": 2.3374,
+      "step": 3764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000190002951111887,
+      "loss": 2.5283,
+      "step": 3765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018994659217161117,
+      "loss": 2.6317,
+      "step": 3766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018989022716093737,
+      "loss": 2.6559,
+      "step": 3767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018983385608843103,
+      "loss": 2.6143,
+      "step": 3768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018977747896265838,
+      "loss": 2.6419,
+      "step": 3769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001897210957921867,
+      "loss": 2.7094,
+      "step": 3770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018966470658558416,
+      "loss": 2.7518,
+      "step": 3771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001896083113514197,
+      "loss": 2.3156,
+      "step": 3772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018955191009826345,
+      "loss": 3.1203,
+      "step": 3773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001894955028346862,
+      "loss": 3.12,
+      "step": 3774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018943908956925982,
+      "loss": 2.3961,
+      "step": 3775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000189382670310557,
+      "loss": 2.8088,
+      "step": 3776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018932624506715138,
+      "loss": 2.4436,
+      "step": 3777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001892698138476175,
+      "loss": 2.4256,
+      "step": 3778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018921337666053081,
+      "loss": 2.8392,
+      "step": 3779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001891569335144677,
+      "loss": 2.5897,
+      "step": 3780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018910048441800534,
+      "loss": 2.3877,
+      "step": 3781
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000189044029379722,
+      "loss": 1.9527,
+      "step": 3782
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018898756840819668,
+      "loss": 2.4379,
+      "step": 3783
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001889311015120094,
+      "loss": 2.755,
+      "step": 3784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018887462869974096,
+      "loss": 2.7263,
+      "step": 3785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018881814997997323,
+      "loss": 2.4634,
+      "step": 3786
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001887616653612888,
+      "loss": 2.9771,
+      "step": 3787
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001887051748522712,
+      "loss": 2.8811,
+      "step": 3788
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018864867846150502,
+      "loss": 2.4158,
+      "step": 3789
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018859217619757548,
+      "loss": 1.8096,
+      "step": 3790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001885356680690689,
+      "loss": 2.7761,
+      "step": 3791
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018847915408457238,
+      "loss": 2.764,
+      "step": 3792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018842263425267397,
+      "loss": 2.3984,
+      "step": 3793
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018836610858196257,
+      "loss": 2.9504,
+      "step": 3794
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000188309577081028,
+      "loss": 2.7248,
+      "step": 3795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018825303975846094,
+      "loss": 2.6931,
+      "step": 3796
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018819649662285296,
+      "loss": 2.8649,
+      "step": 3797
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018813994768279648,
+      "loss": 2.5497,
+      "step": 3798
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001880833929468849,
+      "loss": 2.7926,
+      "step": 3799
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018802683242371247,
+      "loss": 2.8989,
+      "step": 3800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018797026612187413,
+      "loss": 2.6233,
+      "step": 3801
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018791369404996602,
+      "loss": 3.0762,
+      "step": 3802
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018785711621658493,
+      "loss": 2.2355,
+      "step": 3803
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018780053263032853,
+      "loss": 2.3623,
+      "step": 3804
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018774394329979554,
+      "loss": 2.265,
+      "step": 3805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018768734823358538,
+      "loss": 2.7746,
+      "step": 3806
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018763074744029837,
+      "loss": 2.5725,
+      "step": 3807
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018757414092853574,
+      "loss": 2.9791,
+      "step": 3808
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018751752870689963,
+      "loss": 2.2628,
+      "step": 3809
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001874609107839929,
+      "loss": 2.8623,
+      "step": 3810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018740428716841945,
+      "loss": 3.3415,
+      "step": 3811
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018734765786878392,
+      "loss": 3.0464,
+      "step": 3812
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018729102289369193,
+      "loss": 2.0297,
+      "step": 3813
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001872343822517498,
+      "loss": 2.1917,
+      "step": 3814
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018717773595156487,
+      "loss": 2.2716,
+      "step": 3815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018712108400174523,
+      "loss": 2.8775,
+      "step": 3816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001870644264108999,
+      "loss": 2.6646,
+      "step": 3817
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018700776318763876,
+      "loss": 2.4605,
+      "step": 3818
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018695109434057244,
+      "loss": 3.0653,
+      "step": 3819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018689441987831254,
+      "loss": 2.9564,
+      "step": 3820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018683773980947144,
+      "loss": 2.6954,
+      "step": 3821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018678105414266245,
+      "loss": 2.674,
+      "step": 3822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018672436288649967,
+      "loss": 1.1239,
+      "step": 3823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018666766604959805,
+      "loss": 2.5094,
+      "step": 3824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018661096364057346,
+      "loss": 2.4587,
+      "step": 3825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018655425566804244,
+      "loss": 2.6682,
+      "step": 3826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018649754214062264,
+      "loss": 2.618,
+      "step": 3827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001864408230669323,
+      "loss": 2.4708,
+      "step": 3828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018638409845559065,
+      "loss": 2.7106,
+      "step": 3829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018632736831521775,
+      "loss": 2.9634,
+      "step": 3830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018627063265443442,
+      "loss": 2.1788,
+      "step": 3831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018621389148186244,
+      "loss": 1.5218,
+      "step": 3832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001861571448061243,
+      "loss": 2.4113,
+      "step": 3833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001861003926358434,
+      "loss": 1.9624,
+      "step": 3834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018604363497964397,
+      "loss": 2.8464,
+      "step": 3835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018598687184615106,
+      "loss": 2.6399,
+      "step": 3836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001859301032439906,
+      "loss": 2.7231,
+      "step": 3837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018587332918178922,
+      "loss": 2.2897,
+      "step": 3838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018581654966817455,
+      "loss": 2.4483,
+      "step": 3839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018575976471177494,
+      "loss": 2.571,
+      "step": 3840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018570297432121956,
+      "loss": 2.4154,
+      "step": 3841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018564617850513846,
+      "loss": 2.5,
+      "step": 3842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018558937727216257,
+      "loss": 1.7209,
+      "step": 3843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018553257063092343,
+      "loss": 2.762,
+      "step": 3844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018547575859005364,
+      "loss": 2.0829,
+      "step": 3845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018541894115818648,
+      "loss": 2.5863,
+      "step": 3846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018536211834395614,
+      "loss": 3.4038,
+      "step": 3847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018530529015599748,
+      "loss": 2.4791,
+      "step": 3848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018524845660294637,
+      "loss": 2.4343,
+      "step": 3849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001851916176934393,
+      "loss": 2.1783,
+      "step": 3850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018513477343611379,
+      "loss": 2.8328,
+      "step": 3851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018507792383960797,
+      "loss": 2.1612,
+      "step": 3852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018502106891256088,
+      "loss": 2.3072,
+      "step": 3853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018496420866361238,
+      "loss": 2.3777,
+      "step": 3854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018490734310140308,
+      "loss": 2.561,
+      "step": 3855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001848504722345745,
+      "loss": 2.6426,
+      "step": 3856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001847935960717688,
+      "loss": 2.6125,
+      "step": 3857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018473671462162914,
+      "loss": 2.0915,
+      "step": 3858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018467982789279933,
+      "loss": 2.4416,
+      "step": 3859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018462293589392408,
+      "loss": 2.3637,
+      "step": 3860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018456603863364882,
+      "loss": 3.5848,
+      "step": 3861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018450913612061984,
+      "loss": 2.2934,
+      "step": 3862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001844522283634842,
+      "loss": 2.9232,
+      "step": 3863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018439531537088976,
+      "loss": 2.8295,
+      "step": 3864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001843383971514852,
+      "loss": 2.2288,
+      "step": 3865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018428147371391996,
+      "loss": 2.2924,
+      "step": 3866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001842245450668443,
+      "loss": 2.4669,
+      "step": 3867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018416761121890923,
+      "loss": 2.4582,
+      "step": 3868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001841106721787666,
+      "loss": 3.0288,
+      "step": 3869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018405372795506905,
+      "loss": 2.9376,
+      "step": 3870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018399677855646996,
+      "loss": 2.1455,
+      "step": 3871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018393982399162354,
+      "loss": 2.5094,
+      "step": 3872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018388286426918474,
+      "loss": 2.3059,
+      "step": 3873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018382589939780933,
+      "loss": 2.543,
+      "step": 3874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018376892938615393,
+      "loss": 2.1838,
+      "step": 3875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018371195424287572,
+      "loss": 2.1978,
+      "step": 3876
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001836549739766329,
+      "loss": 2.0039,
+      "step": 3877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018359798859608436,
+      "loss": 2.6312,
+      "step": 3878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018354099810988976,
+      "loss": 2.3676,
+      "step": 3879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018348400252670945,
+      "loss": 2.6908,
+      "step": 3880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018342700185520474,
+      "loss": 2.9782,
+      "step": 3881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001833699961040376,
+      "loss": 1.8939,
+      "step": 3882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018331298528187072,
+      "loss": 2.4694,
+      "step": 3883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001832559693973677,
+      "loss": 2.5371,
+      "step": 3884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001831989484591928,
+      "loss": 2.4836,
+      "step": 3885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018314192247601108,
+      "loss": 2.6567,
+      "step": 3886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001830848914564884,
+      "loss": 2.9532,
+      "step": 3887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018302785540929135,
+      "loss": 2.6065,
+      "step": 3888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018297081434308728,
+      "loss": 2.7202,
+      "step": 3889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018291376826654428,
+      "loss": 1.969,
+      "step": 3890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018285671718833133,
+      "loss": 2.5092,
+      "step": 3891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018279966111711798,
+      "loss": 2.6657,
+      "step": 3892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018274260006157468,
+      "loss": 2.0784,
+      "step": 3893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018268553403037258,
+      "loss": 3.1496,
+      "step": 3894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018262846303218356,
+      "loss": 2.7081,
+      "step": 3895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018257138707568034,
+      "loss": 2.5166,
+      "step": 3896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018251430616953636,
+      "loss": 2.4818,
+      "step": 3897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018245722032242571,
+      "loss": 2.4502,
+      "step": 3898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018240012954302343,
+      "loss": 3.1151,
+      "step": 3899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001823430338400051,
+      "loss": 2.9581,
+      "step": 3900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018228593322204719,
+      "loss": 3.299,
+      "step": 3901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001822288276978269,
+      "loss": 2.0702,
+      "step": 3902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001821717172760221,
+      "loss": 2.3194,
+      "step": 3903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018211460196531146,
+      "loss": 1.7962,
+      "step": 3904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018205748177437442,
+      "loss": 2.7894,
+      "step": 3905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018200035671189112,
+      "loss": 2.6512,
+      "step": 3906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018194322678654243,
+      "loss": 2.186,
+      "step": 3907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018188609200700995,
+      "loss": 2.6549,
+      "step": 3908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018182895238197608,
+      "loss": 2.9289,
+      "step": 3909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018177180792012393,
+      "loss": 2.8359,
+      "step": 3910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018171465863013727,
+      "loss": 2.3404,
+      "step": 3911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018165750452070077,
+      "loss": 2.455,
+      "step": 3912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018160034560049966,
+      "loss": 2.3214,
+      "step": 3913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018154318187821994,
+      "loss": 2.6972,
+      "step": 3914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001814860133625485,
+      "loss": 2.6759,
+      "step": 3915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001814288400621727,
+      "loss": 3.2886,
+      "step": 3916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018137166198578083,
+      "loss": 2.5145,
+      "step": 3917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001813144791420618,
+      "loss": 2.6537,
+      "step": 3918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001812572915397053,
+      "loss": 2.4572,
+      "step": 3919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018120009918740171,
+      "loss": 2.5108,
+      "step": 3920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001811429020938421,
+      "loss": 2.9712,
+      "step": 3921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018108570026771848,
+      "loss": 2.4761,
+      "step": 3922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018102849371772316,
+      "loss": 2.7864,
+      "step": 3923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018097128245254956,
+      "loss": 2.6018,
+      "step": 3924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001809140664808916,
+      "loss": 3.0235,
+      "step": 3925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018085684581144404,
+      "loss": 2.5499,
+      "step": 3926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018079962045290223,
+      "loss": 2.6033,
+      "step": 3927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018074239041396237,
+      "loss": 2.4337,
+      "step": 3928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018068515570332123,
+      "loss": 2.6073,
+      "step": 3929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018062791632967643,
+      "loss": 2.2575,
+      "step": 3930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018057067230172616,
+      "loss": 1.8859,
+      "step": 3931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018051342362816945,
+      "loss": 2.7484,
+      "step": 3932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001804561703177059,
+      "loss": 2.8106,
+      "step": 3933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018039891237903595,
+      "loss": 3.0972,
+      "step": 3934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018034164982086065,
+      "loss": 2.8794,
+      "step": 3935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018028438265188182,
+      "loss": 2.5309,
+      "step": 3936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018022711088080192,
+      "loss": 2.6998,
+      "step": 3937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018016983451632406,
+      "loss": 2.4396,
+      "step": 3938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001801125535671522,
+      "loss": 2.2628,
+      "step": 3939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00018005526804199088,
+      "loss": 2.9063,
+      "step": 3940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001799979779495454,
+      "loss": 2.3355,
+      "step": 3941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001799406832985217,
+      "loss": 2.7334,
+      "step": 3942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001798833840976265,
+      "loss": 2.6919,
+      "step": 3943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017982608035556707,
+      "loss": 2.2527,
+      "step": 3944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017976877208105148,
+      "loss": 2.2462,
+      "step": 3945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017971145928278844,
+      "loss": 2.9512,
+      "step": 3946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017965414196948745,
+      "loss": 1.3727,
+      "step": 3947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001795968201498585,
+      "loss": 2.5154,
+      "step": 3948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017953949383261243,
+      "loss": 2.5579,
+      "step": 3949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017948216302646075,
+      "loss": 2.8912,
+      "step": 3950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001794248277401155,
+      "loss": 2.7471,
+      "step": 3951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017936748798228963,
+      "loss": 2.0537,
+      "step": 3952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001793101437616966,
+      "loss": 2.7293,
+      "step": 3953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017925279508705064,
+      "loss": 3.0318,
+      "step": 3954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001791954419670665,
+      "loss": 2.2769,
+      "step": 3955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017913808441045987,
+      "loss": 2.5141,
+      "step": 3956
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017908072242594686,
+      "loss": 2.4476,
+      "step": 3957
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017902335602224445,
+      "loss": 2.5809,
+      "step": 3958
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017896598520807016,
+      "loss": 2.4868,
+      "step": 3959
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001789086099921422,
+      "loss": 3.1747,
+      "step": 3960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017885123038317947,
+      "loss": 2.8403,
+      "step": 3961
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001787938463899016,
+      "loss": 2.0555,
+      "step": 3962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017873645802102877,
+      "loss": 2.4476,
+      "step": 3963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001786790652852819,
+      "loss": 2.5846,
+      "step": 3964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017862166819138252,
+      "loss": 2.5182,
+      "step": 3965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017856426674805293,
+      "loss": 2.7031,
+      "step": 3966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017850686096401597,
+      "loss": 2.2778,
+      "step": 3967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001784494508479952,
+      "loss": 3.0607,
+      "step": 3968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001783920364087148,
+      "loss": 2.7255,
+      "step": 3969
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017833461765489966,
+      "loss": 2.2478,
+      "step": 3970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017827719459527524,
+      "loss": 2.8297,
+      "step": 3971
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017821976723856779,
+      "loss": 2.5737,
+      "step": 3972
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017816233559350407,
+      "loss": 3.1817,
+      "step": 3973
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001781048996688116,
+      "loss": 2.8663,
+      "step": 3974
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001780474594732185,
+      "loss": 2.3896,
+      "step": 3975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017799001501545356,
+      "loss": 2.3537,
+      "step": 3976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017793256630424615,
+      "loss": 2.2162,
+      "step": 3977
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017787511334832636,
+      "loss": 2.67,
+      "step": 3978
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017781765615642497,
+      "loss": 2.6324,
+      "step": 3979
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017776019473727324,
+      "loss": 2.7922,
+      "step": 3980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017770272909960324,
+      "loss": 3.0436,
+      "step": 3981
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017764525925214757,
+      "loss": 2.5621,
+      "step": 3982
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017758778520363957,
+      "loss": 2.3298,
+      "step": 3983
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001775303069628131,
+      "loss": 3.029,
+      "step": 3984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017747282453840272,
+      "loss": 3.0647,
+      "step": 3985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017741533793914368,
+      "loss": 2.888,
+      "step": 3986
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017735784717377173,
+      "loss": 2.3768,
+      "step": 3987
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017730035225102335,
+      "loss": 2.769,
+      "step": 3988
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001772428531796357,
+      "loss": 2.4997,
+      "step": 3989
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017718534996834644,
+      "loss": 2.1335,
+      "step": 3990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017712784262589393,
+      "loss": 3.1719,
+      "step": 3991
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017707033116101713,
+      "loss": 2.8421,
+      "step": 3992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017701281558245573,
+      "loss": 2.3846,
+      "step": 3993
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017695529589894986,
+      "loss": 2.1132,
+      "step": 3994
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017689777211924045,
+      "loss": 2.0517,
+      "step": 3995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001768402442520689,
+      "loss": 2.8029,
+      "step": 3996
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017678271230617742,
+      "loss": 1.8802,
+      "step": 3997
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017672517629030863,
+      "loss": 2.5615,
+      "step": 3998
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017666763621320588,
+      "loss": 2.6965,
+      "step": 3999
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017661009208361315,
+      "loss": 2.445,
+      "step": 4000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017655254391027501,
+      "loss": 2.3353,
+      "step": 4001
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017649499170193663,
+      "loss": 2.7424,
+      "step": 4002
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017643743546734384,
+      "loss": 3.025,
+      "step": 4003
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017637987521524297,
+      "loss": 1.2813,
+      "step": 4004
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017632231095438114,
+      "loss": 2.5381,
+      "step": 4005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017626474269350596,
+      "loss": 2.3767,
+      "step": 4006
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017620717044136559,
+      "loss": 2.3739,
+      "step": 4007
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017614959420670894,
+      "loss": 2.6419,
+      "step": 4008
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017609201399828545,
+      "loss": 2.6584,
+      "step": 4009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017603442982484517,
+      "loss": 2.6039,
+      "step": 4010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017597684169513871,
+      "loss": 2.5726,
+      "step": 4011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017591924961791742,
+      "loss": 2.4193,
+      "step": 4012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017586165360193308,
+      "loss": 2.4863,
+      "step": 4013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001758040536559382,
+      "loss": 2.5623,
+      "step": 4014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017574644978868574,
+      "loss": 2.4198,
+      "step": 4015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017568884200892942,
+      "loss": 2.8965,
+      "step": 4016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001756312303254235,
+      "loss": 2.986,
+      "step": 4017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017557361474692274,
+      "loss": 2.3756,
+      "step": 4018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017551599528218268,
+      "loss": 2.5154,
+      "step": 4019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001754583719399592,
+      "loss": 2.9784,
+      "step": 4020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017540074472900902,
+      "loss": 2.5618,
+      "step": 4021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017534311365808932,
+      "loss": 2.7082,
+      "step": 4022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017528547873595782,
+      "loss": 2.1918,
+      "step": 4023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017522783997137291,
+      "loss": 3.0298,
+      "step": 4024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017517019737309358,
+      "loss": 2.7193,
+      "step": 4025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001751125509498793,
+      "loss": 3.082,
+      "step": 4026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017505490071049025,
+      "loss": 2.6354,
+      "step": 4027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001749972466636871,
+      "loss": 2.4823,
+      "step": 4028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017493958881823114,
+      "loss": 3.1041,
+      "step": 4029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017488192718288415,
+      "loss": 2.6528,
+      "step": 4030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017482426176640862,
+      "loss": 2.1295,
+      "step": 4031
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017476659257756755,
+      "loss": 2.712,
+      "step": 4032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017470891962512446,
+      "loss": 2.534,
+      "step": 4033
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001746512429178436,
+      "loss": 2.572,
+      "step": 4034
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017459356246448958,
+      "loss": 3.057,
+      "step": 4035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017453587827382774,
+      "loss": 2.8487,
+      "step": 4036
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001744781903546239,
+      "loss": 2.7158,
+      "step": 4037
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017442049871564455,
+      "loss": 2.591,
+      "step": 4038
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017436280336565661,
+      "loss": 0.2753,
+      "step": 4039
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017430510431342764,
+      "loss": 2.8875,
+      "step": 4040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017424740156772575,
+      "loss": 2.4436,
+      "step": 4041
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001741896951373197,
+      "loss": 2.8039,
+      "step": 4042
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017413198503097858,
+      "loss": 2.4736,
+      "step": 4043
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001740742712574723,
+      "loss": 2.6076,
+      "step": 4044
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017401655382557116,
+      "loss": 1.6081,
+      "step": 4045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017395883274404602,
+      "loss": 1.8932,
+      "step": 4046
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001739011080216684,
+      "loss": 2.9696,
+      "step": 4047
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017384337966721032,
+      "loss": 2.6053,
+      "step": 4048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017378564768944432,
+      "loss": 2.6792,
+      "step": 4049
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017372791209714348,
+      "loss": 3.1688,
+      "step": 4050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017367017289908156,
+      "loss": 2.5404,
+      "step": 4051
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017361243010403273,
+      "loss": 2.5305,
+      "step": 4052
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001735546837207717,
+      "loss": 2.4084,
+      "step": 4053
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017349693375807383,
+      "loss": 2.6519,
+      "step": 4054
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017343918022471496,
+      "loss": 2.7841,
+      "step": 4055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017338142312947144,
+      "loss": 2.1333,
+      "step": 4056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017332366248112028,
+      "loss": 2.6889,
+      "step": 4057
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017326589828843892,
+      "loss": 2.3181,
+      "step": 4058
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017320813056020533,
+      "loss": 3.3105,
+      "step": 4059
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001731503593051981,
+      "loss": 2.6212,
+      "step": 4060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001730925845321963,
+      "loss": 2.6534,
+      "step": 4061
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017303480624997956,
+      "loss": 2.6212,
+      "step": 4062
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000172977024467328,
+      "loss": 2.4464,
+      "step": 4063
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017291923919302235,
+      "loss": 2.6931,
+      "step": 4064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001728614504358438,
+      "loss": 2.4946,
+      "step": 4065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017280365820457412,
+      "loss": 2.9227,
+      "step": 4066
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017274586250799553,
+      "loss": 2.2771,
+      "step": 4067
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017268806335489088,
+      "loss": 2.2447,
+      "step": 4068
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017263026075404347,
+      "loss": 2.5157,
+      "step": 4069
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017257245471423717,
+      "loss": 2.6292,
+      "step": 4070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001725146452442563,
+      "loss": 2.4369,
+      "step": 4071
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001724568323528858,
+      "loss": 2.5443,
+      "step": 4072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017239901604891112,
+      "loss": 2.6058,
+      "step": 4073
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017234119634111807,
+      "loss": 2.5437,
+      "step": 4074
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017228337323829318,
+      "loss": 1.4445,
+      "step": 4075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001722255467492234,
+      "loss": 2.5418,
+      "step": 4076
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001721677168826962,
+      "loss": 2.7096,
+      "step": 4077
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001721098836474996,
+      "loss": 2.2895,
+      "step": 4078
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017205204705242203,
+      "loss": 2.5216,
+      "step": 4079
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001719942071062526,
+      "loss": 2.4264,
+      "step": 4080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017193636381778078,
+      "loss": 2.7769,
+      "step": 4081
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001718785171957966,
+      "loss": 2.4221,
+      "step": 4082
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017182066724909058,
+      "loss": 2.8084,
+      "step": 4083
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017176281398645384,
+      "loss": 2.4144,
+      "step": 4084
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017170495741667782,
+      "loss": 2.9001,
+      "step": 4085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001716470975485547,
+      "loss": 3.0581,
+      "step": 4086
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017158923439087692,
+      "loss": 2.5666,
+      "step": 4087
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017153136795243753,
+      "loss": 2.0384,
+      "step": 4088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017147349824203013,
+      "loss": 2.7756,
+      "step": 4089
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017141562526844876,
+      "loss": 2.5447,
+      "step": 4090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017135774904048794,
+      "loss": 1.945,
+      "step": 4091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017129986956694276,
+      "loss": 2.498,
+      "step": 4092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017124198685660865,
+      "loss": 2.464,
+      "step": 4093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001711841009182817,
+      "loss": 2.1278,
+      "step": 4094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001711262117607585,
+      "loss": 2.2121,
+      "step": 4095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017106831939283585,
+      "loss": 2.3721,
+      "step": 4096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001710104238233114,
+      "loss": 2.7008,
+      "step": 4097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001709525250609831,
+      "loss": 3.022,
+      "step": 4098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001708946231146493,
+      "loss": 2.7011,
+      "step": 4099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017083671799310912,
+      "loss": 2.6825,
+      "step": 4100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017077880970516187,
+      "loss": 2.6259,
+      "step": 4101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001707208982596075,
+      "loss": 2.1129,
+      "step": 4102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017066298366524637,
+      "loss": 2.5379,
+      "step": 4103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017060506593087938,
+      "loss": 2.6292,
+      "step": 4104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001705471450653079,
+      "loss": 2.2896,
+      "step": 4105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017048922107733367,
+      "loss": 2.9129,
+      "step": 4106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017043129397575901,
+      "loss": 2.9478,
+      "step": 4107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017037336376938666,
+      "loss": 2.4953,
+      "step": 4108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017031543046701998,
+      "loss": 2.5345,
+      "step": 4109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017025749407746253,
+      "loss": 2.3183,
+      "step": 4110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001701995546095186,
+      "loss": 2.7136,
+      "step": 4111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001701416120719928,
+      "loss": 3.1179,
+      "step": 4112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017008366647369022,
+      "loss": 2.1175,
+      "step": 4113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00017002571782341646,
+      "loss": 3.3427,
+      "step": 4114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016996776612997755,
+      "loss": 2.7142,
+      "step": 4115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016990981140218,
+      "loss": 2.7128,
+      "step": 4116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001698518536488308,
+      "loss": 2.6345,
+      "step": 4117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016979389287873731,
+      "loss": 3.1871,
+      "step": 4118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016973592910070752,
+      "loss": 2.0743,
+      "step": 4119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016967796232354967,
+      "loss": 2.4546,
+      "step": 4120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016961999255607263,
+      "loss": 2.84,
+      "step": 4121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016956201980708555,
+      "loss": 2.7491,
+      "step": 4122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016950404408539824,
+      "loss": 2.6326,
+      "step": 4123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001694460653998208,
+      "loss": 2.1937,
+      "step": 4124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016938808375916383,
+      "loss": 2.645,
+      "step": 4125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001693300991722385,
+      "loss": 2.3415,
+      "step": 4126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001692721116478561,
+      "loss": 2.4925,
+      "step": 4127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016921412119482877,
+      "loss": 2.571,
+      "step": 4128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016915612782196883,
+      "loss": 1.9565,
+      "step": 4129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016909813153808908,
+      "loss": 2.5542,
+      "step": 4130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001690401323520029,
+      "loss": 2.5379,
+      "step": 4131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016898213027252392,
+      "loss": 2.1869,
+      "step": 4132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016892412530846636,
+      "loss": 2.5097,
+      "step": 4133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001688661174686448,
+      "loss": 2.734,
+      "step": 4134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016880810676187424,
+      "loss": 2.7663,
+      "step": 4135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001687500931969702,
+      "loss": 2.4967,
+      "step": 4136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016869207678274855,
+      "loss": 2.4204,
+      "step": 4137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016863405752802566,
+      "loss": 2.9563,
+      "step": 4138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016857603544161832,
+      "loss": 2.7124,
+      "step": 4139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016851801053234365,
+      "loss": 2.4768,
+      "step": 4140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016845998280901937,
+      "loss": 2.8712,
+      "step": 4141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001684019522804635,
+      "loss": 2.2809,
+      "step": 4142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016834391895549452,
+      "loss": 2.6113,
+      "step": 4143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016828588284293136,
+      "loss": 2.677,
+      "step": 4144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001682278439515933,
+      "loss": 1.4822,
+      "step": 4145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016816980229030016,
+      "loss": 2.5986,
+      "step": 4146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016811175786787211,
+      "loss": 2.6019,
+      "step": 4147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016805371069312974,
+      "loss": 2.409,
+      "step": 4148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000167995660774894,
+      "loss": 3.0275,
+      "step": 4149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001679376081219864,
+      "loss": 2.4492,
+      "step": 4150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016787955274322877,
+      "loss": 2.773,
+      "step": 4151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001678214946474433,
+      "loss": 3.0592,
+      "step": 4152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016776343384345278,
+      "loss": 2.4871,
+      "step": 4153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001677053703400802,
+      "loss": 2.57,
+      "step": 4154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016764730414614912,
+      "loss": 2.7002,
+      "step": 4155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001675892352704834,
+      "loss": 2.9104,
+      "step": 4156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016753116372190735,
+      "loss": 3.015,
+      "step": 4157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016747308950924576,
+      "loss": 2.1846,
+      "step": 4158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016741501264132367,
+      "loss": 2.4855,
+      "step": 4159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001673569331269666,
+      "loss": 2.4516,
+      "step": 4160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016729885097500056,
+      "loss": 2.5185,
+      "step": 4161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016724076619425188,
+      "loss": 1.8479,
+      "step": 4162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016718267879354718,
+      "loss": 2.8781,
+      "step": 4163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016712458878171365,
+      "loss": 2.7202,
+      "step": 4164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016706649616757884,
+      "loss": 2.5818,
+      "step": 4165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001670084009599706,
+      "loss": 1.9566,
+      "step": 4166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016695030316771732,
+      "loss": 2.9057,
+      "step": 4167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016689220279964768,
+      "loss": 2.7607,
+      "step": 4168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016683409986459075,
+      "loss": 2.6751,
+      "step": 4169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016677599437137604,
+      "loss": 2.6272,
+      "step": 4170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016671788632883342,
+      "loss": 2.0443,
+      "step": 4171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016665977574579317,
+      "loss": 2.5655,
+      "step": 4172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001666016626310859,
+      "loss": 2.6101,
+      "step": 4173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001665435469935427,
+      "loss": 2.4335,
+      "step": 4174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016648542884199494,
+      "loss": 3.0206,
+      "step": 4175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016642730818527444,
+      "loss": 2.592,
+      "step": 4176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016636918503221338,
+      "loss": 2.5285,
+      "step": 4177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016631105939164434,
+      "loss": 2.0114,
+      "step": 4178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001662529312724002,
+      "loss": 2.4866,
+      "step": 4179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016619480068331437,
+      "loss": 2.8719,
+      "step": 4180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016613666763322046,
+      "loss": 2.319,
+      "step": 4181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016607853213095253,
+      "loss": 2.921,
+      "step": 4182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016602039418534505,
+      "loss": 2.3539,
+      "step": 4183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016596225380523285,
+      "loss": 2.42,
+      "step": 4184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016590411099945106,
+      "loss": 2.076,
+      "step": 4185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001658459657768352,
+      "loss": 2.2749,
+      "step": 4186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001657878181462213,
+      "loss": 2.6233,
+      "step": 4187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001657296681164455,
+      "loss": 2.5892,
+      "step": 4188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016567151569634456,
+      "loss": 2.7781,
+      "step": 4189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016561336089475543,
+      "loss": 2.76,
+      "step": 4190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016555520372051548,
+      "loss": 2.6199,
+      "step": 4191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016549704418246248,
+      "loss": 2.4749,
+      "step": 4192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016543888228943445,
+      "loss": 2.1004,
+      "step": 4193
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016538071805026992,
+      "loss": 2.2673,
+      "step": 4194
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016532255147380762,
+      "loss": 2.7483,
+      "step": 4195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016526438256888673,
+      "loss": 2.9628,
+      "step": 4196
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016520621134434677,
+      "loss": 2.9467,
+      "step": 4197
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001651480378090276,
+      "loss": 2.631,
+      "step": 4198
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016508986197176948,
+      "loss": 2.5438,
+      "step": 4199
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001650316838414129,
+      "loss": 2.9497,
+      "step": 4200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016497350342679885,
+      "loss": 2.0477,
+      "step": 4201
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016491532073676849,
+      "loss": 2.9313,
+      "step": 4202
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016485713578016353,
+      "loss": 2.8956,
+      "step": 4203
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001647989485658259,
+      "loss": 2.3984,
+      "step": 4204
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016474075910259784,
+      "loss": 2.7872,
+      "step": 4205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016468256739932203,
+      "loss": 2.2687,
+      "step": 4206
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016462437346484145,
+      "loss": 1.6177,
+      "step": 4207
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016456617730799941,
+      "loss": 2.265,
+      "step": 4208
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016450797893763957,
+      "loss": 2.6945,
+      "step": 4209
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016444977836260592,
+      "loss": 2.8879,
+      "step": 4210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016439157559174273,
+      "loss": 3.0133,
+      "step": 4211
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016433337063389476,
+      "loss": 2.2128,
+      "step": 4212
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001642751634979069,
+      "loss": 2.3553,
+      "step": 4213
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016421695419262453,
+      "loss": 2.5716,
+      "step": 4214
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016415874272689333,
+      "loss": 2.9582,
+      "step": 4215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001641005291095592,
+      "loss": 2.9363,
+      "step": 4216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016404231334946848,
+      "loss": 2.5589,
+      "step": 4217
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016398409545546782,
+      "loss": 2.7119,
+      "step": 4218
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016392587543640418,
+      "loss": 2.2418,
+      "step": 4219
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016386765330112484,
+      "loss": 2.5333,
+      "step": 4220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016380942905847736,
+      "loss": 2.4677,
+      "step": 4221
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016375120271730972,
+      "loss": 2.345,
+      "step": 4222
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016369297428647014,
+      "loss": 2.969,
+      "step": 4223
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016363474377480716,
+      "loss": 2.4432,
+      "step": 4224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016357651119116963,
+      "loss": 2.8067,
+      "step": 4225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001635182765444068,
+      "loss": 3.093,
+      "step": 4226
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016346003984336814,
+      "loss": 3.2488,
+      "step": 4227
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016340180109690347,
+      "loss": 2.6713,
+      "step": 4228
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016334356031386292,
+      "loss": 2.5131,
+      "step": 4229
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001632853175030969,
+      "loss": 2.252,
+      "step": 4230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016322707267345623,
+      "loss": 2.5387,
+      "step": 4231
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016316882583379186,
+      "loss": 2.5588,
+      "step": 4232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016311057699295523,
+      "loss": 2.5315,
+      "step": 4233
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016305232615979794,
+      "loss": 2.8513,
+      "step": 4234
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016299407334317194,
+      "loss": 2.3791,
+      "step": 4235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001629358185519296,
+      "loss": 2.8267,
+      "step": 4236
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016287756179492337,
+      "loss": 2.6753,
+      "step": 4237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001628193030810062,
+      "loss": 2.8581,
+      "step": 4238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016276104241903116,
+      "loss": 2.7621,
+      "step": 4239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016270277981785176,
+      "loss": 2.3722,
+      "step": 4240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001626445152863218,
+      "loss": 2.1522,
+      "step": 4241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016258624883329521,
+      "loss": 2.2682,
+      "step": 4242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016252798046762637,
+      "loss": 1.9306,
+      "step": 4243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016246971019816993,
+      "loss": 2.2287,
+      "step": 4244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016241143803378085,
+      "loss": 3.0713,
+      "step": 4245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016235316398331421,
+      "loss": 2.1167,
+      "step": 4246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001622948880556256,
+      "loss": 2.9186,
+      "step": 4247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016223661025957075,
+      "loss": 2.6083,
+      "step": 4248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016217833060400578,
+      "loss": 3.0776,
+      "step": 4249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016212004909778695,
+      "loss": 3.1645,
+      "step": 4250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016206176574977096,
+      "loss": 2.6865,
+      "step": 4251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016200348056881462,
+      "loss": 1.778,
+      "step": 4252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001619451935637752,
+      "loss": 2.5233,
+      "step": 4253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016188690474351015,
+      "loss": 2.4019,
+      "step": 4254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001618286141168772,
+      "loss": 2.4313,
+      "step": 4255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016177032169273432,
+      "loss": 2.0314,
+      "step": 4256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001617120274799398,
+      "loss": 3.0531,
+      "step": 4257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001616537314873522,
+      "loss": 2.7777,
+      "step": 4258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001615954337238304,
+      "loss": 3.1249,
+      "step": 4259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016153713419823343,
+      "loss": 2.4248,
+      "step": 4260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016147883291942066,
+      "loss": 2.5383,
+      "step": 4261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016142052989625172,
+      "loss": 2.679,
+      "step": 4262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016136222513758654,
+      "loss": 2.1315,
+      "step": 4263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016130391865228527,
+      "loss": 2.462,
+      "step": 4264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001612456104492083,
+      "loss": 2.2826,
+      "step": 4265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016118730053721632,
+      "loss": 2.5527,
+      "step": 4266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016112898892517026,
+      "loss": 2.3363,
+      "step": 4267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016107067562193134,
+      "loss": 2.7879,
+      "step": 4268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016101236063636098,
+      "loss": 2.1262,
+      "step": 4269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000160954043977321,
+      "loss": 3.0928,
+      "step": 4270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016089572565367323,
+      "loss": 2.2143,
+      "step": 4271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016083740567427994,
+      "loss": 2.6062,
+      "step": 4272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001607790840480036,
+      "loss": 3.1135,
+      "step": 4273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016072076078370694,
+      "loss": 2.9586,
+      "step": 4274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016066243589025293,
+      "loss": 2.4429,
+      "step": 4275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016060410937650477,
+      "loss": 2.3577,
+      "step": 4276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016054578125132592,
+      "loss": 2.0418,
+      "step": 4277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016048745152358015,
+      "loss": 2.2788,
+      "step": 4278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001604291202021313,
+      "loss": 2.4546,
+      "step": 4279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016037078729584366,
+      "loss": 2.9924,
+      "step": 4280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016031245281358163,
+      "loss": 2.2778,
+      "step": 4281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016025411676420987,
+      "loss": 2.7018,
+      "step": 4282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016019577915659335,
+      "loss": 2.3183,
+      "step": 4283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00016013743999959717,
+      "loss": 2.6836,
+      "step": 4284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001600790993020867,
+      "loss": 2.3079,
+      "step": 4285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001600207570729276,
+      "loss": 1.8835,
+      "step": 4286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015996241332098568,
+      "loss": 2.2437,
+      "step": 4287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015990406805512706,
+      "loss": 3.1732,
+      "step": 4288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015984572128421802,
+      "loss": 2.8089,
+      "step": 4289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001597873730171251,
+      "loss": 2.4715,
+      "step": 4290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015972902326271513,
+      "loss": 2.2522,
+      "step": 4291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015967067202985504,
+      "loss": 2.5301,
+      "step": 4292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001596123193274121,
+      "loss": 2.8921,
+      "step": 4293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015955396516425366,
+      "loss": 2.6579,
+      "step": 4294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001594956095492475,
+      "loss": 2.4667,
+      "step": 4295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015943725249126144,
+      "loss": 3.0057,
+      "step": 4296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015937889399916363,
+      "loss": 2.8888,
+      "step": 4297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015932053408182237,
+      "loss": 2.1406,
+      "step": 4298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015926217274810618,
+      "loss": 2.757,
+      "step": 4299
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015920381000688382,
+      "loss": 3.0245,
+      "step": 4300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015914544586702426,
+      "loss": 2.7684,
+      "step": 4301
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015908708033739668,
+      "loss": 3.3143,
+      "step": 4302
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001590287134268705,
+      "loss": 2.9914,
+      "step": 4303
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015897034514431526,
+      "loss": 2.5502,
+      "step": 4304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015891197549860084,
+      "loss": 2.7651,
+      "step": 4305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015885360449859723,
+      "loss": 1.2111,
+      "step": 4306
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015879523215317465,
+      "loss": 2.1184,
+      "step": 4307
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015873685847120353,
+      "loss": 2.8045,
+      "step": 4308
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001586784834615545,
+      "loss": 2.5842,
+      "step": 4309
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001586201071330984,
+      "loss": 3.0804,
+      "step": 4310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015856172949470625,
+      "loss": 2.3571,
+      "step": 4311
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001585033505552493,
+      "loss": 2.6121,
+      "step": 4312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000158444970323599,
+      "loss": 2.8622,
+      "step": 4313
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015838658880862688,
+      "loss": 3.1682,
+      "step": 4314
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015832820601920488,
+      "loss": 2.7323,
+      "step": 4315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015826982196420498,
+      "loss": 2.4508,
+      "step": 4316
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015821143665249932,
+      "loss": 2.8791,
+      "step": 4317
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015815305009296037,
+      "loss": 1.5323,
+      "step": 4318
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015809466229446066,
+      "loss": 2.2645,
+      "step": 4319
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015803627326587303,
+      "loss": 2.6961,
+      "step": 4320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015797788301607043,
+      "loss": 3.3116,
+      "step": 4321
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015791949155392597,
+      "loss": 1.3015,
+      "step": 4322
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015786109888831304,
+      "loss": 3.2347,
+      "step": 4323
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015780270502810506,
+      "loss": 2.688,
+      "step": 4324
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015774430998217586,
+      "loss": 2.7729,
+      "step": 4325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001576859137593992,
+      "loss": 2.0798,
+      "step": 4326
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001576275163686492,
+      "loss": 2.6171,
+      "step": 4327
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001575691178188001,
+      "loss": 2.2236,
+      "step": 4328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015751071811872624,
+      "loss": 1.8795,
+      "step": 4329
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015745231727730231,
+      "loss": 3.1621,
+      "step": 4330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000157393915303403,
+      "loss": 2.7207,
+      "step": 4331
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015733551220590324,
+      "loss": 2.7669,
+      "step": 4332
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015727710799367812,
+      "loss": 2.243,
+      "step": 4333
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000157218702675603,
+      "loss": 2.7186,
+      "step": 4334
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001571602962605532,
+      "loss": 2.2808,
+      "step": 4335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015710188875740435,
+      "loss": 2.7337,
+      "step": 4336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001570434801750323,
+      "loss": 1.976,
+      "step": 4337
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001569850705223129,
+      "loss": 2.2769,
+      "step": 4338
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001569266598081223,
+      "loss": 2.2761,
+      "step": 4339
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015686824804133671,
+      "loss": 2.9546,
+      "step": 4340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015680983523083263,
+      "loss": 2.5934,
+      "step": 4341
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015675142138548654,
+      "loss": 2.3914,
+      "step": 4342
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015669300651417523,
+      "loss": 2.7569,
+      "step": 4343
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015663459062577556,
+      "loss": 2.2392,
+      "step": 4344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015657617372916467,
+      "loss": 2.848,
+      "step": 4345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015651775583321964,
+      "loss": 2.3948,
+      "step": 4346
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015645933694681782,
+      "loss": 2.5431,
+      "step": 4347
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001564009170788368,
+      "loss": 2.2826,
+      "step": 4348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015634249623815413,
+      "loss": 2.246,
+      "step": 4349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001562840744336477,
+      "loss": 2.5951,
+      "step": 4350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015622565167419542,
+      "loss": 3.1058,
+      "step": 4351
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015616722796867535,
+      "loss": 2.4141,
+      "step": 4352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015610880332596574,
+      "loss": 3.1325,
+      "step": 4353
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015605037775494497,
+      "loss": 2.4983,
+      "step": 4354
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015599195126449157,
+      "loss": 2.8944,
+      "step": 4355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015593352386348422,
+      "loss": 2.6405,
+      "step": 4356
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001558750955608016,
+      "loss": 2.1492,
+      "step": 4357
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015581666636532278,
+      "loss": 3.0147,
+      "step": 4358
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015575823628592676,
+      "loss": 2.302,
+      "step": 4359
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001556998053314927,
+      "loss": 2.6269,
+      "step": 4360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001556413735109,
+      "loss": 1.8665,
+      "step": 4361
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015558294083302813,
+      "loss": 1.8936,
+      "step": 4362
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015552450730675665,
+      "loss": 2.1188,
+      "step": 4363
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001554660729409653,
+      "loss": 2.3379,
+      "step": 4364
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015540763774453392,
+      "loss": 1.9277,
+      "step": 4365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001553492017263425,
+      "loss": 2.3124,
+      "step": 4366
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015529076489527116,
+      "loss": 2.8133,
+      "step": 4367
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001552323272602001,
+      "loss": 2.9893,
+      "step": 4368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015517388883000966,
+      "loss": 2.6672,
+      "step": 4369
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015511544961358036,
+      "loss": 2.6366,
+      "step": 4370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015505700961979273,
+      "loss": 2.6494,
+      "step": 4371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015499856885752754,
+      "loss": 2.3208,
+      "step": 4372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015494012733566556,
+      "loss": 2.7498,
+      "step": 4373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015488168506308782,
+      "loss": 3.1347,
+      "step": 4374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015482324204867524,
+      "loss": 2.7026,
+      "step": 4375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015476479830130903,
+      "loss": 2.5603,
+      "step": 4376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015470635382987057,
+      "loss": 2.4607,
+      "step": 4377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015464790864324113,
+      "loss": 2.7684,
+      "step": 4378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015458946275030225,
+      "loss": 2.3732,
+      "step": 4379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015453101615993552,
+      "loss": 2.6615,
+      "step": 4380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001544725688810227,
+      "loss": 2.3725,
+      "step": 4381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015441412092244552,
+      "loss": 2.5043,
+      "step": 4382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015435567229308598,
+      "loss": 1.8602,
+      "step": 4383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015429722300182606,
+      "loss": 2.2727,
+      "step": 4384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015423877305754792,
+      "loss": 3.0989,
+      "step": 4385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015418032246913372,
+      "loss": 2.3763,
+      "step": 4386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001541218712454658,
+      "loss": 2.4752,
+      "step": 4387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015406341939542658,
+      "loss": 2.4686,
+      "step": 4388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001540049669278986,
+      "loss": 3.0802,
+      "step": 4389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015394651385176435,
+      "loss": 2.8014,
+      "step": 4390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015388806017590674,
+      "loss": 3.2674,
+      "step": 4391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015382960590920833,
+      "loss": 2.9427,
+      "step": 4392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015377115106055214,
+      "loss": 2.7936,
+      "step": 4393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015371269563882106,
+      "loss": 2.2048,
+      "step": 4394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015365423965289817,
+      "loss": 2.7655,
+      "step": 4395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001535957831116666,
+      "loss": 2.7487,
+      "step": 4396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015353732602400963,
+      "loss": 2.8071,
+      "step": 4397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015347886839881049,
+      "loss": 2.4228,
+      "step": 4398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001534204102449526,
+      "loss": 2.2096,
+      "step": 4399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001533619515713194,
+      "loss": 2.8637,
+      "step": 4400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015330349238679448,
+      "loss": 2.7617,
+      "step": 4401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015324503270026143,
+      "loss": 2.5075,
+      "step": 4402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015318657252060396,
+      "loss": 2.5064,
+      "step": 4403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001531281118567059,
+      "loss": 2.861,
+      "step": 4404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000153069650717451,
+      "loss": 2.676,
+      "step": 4405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015301118911172329,
+      "loss": 2.4221,
+      "step": 4406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015295272704840662,
+      "loss": 2.357,
+      "step": 4407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001528942645363852,
+      "loss": 2.3968,
+      "step": 4408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015283580158454306,
+      "loss": 4.2249,
+      "step": 4409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001527773382017644,
+      "loss": 2.2842,
+      "step": 4410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00015271887439693356,
+      "loss": 2.5497,
+      "step": 4411
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001526604101789348,
+      "loss": 2.603,
+      "step": 4412
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015260194555665255,
+      "loss": 1.6549,
+      "step": 4413
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001525434805389712,
+      "loss": 2.8153,
+      "step": 4414
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015248501513477535,
+      "loss": 2.2302,
+      "step": 4415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001524265493529495,
+      "loss": 2.6134,
+      "step": 4416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015236808320237826,
+      "loss": 2.8008,
+      "step": 4417
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001523096166919464,
+      "loss": 2.8076,
+      "step": 4418
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015225114983053858,
+      "loss": 2.5783,
+      "step": 4419
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015219268262703966,
+      "loss": 2.8615,
+      "step": 4420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015213421509033438,
+      "loss": 2.194,
+      "step": 4421
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015207574722930775,
+      "loss": 2.317,
+      "step": 4422
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001520172790528446,
+      "loss": 2.2465,
+      "step": 4423
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015195881056983,
+      "loss": 2.1413,
+      "step": 4424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015190034178914896,
+      "loss": 2.8946,
+      "step": 4425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015184187271968658,
+      "loss": 2.7642,
+      "step": 4426
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015178340337032795,
+      "loss": 2.6434,
+      "step": 4427
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015172493374995827,
+      "loss": 2.0589,
+      "step": 4428
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001516664638674628,
+      "loss": 2.464,
+      "step": 4429
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015160799373172665,
+      "loss": 2.0905,
+      "step": 4430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001515495233516352,
+      "loss": 3.0755,
+      "step": 4431
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001514910527360738,
+      "loss": 2.3485,
+      "step": 4432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015143258189392777,
+      "loss": 2.9569,
+      "step": 4433
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015137411083408252,
+      "loss": 2.3607,
+      "step": 4434
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001513156395654235,
+      "loss": 2.8969,
+      "step": 4435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015125716809683608,
+      "loss": 2.2728,
+      "step": 4436
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015119869643720585,
+      "loss": 2.051,
+      "step": 4437
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015114022459541827,
+      "loss": 2.1462,
+      "step": 4438
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015108175258035894,
+      "loss": 2.6545,
+      "step": 4439
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001510232804009134,
+      "loss": 2.4428,
+      "step": 4440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001509648080659672,
+      "loss": 2.8479,
+      "step": 4441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001509063355844061,
+      "loss": 2.3266,
+      "step": 4442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015084786296511561,
+      "loss": 2.6929,
+      "step": 4443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015078939021698144,
+      "loss": 2.809,
+      "step": 4444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001507309173488893,
+      "loss": 2.272,
+      "step": 4445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015067244436972488,
+      "loss": 1.9906,
+      "step": 4446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001506139712883739,
+      "loss": 1.9428,
+      "step": 4447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001505554981137221,
+      "loss": 2.8664,
+      "step": 4448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015049702485465526,
+      "loss": 2.9306,
+      "step": 4449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015043855152005906,
+      "loss": 2.6686,
+      "step": 4450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015038007811881928,
+      "loss": 2.5717,
+      "step": 4451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001503216046598218,
+      "loss": 3.0265,
+      "step": 4452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015026313115195237,
+      "loss": 2.4443,
+      "step": 4453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015020465760409675,
+      "loss": 2.7168,
+      "step": 4454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015014618402514076,
+      "loss": 2.8719,
+      "step": 4455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015008771042397028,
+      "loss": 2.4637,
+      "step": 4456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00015002923680947105,
+      "loss": 2.7771,
+      "step": 4457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014997076319052895,
+      "loss": 2.9686,
+      "step": 4458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014991228957602967,
+      "loss": 2.5199,
+      "step": 4459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014985381597485922,
+      "loss": 2.8802,
+      "step": 4460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001497953423959032,
+      "loss": 2.5378,
+      "step": 4461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014973686884804763,
+      "loss": 3.1396,
+      "step": 4462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014967839534017814,
+      "loss": 1.9524,
+      "step": 4463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001496199218811807,
+      "loss": 2.0853,
+      "step": 4464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014956144847994094,
+      "loss": 2.6192,
+      "step": 4465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014950297514534474,
+      "loss": 2.7605,
+      "step": 4466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014944450188627786,
+      "loss": 1.9389,
+      "step": 4467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001493860287116261,
+      "loss": 2.4578,
+      "step": 4468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014932755563027507,
+      "loss": 3.3715,
+      "step": 4469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014926908265111067,
+      "loss": 2.6234,
+      "step": 4470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014921060978301853,
+      "loss": 2.9567,
+      "step": 4471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001491521370348844,
+      "loss": 2.6748,
+      "step": 4472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014909366441559388,
+      "loss": 2.4914,
+      "step": 4473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014903519193403277,
+      "loss": 1.7941,
+      "step": 4474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014897671959908658,
+      "loss": 2.0164,
+      "step": 4475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014891824741964106,
+      "loss": 2.3144,
+      "step": 4476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014885977540458168,
+      "loss": 2.3724,
+      "step": 4477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014880130356279415,
+      "loss": 2.8549,
+      "step": 4478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014874283190316392,
+      "loss": 3.0197,
+      "step": 4479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001486843604345765,
+      "loss": 2.4833,
+      "step": 4480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014862588916591746,
+      "loss": 2.7609,
+      "step": 4481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014856741810607223,
+      "loss": 1.8935,
+      "step": 4482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014850894726392615,
+      "loss": 2.5461,
+      "step": 4483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014845047664836477,
+      "loss": 2.9919,
+      "step": 4484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014839200626827332,
+      "loss": 2.3431,
+      "step": 4485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014833353613253721,
+      "loss": 2.4133,
+      "step": 4486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014827506625004165,
+      "loss": 2.5876,
+      "step": 4487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014821659662967202,
+      "loss": 2.2104,
+      "step": 4488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001481581272803134,
+      "loss": 2.4083,
+      "step": 4489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000148099658210851,
+      "loss": 1.2705,
+      "step": 4490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014804118943016994,
+      "loss": 2.5005,
+      "step": 4491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001479827209471554,
+      "loss": 2.7928,
+      "step": 4492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014792425277069225,
+      "loss": 2.986,
+      "step": 4493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001478657849096656,
+      "loss": 2.1457,
+      "step": 4494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014780731737296037,
+      "loss": 2.5703,
+      "step": 4495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001477488501694614,
+      "loss": 2.64,
+      "step": 4496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014769038330805358,
+      "loss": 2.2901,
+      "step": 4497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014763191679762172,
+      "loss": 2.9809,
+      "step": 4498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014757345064705054,
+      "loss": 1.9999,
+      "step": 4499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014751498486522465,
+      "loss": 2.6175,
+      "step": 4500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001474565194610288,
+      "loss": 2.4069,
+      "step": 4501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014739805444334743,
+      "loss": 2.4655,
+      "step": 4502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014733958982106522,
+      "loss": 2.1588,
+      "step": 4503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014728112560306641,
+      "loss": 2.8891,
+      "step": 4504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001472226617982356,
+      "loss": 2.5234,
+      "step": 4505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014716419841545695,
+      "loss": 3.1343,
+      "step": 4506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014710573546361484,
+      "loss": 2.126,
+      "step": 4507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014704727295159335,
+      "loss": 2.5032,
+      "step": 4508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014698881088827677,
+      "loss": 2.414,
+      "step": 4509
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000146930349282549,
+      "loss": 2.311,
+      "step": 4510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014687188814329412,
+      "loss": 2.3194,
+      "step": 4511
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014681342747939602,
+      "loss": 2.1684,
+      "step": 4512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001467549672997386,
+      "loss": 2.8967,
+      "step": 4513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014669650761320552,
+      "loss": 2.7157,
+      "step": 4514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014663804842868062,
+      "loss": 2.2808,
+      "step": 4515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001465795897550474,
+      "loss": 2.5586,
+      "step": 4516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014652113160118957,
+      "loss": 1.9469,
+      "step": 4517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014646267397599037,
+      "loss": 2.9585,
+      "step": 4518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001464042168883334,
+      "loss": 2.5881,
+      "step": 4519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014634576034710183,
+      "loss": 1.2484,
+      "step": 4520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014628730436117896,
+      "loss": 3.216,
+      "step": 4521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014622884893944786,
+      "loss": 2.5222,
+      "step": 4522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014617039409079168,
+      "loss": 2.0856,
+      "step": 4523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014611193982409326,
+      "loss": 2.5412,
+      "step": 4524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014605348614823562,
+      "loss": 2.2553,
+      "step": 4525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014599503307210143,
+      "loss": 2.644,
+      "step": 4526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014593658060457342,
+      "loss": 2.4239,
+      "step": 4527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001458781287545342,
+      "loss": 2.295,
+      "step": 4528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001458196775308663,
+      "loss": 2.8954,
+      "step": 4529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014576122694245208,
+      "loss": 2.8818,
+      "step": 4530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001457027769981739,
+      "loss": 2.728,
+      "step": 4531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000145644327706914,
+      "loss": 2.7,
+      "step": 4532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014558587907755445,
+      "loss": 2.3427,
+      "step": 4533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014552743111897727,
+      "loss": 1.9764,
+      "step": 4534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014546898384006445,
+      "loss": 2.1911,
+      "step": 4535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014541053724969772,
+      "loss": 2.9022,
+      "step": 4536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014535209135675887,
+      "loss": 2.4294,
+      "step": 4537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001452936461701294,
+      "loss": 2.5865,
+      "step": 4538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014523520169869094,
+      "loss": 2.2765,
+      "step": 4539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014517675795132476,
+      "loss": 2.2681,
+      "step": 4540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001451183149369122,
+      "loss": 2.8464,
+      "step": 4541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014505987266433438,
+      "loss": 2.679,
+      "step": 4542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014500143114247246,
+      "loss": 2.0934,
+      "step": 4543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014494299038020722,
+      "loss": 2.6306,
+      "step": 4544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014488455038641964,
+      "loss": 2.5218,
+      "step": 4545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014482611116999029,
+      "loss": 2.7973,
+      "step": 4546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014476767273979992,
+      "loss": 3.0175,
+      "step": 4547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014470923510472882,
+      "loss": 2.2764,
+      "step": 4548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014465079827365748,
+      "loss": 3.5739,
+      "step": 4549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014459236225546605,
+      "loss": 2.144,
+      "step": 4550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001445339270590347,
+      "loss": 3.2617,
+      "step": 4551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001444754926932433,
+      "loss": 2.4807,
+      "step": 4552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014441705916697184,
+      "loss": 2.1188,
+      "step": 4553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014435862648909997,
+      "loss": 2.648,
+      "step": 4554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001443001946685073,
+      "loss": 3.0433,
+      "step": 4555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014424176371407324,
+      "loss": 2.7607,
+      "step": 4556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014418333363467724,
+      "loss": 2.3943,
+      "step": 4557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014412490443919834,
+      "loss": 2.6843,
+      "step": 4558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001440664761365158,
+      "loss": 2.7069,
+      "step": 4559
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014400804873550835,
+      "loss": 2.8604,
+      "step": 4560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000143949622245055,
+      "loss": 2.9786,
+      "step": 4561
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001438911966740342,
+      "loss": 2.767,
+      "step": 4562
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014383277203132465,
+      "loss": 2.5504,
+      "step": 4563
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014377434832580456,
+      "loss": 2.7626,
+      "step": 4564
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014371592556635228,
+      "loss": 2.5398,
+      "step": 4565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001436575037618458,
+      "loss": 3.0318,
+      "step": 4566
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014359908292116321,
+      "loss": 2.4651,
+      "step": 4567
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014354066305318218,
+      "loss": 1.8278,
+      "step": 4568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014348224416678036,
+      "loss": 1.7024,
+      "step": 4569
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014342382627083536,
+      "loss": 2.2868,
+      "step": 4570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014336540937422441,
+      "loss": 1.4303,
+      "step": 4571
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014330699348582475,
+      "loss": 2.5888,
+      "step": 4572
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014324857861451346,
+      "loss": 2.4617,
+      "step": 4573
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001431901647691674,
+      "loss": 2.919,
+      "step": 4574
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014313175195866326,
+      "loss": 2.9256,
+      "step": 4575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014307334019187775,
+      "loss": 2.4424,
+      "step": 4576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014301492947768707,
+      "loss": 2.5385,
+      "step": 4577
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014295651982496773,
+      "loss": 2.528,
+      "step": 4578
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014289811124259565,
+      "loss": 2.111,
+      "step": 4579
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014283970373944686,
+      "loss": 2.3258,
+      "step": 4580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014278129732439704,
+      "loss": 2.8264,
+      "step": 4581
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014272289200632188,
+      "loss": 1.862,
+      "step": 4582
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014266448779409676,
+      "loss": 2.5076,
+      "step": 4583
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014260608469659702,
+      "loss": 2.8167,
+      "step": 4584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001425476827226977,
+      "loss": 2.721,
+      "step": 4585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014248928188127373,
+      "loss": 2.3593,
+      "step": 4586
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014243088218119992,
+      "loss": 2.2044,
+      "step": 4587
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001423724836313508,
+      "loss": 2.5387,
+      "step": 4588
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014231408624060078,
+      "loss": 2.8137,
+      "step": 4589
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001422556900178242,
+      "loss": 2.8876,
+      "step": 4590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001421972949718949,
+      "loss": 2.2103,
+      "step": 4591
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014213890111168702,
+      "loss": 2.4704,
+      "step": 4592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014208050844607403,
+      "loss": 2.545,
+      "step": 4593
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001420221169839296,
+      "loss": 2.8138,
+      "step": 4594
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014196372673412695,
+      "loss": 2.4476,
+      "step": 4595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014190533770553935,
+      "loss": 2.4232,
+      "step": 4596
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014184694990703963,
+      "loss": 2.2107,
+      "step": 4597
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001417885633475007,
+      "loss": 2.455,
+      "step": 4598
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014173017803579502,
+      "loss": 3.0887,
+      "step": 4599
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001416717939807951,
+      "loss": 2.5904,
+      "step": 4600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001416134111913731,
+      "loss": 2.8741,
+      "step": 4601
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000141555029676401,
+      "loss": 2.3974,
+      "step": 4602
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014149664944475065,
+      "loss": 1.9438,
+      "step": 4603
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014143827050529373,
+      "loss": 2.7149,
+      "step": 4604
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014137989286690155,
+      "loss": 2.9968,
+      "step": 4605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014132151653844548,
+      "loss": 2.4867,
+      "step": 4606
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014126314152879642,
+      "loss": 2.0526,
+      "step": 4607
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014120476784682535,
+      "loss": 2.2307,
+      "step": 4608
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014114639550140274,
+      "loss": 2.78,
+      "step": 4609
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014108802450139916,
+      "loss": 2.1776,
+      "step": 4610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001410296548556847,
+      "loss": 2.5981,
+      "step": 4611
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014097128657312953,
+      "loss": 2.4155,
+      "step": 4612
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014091291966260327,
+      "loss": 2.5396,
+      "step": 4613
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014085455413297574,
+      "loss": 2.6959,
+      "step": 4614
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014079618999311618,
+      "loss": 2.4993,
+      "step": 4615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001407378272518938,
+      "loss": 2.4977,
+      "step": 4616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001406794659181776,
+      "loss": 2.7454,
+      "step": 4617
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014062110600083634,
+      "loss": 2.3279,
+      "step": 4618
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001405627475087385,
+      "loss": 1.5314,
+      "step": 4619
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014050439045075247,
+      "loss": 2.3278,
+      "step": 4620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014044603483574626,
+      "loss": 2.2929,
+      "step": 4621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001403876806725879,
+      "loss": 2.3414,
+      "step": 4622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014032932797014493,
+      "loss": 3.1559,
+      "step": 4623
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014027097673728487,
+      "loss": 2.7734,
+      "step": 4624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014021262698287484,
+      "loss": 3.1331,
+      "step": 4625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014015427871578198,
+      "loss": 2.4025,
+      "step": 4626
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001400959319448729,
+      "loss": 2.5577,
+      "step": 4627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00014003758667901432,
+      "loss": 2.4068,
+      "step": 4628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001399792429270724,
+      "loss": 2.5856,
+      "step": 4629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013992090069791328,
+      "loss": 2.9206,
+      "step": 4630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013986256000040283,
+      "loss": 2.9525,
+      "step": 4631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013980422084340665,
+      "loss": 1.2485,
+      "step": 4632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013974588323579007,
+      "loss": 2.3575,
+      "step": 4633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013968754718641837,
+      "loss": 2.5592,
+      "step": 4634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001396292127041563,
+      "loss": 2.8151,
+      "step": 4635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013957087979786867,
+      "loss": 2.6339,
+      "step": 4636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013951254847641983,
+      "loss": 1.6272,
+      "step": 4637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013945421874867408,
+      "loss": 1.8212,
+      "step": 4638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001393958906234952,
+      "loss": 2.7373,
+      "step": 4639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013933756410974707,
+      "loss": 2.5187,
+      "step": 4640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001392792392162931,
+      "loss": 2.4585,
+      "step": 4641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001392209159519964,
+      "loss": 2.1192,
+      "step": 4642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001391625943257201,
+      "loss": 2.5316,
+      "step": 4643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013910427434632677,
+      "loss": 2.8677,
+      "step": 4644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013904595602267904,
+      "loss": 2.2403,
+      "step": 4645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000138987639363639,
+      "loss": 2.8129,
+      "step": 4646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013892932437806866,
+      "loss": 2.644,
+      "step": 4647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013887101107482974,
+      "loss": 2.6659,
+      "step": 4648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001388126994627837,
+      "loss": 2.4971,
+      "step": 4649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013875438955079168,
+      "loss": 2.5878,
+      "step": 4650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013869608134771475,
+      "loss": 2.4124,
+      "step": 4651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013863777486241343,
+      "loss": 2.4643,
+      "step": 4652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013857947010374828,
+      "loss": 1.0828,
+      "step": 4653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013852116708057934,
+      "loss": 2.5333,
+      "step": 4654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001384628658017666,
+      "loss": 2.7642,
+      "step": 4655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001384045662761696,
+      "loss": 2.8453,
+      "step": 4656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001383462685126478,
+      "loss": 2.5957,
+      "step": 4657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013828797252006018,
+      "loss": 2.7767,
+      "step": 4658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001382296783072657,
+      "loss": 2.931,
+      "step": 4659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001381713858831228,
+      "loss": 1.9241,
+      "step": 4660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013811309525648985,
+      "loss": 2.5943,
+      "step": 4661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013805480643622477,
+      "loss": 2.2451,
+      "step": 4662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013799651943118536,
+      "loss": 1.1267,
+      "step": 4663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013793823425022904,
+      "loss": 2.487,
+      "step": 4664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013787995090221308,
+      "loss": 2.6329,
+      "step": 4665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013782166939599422,
+      "loss": 2.7765,
+      "step": 4666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013776338974042925,
+      "loss": 2.2568,
+      "step": 4667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013770511194437438,
+      "loss": 2.0462,
+      "step": 4668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013764683601668581,
+      "loss": 2.7251,
+      "step": 4669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013758856196621918,
+      "loss": 2.5882,
+      "step": 4670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013753028980183007,
+      "loss": 2.5524,
+      "step": 4671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001374720195323736,
+      "loss": 2.7956,
+      "step": 4672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001374137511667048,
+      "loss": 2.552,
+      "step": 4673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001373554847136782,
+      "loss": 2.4787,
+      "step": 4674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001372972201821482,
+      "loss": 2.4185,
+      "step": 4675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001372389575809688,
+      "loss": 2.7248,
+      "step": 4676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001371806969189938,
+      "loss": 2.4495,
+      "step": 4677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001371224382050766,
+      "loss": 2.7762,
+      "step": 4678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001370641814480704,
+      "loss": 2.632,
+      "step": 4679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013700592665682798,
+      "loss": 3.108,
+      "step": 4680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013694767384020206,
+      "loss": 0.8831,
+      "step": 4681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013688942300704474,
+      "loss": 2.6275,
+      "step": 4682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013683117416620814,
+      "loss": 2.3572,
+      "step": 4683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013677292732654374,
+      "loss": 2.1792,
+      "step": 4684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013671468249690306,
+      "loss": 2.998,
+      "step": 4685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013665643968613706,
+      "loss": 2.098,
+      "step": 4686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013659819890309653,
+      "loss": 1.8209,
+      "step": 4687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013653996015663183,
+      "loss": 2.355,
+      "step": 4688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001364817234555932,
+      "loss": 2.0718,
+      "step": 4689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013642348880883037,
+      "loss": 2.5649,
+      "step": 4690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013636525622519286,
+      "loss": 2.9049,
+      "step": 4691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013630702571352986,
+      "loss": 2.6581,
+      "step": 4692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013624879728269028,
+      "loss": 2.3006,
+      "step": 4693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001361905709415226,
+      "loss": 2.3012,
+      "step": 4694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013613234669887517,
+      "loss": 1.7491,
+      "step": 4695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013607412456359576,
+      "loss": 2.6481,
+      "step": 4696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013601590454453215,
+      "loss": 2.3636,
+      "step": 4697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013595768665053146,
+      "loss": 2.7094,
+      "step": 4698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001358994708904408,
+      "loss": 2.4409,
+      "step": 4699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013584125727310664,
+      "loss": 2.7417,
+      "step": 4700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013578304580737544,
+      "loss": 2.1004,
+      "step": 4701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013572483650209306,
+      "loss": 2.7714,
+      "step": 4702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013566662936610524,
+      "loss": 2.8876,
+      "step": 4703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013560842440825724,
+      "loss": 2.1481,
+      "step": 4704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013555022163739408,
+      "loss": 2.2086,
+      "step": 4705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001354920210623604,
+      "loss": 2.3613,
+      "step": 4706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013543382269200059,
+      "loss": 2.5492,
+      "step": 4707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001353756265351585,
+      "loss": 2.1672,
+      "step": 4708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013531743260067797,
+      "loss": 2.7246,
+      "step": 4709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001352592408974021,
+      "loss": 2.7718,
+      "step": 4710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001352010514341741,
+      "loss": 2.4042,
+      "step": 4711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001351428642198365,
+      "loss": 2.4634,
+      "step": 4712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001350846792632315,
+      "loss": 2.5467,
+      "step": 4713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001350264965732012,
+      "loss": 2.3912,
+      "step": 4714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001349683161585871,
+      "loss": 2.9097,
+      "step": 4715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013491013802823055,
+      "loss": 2.0861,
+      "step": 4716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013485196219097237,
+      "loss": 2.4267,
+      "step": 4717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013479378865565323,
+      "loss": 2.6669,
+      "step": 4718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013473561743111327,
+      "loss": 2.3575,
+      "step": 4719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001346774485261924,
+      "loss": 2.7199,
+      "step": 4720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013461928194973008,
+      "loss": 2.3024,
+      "step": 4721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013456111771056555,
+      "loss": 2.9097,
+      "step": 4722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013450295581753753,
+      "loss": 2.9022,
+      "step": 4723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001344447962794845,
+      "loss": 2.5244,
+      "step": 4724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013438663910524457,
+      "loss": 2.4557,
+      "step": 4725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013432848430365544,
+      "loss": 2.7948,
+      "step": 4726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013427033188355447,
+      "loss": 2.847,
+      "step": 4727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013421218185377873,
+      "loss": 2.1459,
+      "step": 4728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013415403422316477,
+      "loss": 2.5958,
+      "step": 4729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000134095889000549,
+      "loss": 2.6452,
+      "step": 4730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013403774619476715,
+      "loss": 2.698,
+      "step": 4731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013397960581465495,
+      "loss": 2.702,
+      "step": 4732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013392146786904747,
+      "loss": 2.6399,
+      "step": 4733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013386333236677957,
+      "loss": 3.1887,
+      "step": 4734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001338051993166856,
+      "loss": 2.1269,
+      "step": 4735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013374706872759976,
+      "loss": 2.589,
+      "step": 4736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013368894060835566,
+      "loss": 2.1973,
+      "step": 4737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001336308149677866,
+      "loss": 2.6203,
+      "step": 4738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013357269181472556,
+      "loss": 2.6595,
+      "step": 4739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013351457115800508,
+      "loss": 2.3609,
+      "step": 4740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001334564530064573,
+      "loss": 2.563,
+      "step": 4741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013339833736891412,
+      "loss": 2.9308,
+      "step": 4742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013334022425420683,
+      "loss": 1.1567,
+      "step": 4743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013328211367116658,
+      "loss": 1.1004,
+      "step": 4744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013322400562862396,
+      "loss": 2.4587,
+      "step": 4745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013316590013540925,
+      "loss": 2.5611,
+      "step": 4746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001331077972003523,
+      "loss": 2.7283,
+      "step": 4747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013304969683228268,
+      "loss": 2.8275,
+      "step": 4748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013299159904002936,
+      "loss": 2.7907,
+      "step": 4749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013293350383242116,
+      "loss": 2.9788,
+      "step": 4750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013287541121828633,
+      "loss": 2.7852,
+      "step": 4751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013281732120645282,
+      "loss": 2.6788,
+      "step": 4752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013275923380574813,
+      "loss": 2.7189,
+      "step": 4753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013270114902499941,
+      "loss": 2.5028,
+      "step": 4754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013264306687303334,
+      "loss": 2.763,
+      "step": 4755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013258498735867634,
+      "loss": 2.6454,
+      "step": 4756
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013252691049075421,
+      "loss": 2.638,
+      "step": 4757
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013246883627809263,
+      "loss": 2.5937,
+      "step": 4758
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013241076472951657,
+      "loss": 1.9887,
+      "step": 4759
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013235269585385088,
+      "loss": 2.5069,
+      "step": 4760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013229462965991975,
+      "loss": 2.4579,
+      "step": 4761
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013223656615654722,
+      "loss": 2.279,
+      "step": 4762
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013217850535255664,
+      "loss": 2.2172,
+      "step": 4763
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013212044725677125,
+      "loss": 2.0299,
+      "step": 4764
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013206239187801358,
+      "loss": 2.6887,
+      "step": 4765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013200433922510598,
+      "loss": 2.7381,
+      "step": 4766
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013194628930687026,
+      "loss": 2.7415,
+      "step": 4767
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001318882421321279,
+      "loss": 2.3276,
+      "step": 4768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013183019770969978,
+      "loss": 2.2377,
+      "step": 4769
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013177215604840667,
+      "loss": 2.1175,
+      "step": 4770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001317141171570686,
+      "loss": 2.2204,
+      "step": 4771
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013165608104450549,
+      "loss": 2.5214,
+      "step": 4772
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013159804771953646,
+      "loss": 2.6982,
+      "step": 4773
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001315400171909806,
+      "loss": 3.0425,
+      "step": 4774
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001314819894676563,
+      "loss": 2.4497,
+      "step": 4775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013142396455838169,
+      "loss": 2.8322,
+      "step": 4776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013136594247197428,
+      "loss": 2.6603,
+      "step": 4777
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013130792321725145,
+      "loss": 2.3701,
+      "step": 4778
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001312499068030298,
+      "loss": 2.7288,
+      "step": 4779
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013119189323812576,
+      "loss": 2.6041,
+      "step": 4780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001311338825313552,
+      "loss": 1.5182,
+      "step": 4781
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013107587469153361,
+      "loss": 2.4692,
+      "step": 4782
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013101786972747603,
+      "loss": 2.5707,
+      "step": 4783
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001309598676479971,
+      "loss": 2.2276,
+      "step": 4784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001309018684619109,
+      "loss": 2.4837,
+      "step": 4785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001308438721780312,
+      "loss": 1.9299,
+      "step": 4786
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013078587880517125,
+      "loss": 1.9339,
+      "step": 4787
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013072788835214388,
+      "loss": 3.1482,
+      "step": 4788
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013066990082776157,
+      "loss": 2.5928,
+      "step": 4789
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013061191624083615,
+      "loss": 2.5491,
+      "step": 4790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013055393460017923,
+      "loss": 2.8409,
+      "step": 4791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013049595591460176,
+      "loss": 2.6159,
+      "step": 4792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013043798019291448,
+      "loss": 2.2344,
+      "step": 4793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001303800074439274,
+      "loss": 2.6329,
+      "step": 4794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013032203767645036,
+      "loss": 2.5262,
+      "step": 4795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013026407089929246,
+      "loss": 2.3005,
+      "step": 4796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013020610712126266,
+      "loss": 2.4952,
+      "step": 4797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001301481463511692,
+      "loss": 2.7247,
+      "step": 4798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013009018859782,
+      "loss": 2.462,
+      "step": 4799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00013003223387002245,
+      "loss": 2.4326,
+      "step": 4800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012997428217658357,
+      "loss": 2.9921,
+      "step": 4801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012991633352630978,
+      "loss": 2.4257,
+      "step": 4802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012985838792800724,
+      "loss": 2.6835,
+      "step": 4803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012980044539048137,
+      "loss": 2.4777,
+      "step": 4804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012974250592253747,
+      "loss": 2.7306,
+      "step": 4805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012968456953298002,
+      "loss": 2.4434,
+      "step": 4806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012962663623061334,
+      "loss": 2.4531,
+      "step": 4807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000129568706024241,
+      "loss": 2.7897,
+      "step": 4808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012951077892266638,
+      "loss": 2.7601,
+      "step": 4809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001294528549346921,
+      "loss": 2.645,
+      "step": 4810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001293949340691206,
+      "loss": 2.473,
+      "step": 4811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001293370163347536,
+      "loss": 2.9,
+      "step": 4812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001292791017403925,
+      "loss": 1.5271,
+      "step": 4813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012922119029483813,
+      "loss": 2.8935,
+      "step": 4814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001291632820068909,
+      "loss": 2.7938,
+      "step": 4815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012910537688535066,
+      "loss": 2.3846,
+      "step": 4816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012904747493901696,
+      "loss": 2.5464,
+      "step": 4817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001289895761766886,
+      "loss": 2.7135,
+      "step": 4818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012893168060716415,
+      "loss": 2.3647,
+      "step": 4819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001288737882392415,
+      "loss": 3.0359,
+      "step": 4820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001288158990817183,
+      "loss": 2.8284,
+      "step": 4821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001287580131433913,
+      "loss": 3.469,
+      "step": 4822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012870013043305727,
+      "loss": 2.2655,
+      "step": 4823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000128642250959512,
+      "loss": 2.503,
+      "step": 4824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012858437473155124,
+      "loss": 2.6654,
+      "step": 4825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012852650175796984,
+      "loss": 2.5698,
+      "step": 4826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012846863204756244,
+      "loss": 2.1068,
+      "step": 4827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012841076560912308,
+      "loss": 1.0898,
+      "step": 4828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012835290245144534,
+      "loss": 2.2967,
+      "step": 4829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012829504258332213,
+      "loss": 2.6498,
+      "step": 4830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012823718601354616,
+      "loss": 2.7256,
+      "step": 4831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001281793327509094,
+      "loss": 2.4023,
+      "step": 4832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001281214828042034,
+      "loss": 2.3348,
+      "step": 4833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001280636361822192,
+      "loss": 2.8805,
+      "step": 4834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001280057928937474,
+      "loss": 2.1241,
+      "step": 4835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012794795294757792,
+      "loss": 2.6465,
+      "step": 4836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001278901163525004,
+      "loss": 1.6527,
+      "step": 4837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012783228311730378,
+      "loss": 2.4482,
+      "step": 4838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001277744532507766,
+      "loss": 2.8392,
+      "step": 4839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012771662676170682,
+      "loss": 2.6375,
+      "step": 4840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001276588036588819,
+      "loss": 2.5767,
+      "step": 4841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012760098395108889,
+      "loss": 2.029,
+      "step": 4842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012754316764711416,
+      "loss": 2.5441,
+      "step": 4843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012748535475574365,
+      "loss": 2.7877,
+      "step": 4844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012742754528576283,
+      "loss": 2.883,
+      "step": 4845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012736973924595648,
+      "loss": 2.5361,
+      "step": 4846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001273119366451091,
+      "loss": 1.9554,
+      "step": 4847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012725413749200442,
+      "loss": 2.6869,
+      "step": 4848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012719634179542586,
+      "loss": 2.6605,
+      "step": 4849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012713854956415615,
+      "loss": 2.3664,
+      "step": 4850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012708076080697762,
+      "loss": 2.5547,
+      "step": 4851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012702297553267194,
+      "loss": 2.9861,
+      "step": 4852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012696519375002044,
+      "loss": 2.738,
+      "step": 4853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012690741546780366,
+      "loss": 2.8246,
+      "step": 4854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001268496406948019,
+      "loss": 2.6476,
+      "step": 4855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012679186943979467,
+      "loss": 2.3335,
+      "step": 4856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012673410171156106,
+      "loss": 2.8602,
+      "step": 4857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001266763375188797,
+      "loss": 2.6638,
+      "step": 4858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012661857687052853,
+      "loss": 1.7893,
+      "step": 4859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012656081977528504,
+      "loss": 2.51,
+      "step": 4860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012650306624192615,
+      "loss": 2.6957,
+      "step": 4861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012644531627922833,
+      "loss": 2.6763,
+      "step": 4862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012638756989596727,
+      "loss": 2.9073,
+      "step": 4863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012632982710091844,
+      "loss": 2.5318,
+      "step": 4864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001262720879028565,
+      "loss": 3.1688,
+      "step": 4865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001262143523105557,
+      "loss": 1.576,
+      "step": 4866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012615662033278968,
+      "loss": 2.3809,
+      "step": 4867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001260988919783316,
+      "loss": 3.166,
+      "step": 4868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012604116725595398,
+      "loss": 2.8451,
+      "step": 4869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001259834461744289,
+      "loss": 2.673,
+      "step": 4870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012592572874252768,
+      "loss": 2.8236,
+      "step": 4871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012586801496902142,
+      "loss": 2.5042,
+      "step": 4872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001258103048626803,
+      "loss": 2.3953,
+      "step": 4873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012575259843227422,
+      "loss": 2.6781,
+      "step": 4874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012569489568657233,
+      "loss": 2.5499,
+      "step": 4875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012563719663434341,
+      "loss": 2.2737,
+      "step": 4876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012557950128435545,
+      "loss": 2.4485,
+      "step": 4877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001255218096453761,
+      "loss": 2.0605,
+      "step": 4878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012546412172617226,
+      "loss": 2.4759,
+      "step": 4879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012540643753551045,
+      "loss": 2.0677,
+      "step": 4880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001253487570821564,
+      "loss": 3.0349,
+      "step": 4881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012529108037487554,
+      "loss": 3.1156,
+      "step": 4882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012523340742243247,
+      "loss": 2.3845,
+      "step": 4883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001251757382335914,
+      "loss": 2.6003,
+      "step": 4884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012511807281711585,
+      "loss": 2.2038,
+      "step": 4885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012506041118176892,
+      "loss": 2.6577,
+      "step": 4886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001250027533363129,
+      "loss": 2.5756,
+      "step": 4887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012494509928950975,
+      "loss": 1.0951,
+      "step": 4888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001248874490501207,
+      "loss": 2.265,
+      "step": 4889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012482980262690647,
+      "loss": 3.1466,
+      "step": 4890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012477216002862706,
+      "loss": 2.1056,
+      "step": 4891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001247145212640422,
+      "loss": 2.3866,
+      "step": 4892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012465688634191068,
+      "loss": 2.5564,
+      "step": 4893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012459925527099095,
+      "loss": 2.7679,
+      "step": 4894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012454162806004076,
+      "loss": 2.8071,
+      "step": 4895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012448400471781732,
+      "loss": 2.1862,
+      "step": 4896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012442638525307718,
+      "loss": 3.1265,
+      "step": 4897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001243687696745765,
+      "loss": 2.6684,
+      "step": 4898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012431115799107053,
+      "loss": 2.6715,
+      "step": 4899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012425355021131426,
+      "loss": 1.4357,
+      "step": 4900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001241959463440618,
+      "loss": 2.6371,
+      "step": 4901
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001241383463980669,
+      "loss": 2.4697,
+      "step": 4902
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012408075038208258,
+      "loss": 2.5303,
+      "step": 4903
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001240231583048613,
+      "loss": 2.167,
+      "step": 4904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001239655701751548,
+      "loss": 1.9817,
+      "step": 4905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012390798600171455,
+      "loss": 2.8864,
+      "step": 4906
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012385040579329103,
+      "loss": 2.6953,
+      "step": 4907
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012379282955863441,
+      "loss": 2.5091,
+      "step": 4908
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012373525730649402,
+      "loss": 2.9732,
+      "step": 4909
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012367768904561884,
+      "loss": 2.6379,
+      "step": 4910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012362012478475695,
+      "loss": 2.6649,
+      "step": 4911
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012356256453265616,
+      "loss": 2.4874,
+      "step": 4912
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012350500829806332,
+      "loss": 2.5677,
+      "step": 4913
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000123447456089725,
+      "loss": 2.2912,
+      "step": 4914
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001233899079163868,
+      "loss": 2.8644,
+      "step": 4915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012333236378679412,
+      "loss": 1.1667,
+      "step": 4916
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012327482370969137,
+      "loss": 2.5036,
+      "step": 4917
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012321728769382258,
+      "loss": 3.043,
+      "step": 4918
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012315975574793105,
+      "loss": 2.5794,
+      "step": 4919
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012310222788075955,
+      "loss": 2.484,
+      "step": 4920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001230447041010501,
+      "loss": 2.5553,
+      "step": 4921
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012298718441754427,
+      "loss": 2.6185,
+      "step": 4922
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001229296688389828,
+      "loss": 2.7649,
+      "step": 4923
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012287215737410607,
+      "loss": 2.8622,
+      "step": 4924
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001228146500316535,
+      "loss": 1.4145,
+      "step": 4925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001227571468203643,
+      "loss": 2.4231,
+      "step": 4926
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001226996477489766,
+      "loss": 2.3402,
+      "step": 4927
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012264215282622827,
+      "loss": 2.3175,
+      "step": 4928
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001225846620608563,
+      "loss": 2.6179,
+      "step": 4929
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012252717546159728,
+      "loss": 2.8531,
+      "step": 4930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012246969303718693,
+      "loss": 3.1986,
+      "step": 4931
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012241221479636043,
+      "loss": 2.5611,
+      "step": 4932
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001223547407478524,
+      "loss": 2.3891,
+      "step": 4933
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012229727090039676,
+      "loss": 2.529,
+      "step": 4934
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012223980526272676,
+      "loss": 2.4626,
+      "step": 4935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012218234384357504,
+      "loss": 2.625,
+      "step": 4936
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012212488665167364,
+      "loss": 2.048,
+      "step": 4937
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012206743369575385,
+      "loss": 2.7639,
+      "step": 4938
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001220099849845465,
+      "loss": 2.0959,
+      "step": 4939
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001219525405267815,
+      "loss": 2.8262,
+      "step": 4940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012189510033118841,
+      "loss": 2.3204,
+      "step": 4941
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012183766440649593,
+      "loss": 2.2886,
+      "step": 4942
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012178023276143224,
+      "loss": 2.4923,
+      "step": 4943
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012172280540472475,
+      "loss": 2.4761,
+      "step": 4944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001216653823451004,
+      "loss": 2.3323,
+      "step": 4945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012160796359128519,
+      "loss": 2.2348,
+      "step": 4946
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012155054915200481,
+      "loss": 2.0654,
+      "step": 4947
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012149313903598403,
+      "loss": 3.1069,
+      "step": 4948
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012143573325194705,
+      "loss": 2.5559,
+      "step": 4949
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012137833180861745,
+      "loss": 1.1237,
+      "step": 4950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012132093471471813,
+      "loss": 2.6067,
+      "step": 4951
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012126354197897123,
+      "loss": 2.621,
+      "step": 4952
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012120615361009843,
+      "loss": 2.6936,
+      "step": 4953
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012114876961682051,
+      "loss": 3.2207,
+      "step": 4954
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012109139000785785,
+      "loss": 2.9908,
+      "step": 4955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012103401479192985,
+      "loss": 2.7454,
+      "step": 4956
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012097664397775556,
+      "loss": 2.4788,
+      "step": 4957
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001209192775740531,
+      "loss": 2.5154,
+      "step": 4958
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012086191558954016,
+      "loss": 3.0421,
+      "step": 4959
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012080455803293346,
+      "loss": 2.5974,
+      "step": 4960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001207472049129494,
+      "loss": 2.6435,
+      "step": 4961
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001206898562383034,
+      "loss": 2.8989,
+      "step": 4962
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012063251201771035,
+      "loss": 2.1354,
+      "step": 4963
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012057517225988447,
+      "loss": 2.6929,
+      "step": 4964
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012051783697353929,
+      "loss": 1.933,
+      "step": 4965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012046050616738753,
+      "loss": 2.6483,
+      "step": 4966
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001204031798501415,
+      "loss": 2.1175,
+      "step": 4967
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012034585803051253,
+      "loss": 2.7477,
+      "step": 4968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012028854071721152,
+      "loss": 2.2383,
+      "step": 4969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012023122791894848,
+      "loss": 3.0223,
+      "step": 4970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012017391964443294,
+      "loss": 2.9324,
+      "step": 4971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012011661590237345,
+      "loss": 2.8762,
+      "step": 4972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012005931670147826,
+      "loss": 2.5122,
+      "step": 4973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00012000202205045456,
+      "loss": 2.749,
+      "step": 4974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001199447319580091,
+      "loss": 2.7128,
+      "step": 4975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011988744643284777,
+      "loss": 2.695,
+      "step": 4976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011983016548367594,
+      "loss": 2.2479,
+      "step": 4977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001197728891191981,
+      "loss": 2.501,
+      "step": 4978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011971561734811816,
+      "loss": 2.0515,
+      "step": 4979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001196583501791393,
+      "loss": 2.2312,
+      "step": 4980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011960108762096403,
+      "loss": 2.5415,
+      "step": 4981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011954382968229405,
+      "loss": 2.5798,
+      "step": 4982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011948657637183055,
+      "loss": 3.6077,
+      "step": 4983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001194293276982738,
+      "loss": 2.5109,
+      "step": 4984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011937208367032358,
+      "loss": 1.9,
+      "step": 4985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011931484429667872,
+      "loss": 1.6866,
+      "step": 4986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011925760958603764,
+      "loss": 1.8784,
+      "step": 4987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011920037954709771,
+      "loss": 2.3499,
+      "step": 4988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011914315418855597,
+      "loss": 2.2887,
+      "step": 4989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011908593351910834,
+      "loss": 2.0785,
+      "step": 4990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011902871754745045,
+      "loss": 2.6363,
+      "step": 4991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011897150628227681,
+      "loss": 2.1285,
+      "step": 4992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011891429973228154,
+      "loss": 2.4946,
+      "step": 4993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011885709790615782,
+      "loss": 2.1481,
+      "step": 4994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011879990081259829,
+      "loss": 2.7206,
+      "step": 4995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011874270846029466,
+      "loss": 2.9558,
+      "step": 4996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001186855208579382,
+      "loss": 3.196,
+      "step": 4997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011862833801421914,
+      "loss": 2.0245,
+      "step": 4998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011857115993782729,
+      "loss": 3.071,
+      "step": 4999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011851398663745149,
+      "loss": 2.5626,
+      "step": 5000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011845681812178003,
+      "loss": 2.1618,
+      "step": 5001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011839965439950032,
+      "loss": 3.0039,
+      "step": 5002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011834249547929923,
+      "loss": 2.5151,
+      "step": 5003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011828534136986273,
+      "loss": 2.3039,
+      "step": 5004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011822819207987606,
+      "loss": 2.6801,
+      "step": 5005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011817104761802394,
+      "loss": 1.0666,
+      "step": 5006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011811390799299003,
+      "loss": 2.2146,
+      "step": 5007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011805677321345757,
+      "loss": 2.976,
+      "step": 5008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011799964328810888,
+      "loss": 2.1781,
+      "step": 5009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011794251822562554,
+      "loss": 2.6242,
+      "step": 5010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001178853980346885,
+      "loss": 2.2641,
+      "step": 5011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011782828272397792,
+      "loss": 2.2638,
+      "step": 5012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001177711723021731,
+      "loss": 2.8392,
+      "step": 5013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011771406677795283,
+      "loss": 2.301,
+      "step": 5014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011765696615999489,
+      "loss": 2.5941,
+      "step": 5015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011759987045697662,
+      "loss": 2.5481,
+      "step": 5016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011754277967757427,
+      "loss": 2.0623,
+      "step": 5017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001174856938304637,
+      "loss": 2.5408,
+      "step": 5018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011742861292431964,
+      "loss": 2.7254,
+      "step": 5019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011737153696781646,
+      "loss": 2.7209,
+      "step": 5020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011731446596962742,
+      "loss": 2.0513,
+      "step": 5021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011725739993842533,
+      "loss": 2.6378,
+      "step": 5022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011720033888288203,
+      "loss": 2.5328,
+      "step": 5023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011714328281166868,
+      "loss": 2.712,
+      "step": 5024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011708623173345568,
+      "loss": 2.3634,
+      "step": 5025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011702918565691274,
+      "loss": 2.2351,
+      "step": 5026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011697214459070865,
+      "loss": 2.7286,
+      "step": 5027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011691510854351161,
+      "loss": 2.3158,
+      "step": 5028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011685807752398891,
+      "loss": 2.3951,
+      "step": 5029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011680105154080722,
+      "loss": 2.4739,
+      "step": 5030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001167440306026323,
+      "loss": 2.2327,
+      "step": 5031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001166870147181293,
+      "loss": 2.8372,
+      "step": 5032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011663000389596242,
+      "loss": 2.5964,
+      "step": 5033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011657299814479527,
+      "loss": 1.795,
+      "step": 5034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011651599747329053,
+      "loss": 2.677,
+      "step": 5035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011645900189011028,
+      "loss": 2.4303,
+      "step": 5036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011640201140391564,
+      "loss": 2.9014,
+      "step": 5037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011634502602336708,
+      "loss": 1.4011,
+      "step": 5038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011628804575712427,
+      "loss": 2.5066,
+      "step": 5039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011623107061384609,
+      "loss": 2.9559,
+      "step": 5040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011617410060219062,
+      "loss": 2.7716,
+      "step": 5041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011611713573081526,
+      "loss": 1.9954,
+      "step": 5042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011606017600837644,
+      "loss": 2.1336,
+      "step": 5043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011600322144353004,
+      "loss": 2.518,
+      "step": 5044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001159462720449309,
+      "loss": 2.2609,
+      "step": 5045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011588932782123339,
+      "loss": 2.468,
+      "step": 5046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011583238878109072,
+      "loss": 2.3869,
+      "step": 5047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001157754549331557,
+      "loss": 3.033,
+      "step": 5048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011571852628608,
+      "loss": 3.0645,
+      "step": 5049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011566160284851479,
+      "loss": 2.0704,
+      "step": 5050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001156046846291102,
+      "loss": 2.5963,
+      "step": 5051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011554777163651579,
+      "loss": 2.6877,
+      "step": 5052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011549086387938015,
+      "loss": 2.2268,
+      "step": 5053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011543396136635114,
+      "loss": 2.3722,
+      "step": 5054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001153770641060759,
+      "loss": 2.2058,
+      "step": 5055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011532017210720066,
+      "loss": 2.6177,
+      "step": 5056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011526328537837081,
+      "loss": 2.6688,
+      "step": 5057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011520640392823118,
+      "loss": 2.9623,
+      "step": 5058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011514952776542547,
+      "loss": 2.8731,
+      "step": 5059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011509265689859691,
+      "loss": 2.5241,
+      "step": 5060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001150357913363876,
+      "loss": 2.0755,
+      "step": 5061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011497893108743912,
+      "loss": 2.5536,
+      "step": 5062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000114922076160392,
+      "loss": 2.2812,
+      "step": 5063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011486522656388622,
+      "loss": 2.7653,
+      "step": 5064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011480838230656062,
+      "loss": 2.6908,
+      "step": 5065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011475154339705363,
+      "loss": 2.941,
+      "step": 5066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011469470984400249,
+      "loss": 2.4303,
+      "step": 5067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011463788165604386,
+      "loss": 2.4982,
+      "step": 5068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011458105884181347,
+      "loss": 3.1811,
+      "step": 5069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011452424140994636,
+      "loss": 2.1415,
+      "step": 5070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011446742936907653,
+      "loss": 2.4325,
+      "step": 5071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011441062272783745,
+      "loss": 2.7333,
+      "step": 5072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011435382149486147,
+      "loss": 2.6575,
+      "step": 5073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011429702567878044,
+      "loss": 2.4745,
+      "step": 5074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001142402352882251,
+      "loss": 2.9274,
+      "step": 5075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011418345033182546,
+      "loss": 2.3115,
+      "step": 5076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011412667081821079,
+      "loss": 2.4902,
+      "step": 5077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011406989675600941,
+      "loss": 2.0076,
+      "step": 5078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011401312815384894,
+      "loss": 2.8635,
+      "step": 5079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011395636502035601,
+      "loss": 2.1927,
+      "step": 5080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011389960736415661,
+      "loss": 2.5363,
+      "step": 5081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011384285519387569,
+      "loss": 2.6082,
+      "step": 5082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011378610851813758,
+      "loss": 2.3816,
+      "step": 5083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011372936734556557,
+      "loss": 2.6012,
+      "step": 5084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011367263168478226,
+      "loss": 2.1844,
+      "step": 5085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011361590154440933,
+      "loss": 2.3629,
+      "step": 5086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011355917693306773,
+      "loss": 2.9442,
+      "step": 5087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011350245785937738,
+      "loss": 2.9668,
+      "step": 5088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011344574433195757,
+      "loss": 2.463,
+      "step": 5089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00011338903635942656,
+      "loss": 2.1719,
+      "step": 5090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011333233395040196,
+      "loss": 2.6282,
+      "step": 5091
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001132756371135003,
+      "loss": 2.7647,
+      "step": 5092
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011321894585733756,
+      "loss": 2.6656,
+      "step": 5093
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011316226019052856,
+      "loss": 2.399,
+      "step": 5094
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001131055801216875,
+      "loss": 2.6131,
+      "step": 5095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011304890565942756,
+      "loss": 2.5332,
+      "step": 5096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011299223681236127,
+      "loss": 2.4403,
+      "step": 5097
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011293557358910009,
+      "loss": 2.3366,
+      "step": 5098
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011287891599825474,
+      "loss": 2.3123,
+      "step": 5099
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011282226404843512,
+      "loss": 2.2137,
+      "step": 5100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011276561774825021,
+      "loss": 2.7051,
+      "step": 5101
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011270897710630807,
+      "loss": 2.2838,
+      "step": 5102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011265234213121608,
+      "loss": 2.3743,
+      "step": 5103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011259571283158055,
+      "loss": 2.4757,
+      "step": 5104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011253908921600713,
+      "loss": 2.5411,
+      "step": 5105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011248247129310038,
+      "loss": 2.4044,
+      "step": 5106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011242585907146428,
+      "loss": 2.0369,
+      "step": 5107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011236925255970162,
+      "loss": 2.3216,
+      "step": 5108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011231265176641465,
+      "loss": 2.0149,
+      "step": 5109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011225605670020442,
+      "loss": 2.3825,
+      "step": 5110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011219946736967145,
+      "loss": 2.8086,
+      "step": 5111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011214288378341506,
+      "loss": 2.7175,
+      "step": 5112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011208630595003397,
+      "loss": 2.4073,
+      "step": 5113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011202973387812584,
+      "loss": 2.241,
+      "step": 5114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011197316757628754,
+      "loss": 2.6681,
+      "step": 5115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011191660705311504,
+      "loss": 2.0296,
+      "step": 5116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001118600523172035,
+      "loss": 2.5497,
+      "step": 5117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011180350337714701,
+      "loss": 2.68,
+      "step": 5118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011174696024153906,
+      "loss": 2.8439,
+      "step": 5119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011169042291897196,
+      "loss": 2.5893,
+      "step": 5120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011163389141803741,
+      "loss": 2.5056,
+      "step": 5121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011157736574732599,
+      "loss": 2.9727,
+      "step": 5122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001115208459154276,
+      "loss": 2.7223,
+      "step": 5123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011146433193093107,
+      "loss": 2.046,
+      "step": 5124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001114078238024245,
+      "loss": 2.242,
+      "step": 5125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011135132153849495,
+      "loss": 2.3345,
+      "step": 5126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011129482514772876,
+      "loss": 3.0234,
+      "step": 5127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011123833463871119,
+      "loss": 3.3311,
+      "step": 5128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011118185002002674,
+      "loss": 2.1037,
+      "step": 5129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011112537130025898,
+      "loss": 2.5139,
+      "step": 5130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001110688984879906,
+      "loss": 2.0107,
+      "step": 5131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011101243159180326,
+      "loss": 2.2035,
+      "step": 5132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000110955970620278,
+      "loss": 2.8904,
+      "step": 5133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001108995155819946,
+      "loss": 2.6534,
+      "step": 5134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011084306648553231,
+      "loss": 2.2577,
+      "step": 5135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011078662333946915,
+      "loss": 2.3188,
+      "step": 5136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001107301861523825,
+      "loss": 2.6906,
+      "step": 5137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011067375493284858,
+      "loss": 2.782,
+      "step": 5138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011061732968944298,
+      "loss": 2.424,
+      "step": 5139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011056091043074015,
+      "loss": 2.5913,
+      "step": 5140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011050449716531378,
+      "loss": 2.6322,
+      "step": 5141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011044808990173656,
+      "loss": 2.6092,
+      "step": 5142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011039168864858029,
+      "loss": 2.9813,
+      "step": 5143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011033529341441586,
+      "loss": 2.6487,
+      "step": 5144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001102789042078133,
+      "loss": 2.1569,
+      "step": 5145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011022252103734157,
+      "loss": 2.7862,
+      "step": 5146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011016614391156897,
+      "loss": 2.6147,
+      "step": 5147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00011010977283906263,
+      "loss": 2.5638,
+      "step": 5148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001100534078283888,
+      "loss": 1.4099,
+      "step": 5149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010999704888811304,
+      "loss": 2.1164,
+      "step": 5150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010994069602679961,
+      "loss": 3.3021,
+      "step": 5151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010988434925301225,
+      "loss": 2.4549,
+      "step": 5152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010982800857531341,
+      "loss": 2.8005,
+      "step": 5153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010977167400226493,
+      "loss": 1.6996,
+      "step": 5154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010971534554242742,
+      "loss": 2.8057,
+      "step": 5155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010965902320436087,
+      "loss": 2.2759,
+      "step": 5156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000109602706996624,
+      "loss": 2.5376,
+      "step": 5157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010954639692777499,
+      "loss": 2.5473,
+      "step": 5158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010949009300637072,
+      "loss": 1.9753,
+      "step": 5159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010943379524096736,
+      "loss": 2.5673,
+      "step": 5160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010937750364012005,
+      "loss": 2.74,
+      "step": 5161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010932121821238308,
+      "loss": 2.6356,
+      "step": 5162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010926493896630963,
+      "loss": 3.0268,
+      "step": 5163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010920866591045219,
+      "loss": 2.6515,
+      "step": 5164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010915239905336207,
+      "loss": 2.4126,
+      "step": 5165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010909613840358983,
+      "loss": 2.9469,
+      "step": 5166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010903988396968488,
+      "loss": 2.5039,
+      "step": 5167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010898363576019594,
+      "loss": 3.1524,
+      "step": 5168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010892739378367052,
+      "loss": 1.3537,
+      "step": 5169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010887115804865544,
+      "loss": 2.5044,
+      "step": 5170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010881492856369633,
+      "loss": 2.7367,
+      "step": 5171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010875870533733806,
+      "loss": 3.0923,
+      "step": 5172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001087024883781244,
+      "loss": 2.6847,
+      "step": 5173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010864627769459834,
+      "loss": 2.5842,
+      "step": 5174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010859007329530171,
+      "loss": 2.0794,
+      "step": 5175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010853387518877556,
+      "loss": 1.6814,
+      "step": 5176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010847768338355989,
+      "loss": 2.1261,
+      "step": 5177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010842149788819379,
+      "loss": 2.2218,
+      "step": 5178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010836531871121527,
+      "loss": 2.8779,
+      "step": 5179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010830914586116163,
+      "loss": 2.03,
+      "step": 5180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010825297934656891,
+      "loss": 2.3351,
+      "step": 5181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010819681917597249,
+      "loss": 2.7229,
+      "step": 5182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010814066535790646,
+      "loss": 2.2706,
+      "step": 5183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010808451790090427,
+      "loss": 2.0971,
+      "step": 5184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001080283768134981,
+      "loss": 1.9266,
+      "step": 5185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010797224210421946,
+      "loss": 3.0599,
+      "step": 5186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010791611378159859,
+      "loss": 2.6102,
+      "step": 5187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010785999185416506,
+      "loss": 2.7259,
+      "step": 5188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001078038763304472,
+      "loss": 2.9705,
+      "step": 5189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010774776721897255,
+      "loss": 2.3268,
+      "step": 5190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010769166452826756,
+      "loss": 3.5644,
+      "step": 5191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010763556826685783,
+      "loss": 3.0267,
+      "step": 5192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001075794784432678,
+      "loss": 1.9135,
+      "step": 5193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010752339506602115,
+      "loss": 2.4249,
+      "step": 5194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010746731814364036,
+      "loss": 1.473,
+      "step": 5195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010741124768464717,
+      "loss": 2.3947,
+      "step": 5196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010735518369756206,
+      "loss": 2.5851,
+      "step": 5197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010729912619090482,
+      "loss": 2.4329,
+      "step": 5198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000107243075173194,
+      "loss": 2.6545,
+      "step": 5199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010718703065294733,
+      "loss": 2.7099,
+      "step": 5200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010713099263868143,
+      "loss": 2.4411,
+      "step": 5201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010707496113891211,
+      "loss": 2.307,
+      "step": 5202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010701893616215394,
+      "loss": 2.4731,
+      "step": 5203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010696291771692074,
+      "loss": 2.6778,
+      "step": 5204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001069069058117252,
+      "loss": 3.2355,
+      "step": 5205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010685090045507905,
+      "loss": 2.4339,
+      "step": 5206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010679490165549297,
+      "loss": 2.895,
+      "step": 5207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001067389094214768,
+      "loss": 1.2379,
+      "step": 5208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010668292376153918,
+      "loss": 2.5195,
+      "step": 5209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010662694468418792,
+      "loss": 2.2961,
+      "step": 5210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010657097219792971,
+      "loss": 1.9494,
+      "step": 5211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010651500631127035,
+      "loss": 2.5258,
+      "step": 5212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010645904703271448,
+      "loss": 2.3522,
+      "step": 5213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010640309437076595,
+      "loss": 2.4202,
+      "step": 5214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010634714833392735,
+      "loss": 2.76,
+      "step": 5215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010629120893070057,
+      "loss": 2.2428,
+      "step": 5216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010623527616958618,
+      "loss": 2.0191,
+      "step": 5217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010617935005908393,
+      "loss": 2.8932,
+      "step": 5218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010612343060769252,
+      "loss": 2.5467,
+      "step": 5219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010606751782390963,
+      "loss": 2.0735,
+      "step": 5220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010601161171623188,
+      "loss": 3.0538,
+      "step": 5221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010595571229315497,
+      "loss": 2.8598,
+      "step": 5222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010589981956317355,
+      "loss": 2.4002,
+      "step": 5223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010584393353478116,
+      "loss": 2.77,
+      "step": 5224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010578805421647048,
+      "loss": 2.2401,
+      "step": 5225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010573218161673302,
+      "loss": 2.1986,
+      "step": 5226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010567631574405943,
+      "loss": 1.7051,
+      "step": 5227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010562045660693914,
+      "loss": 2.4108,
+      "step": 5228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001055646042138608,
+      "loss": 2.5438,
+      "step": 5229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010550875857331173,
+      "loss": 2.1075,
+      "step": 5230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010545291969377855,
+      "loss": 1.5003,
+      "step": 5231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010539708758374658,
+      "loss": 2.1575,
+      "step": 5232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001053412622517003,
+      "loss": 2.7778,
+      "step": 5233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010528544370612305,
+      "loss": 2.3141,
+      "step": 5234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010522963195549721,
+      "loss": 2.1472,
+      "step": 5235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010517382700830406,
+      "loss": 2.261,
+      "step": 5236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010511802887302386,
+      "loss": 2.7908,
+      "step": 5237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001050622375581359,
+      "loss": 2.7596,
+      "step": 5238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001050064530721184,
+      "loss": 2.2724,
+      "step": 5239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010495067542344843,
+      "loss": 2.463,
+      "step": 5240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010489490462060225,
+      "loss": 2.428,
+      "step": 5241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010483914067205483,
+      "loss": 1.9542,
+      "step": 5242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010478338358628034,
+      "loss": 2.349,
+      "step": 5243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010472763337175164,
+      "loss": 2.1796,
+      "step": 5244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010467189003694084,
+      "loss": 2.342,
+      "step": 5245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010461615359031873,
+      "loss": 2.5066,
+      "step": 5246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010456042404035533,
+      "loss": 1.1145,
+      "step": 5247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010450470139551926,
+      "loss": 2.2394,
+      "step": 5248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001044489856642785,
+      "loss": 2.6945,
+      "step": 5249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010439327685509961,
+      "loss": 2.265,
+      "step": 5250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010433757497644836,
+      "loss": 2.8004,
+      "step": 5251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010428188003678932,
+      "loss": 1.956,
+      "step": 5252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010422619204458607,
+      "loss": 2.7971,
+      "step": 5253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010417051100830105,
+      "loss": 2.4835,
+      "step": 5254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010411483693639582,
+      "loss": 2.9543,
+      "step": 5255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010405916983733068,
+      "loss": 2.3953,
+      "step": 5256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010400350971956506,
+      "loss": 2.7156,
+      "step": 5257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010394785659155708,
+      "loss": 2.661,
+      "step": 5258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001038922104617641,
+      "loss": 2.1713,
+      "step": 5259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010383657133864215,
+      "loss": 2.36,
+      "step": 5260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010378093923064641,
+      "loss": 2.4151,
+      "step": 5261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010372531414623078,
+      "loss": 2.6048,
+      "step": 5262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010366969609384835,
+      "loss": 2.407,
+      "step": 5263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010361408508195089,
+      "loss": 2.1025,
+      "step": 5264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001035584811189892,
+      "loss": 2.5618,
+      "step": 5265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010350288421341305,
+      "loss": 2.4869,
+      "step": 5266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010344729437367118,
+      "loss": 2.3364,
+      "step": 5267
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000103391711608211,
+      "loss": 2.3199,
+      "step": 5268
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010333613592547921,
+      "loss": 2.6769,
+      "step": 5269
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010328056733392112,
+      "loss": 1.9652,
+      "step": 5270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010322500584198118,
+      "loss": 2.7464,
+      "step": 5271
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010316945145810258,
+      "loss": 2.423,
+      "step": 5272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010311390419072765,
+      "loss": 2.8342,
+      "step": 5273
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010305836404829736,
+      "loss": 2.4649,
+      "step": 5274
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001030028310392519,
+      "loss": 2.8042,
+      "step": 5275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001029473051720301,
+      "loss": 2.4474,
+      "step": 5276
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010289178645506992,
+      "loss": 2.4275,
+      "step": 5277
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010283627489680808,
+      "loss": 2.3127,
+      "step": 5278
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010278077050568028,
+      "loss": 2.2776,
+      "step": 5279
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010272527329012117,
+      "loss": 2.3671,
+      "step": 5280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010266978325856424,
+      "loss": 2.4924,
+      "step": 5281
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010261430041944188,
+      "loss": 1.2278,
+      "step": 5282
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010255882478118549,
+      "loss": 2.9779,
+      "step": 5283
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010250335635222521,
+      "loss": 2.6205,
+      "step": 5284
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001024478951409903,
+      "loss": 2.1273,
+      "step": 5285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010239244115590868,
+      "loss": 2.627,
+      "step": 5286
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001023369944054074,
+      "loss": 2.4225,
+      "step": 5287
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010228155489791223,
+      "loss": 2.1879,
+      "step": 5288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000102226122641848,
+      "loss": 2.6303,
+      "step": 5289
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010217069764563822,
+      "loss": 2.523,
+      "step": 5290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001021152799177056,
+      "loss": 2.4475,
+      "step": 5291
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010205986946647146,
+      "loss": 3.069,
+      "step": 5292
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010200446630035614,
+      "loss": 2.7284,
+      "step": 5293
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010194907042777892,
+      "loss": 2.6374,
+      "step": 5294
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010189368185715783,
+      "loss": 2.6153,
+      "step": 5295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010183830059690994,
+      "loss": 3.1533,
+      "step": 5296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010178292665545113,
+      "loss": 2.8611,
+      "step": 5297
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010172756004119614,
+      "loss": 2.1105,
+      "step": 5298
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001016722007625587,
+      "loss": 2.068,
+      "step": 5299
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010161684882795138,
+      "loss": 2.3415,
+      "step": 5300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010156150424578549,
+      "loss": 2.5776,
+      "step": 5301
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010150616702447152,
+      "loss": 2.6175,
+      "step": 5302
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010145083717241853,
+      "loss": 2.9082,
+      "step": 5303
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010139551469803471,
+      "loss": 2.1198,
+      "step": 5304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001013401996097269,
+      "loss": 0.7539,
+      "step": 5305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001012848919159011,
+      "loss": 2.554,
+      "step": 5306
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010122959162496185,
+      "loss": 2.564,
+      "step": 5307
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010117429874531289,
+      "loss": 2.3342,
+      "step": 5308
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010111901328535656,
+      "loss": 2.3013,
+      "step": 5309
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010106373525349431,
+      "loss": 1.1216,
+      "step": 5310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010100846465812626,
+      "loss": 2.7105,
+      "step": 5311
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010095320150765151,
+      "loss": 2.5385,
+      "step": 5312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010089794581046805,
+      "loss": 2.6519,
+      "step": 5313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010084269757497266,
+      "loss": 2.4678,
+      "step": 5314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010078745680956097,
+      "loss": 2.101,
+      "step": 5315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010073222352262761,
+      "loss": 2.3122,
+      "step": 5316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001006769977225659,
+      "loss": 2.3521,
+      "step": 5317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010062177941776824,
+      "loss": 2.6983,
+      "step": 5318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001005665686166256,
+      "loss": 2.3718,
+      "step": 5319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001005113653275281,
+      "loss": 2.8423,
+      "step": 5320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001004561695588645,
+      "loss": 2.585,
+      "step": 5321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010040098131902258,
+      "loss": 2.363,
+      "step": 5322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010034580061638882,
+      "loss": 2.0498,
+      "step": 5323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010029062745934875,
+      "loss": 3.0001,
+      "step": 5324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010023546185628656,
+      "loss": 2.289,
+      "step": 5325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010018030381558535,
+      "loss": 2.1051,
+      "step": 5326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010012515334562715,
+      "loss": 2.7732,
+      "step": 5327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010007001045479279,
+      "loss": 2.5449,
+      "step": 5328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00010001487515146186,
+      "loss": 2.1395,
+      "step": 5329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.995974744401298e-05,
+      "loss": 2.5708,
+      "step": 5330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990462734082342e-05,
+      "loss": 2.6489,
+      "step": 5331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.984951485026949e-05,
+      "loss": 2.8695,
+      "step": 5332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.979440998072613e-05,
+      "loss": 2.714,
+      "step": 5333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.97393127405673e-05,
+      "loss": 2.7354,
+      "step": 5334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.96842231381657e-05,
+      "loss": 2.8619,
+      "step": 5335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.962914118189298e-05,
+      "loss": 2.545,
+      "step": 5336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.957406688011944e-05,
+      "loss": 2.3152,
+      "step": 5337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.951900024121443e-05,
+      "loss": 2.4324,
+      "step": 5338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946394127354596e-05,
+      "loss": 2.8913,
+      "step": 5339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940888998548096e-05,
+      "loss": 2.7756,
+      "step": 5340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935384638538521e-05,
+      "loss": 2.6851,
+      "step": 5341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929881048162331e-05,
+      "loss": 3.0725,
+      "step": 5342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924378228255857e-05,
+      "loss": 1.6269,
+      "step": 5343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.918876179655336e-05,
+      "loss": 2.9371,
+      "step": 5344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.913374903196864e-05,
+      "loss": 2.1187,
+      "step": 5345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.90787439971644e-05,
+      "loss": 2.9611,
+      "step": 5346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.902374670049926e-05,
+      "loss": 2.2799,
+      "step": 5347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.89687571503309e-05,
+      "loss": 2.7128,
+      "step": 5348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.891377535501554e-05,
+      "loss": 2.2749,
+      "step": 5349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.885880132290851e-05,
+      "loss": 2.4852,
+      "step": 5350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.88038350623637e-05,
+      "loss": 2.7893,
+      "step": 5351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.874887658173405e-05,
+      "loss": 2.2698,
+      "step": 5352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.869392588937111e-05,
+      "loss": 2.5027,
+      "step": 5353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.863898299362541e-05,
+      "loss": 2.9209,
+      "step": 5354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858404790284618e-05,
+      "loss": 2.7444,
+      "step": 5355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852912062538157e-05,
+      "loss": 1.2217,
+      "step": 5356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84742011695784e-05,
+      "loss": 2.7536,
+      "step": 5357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841928954378249e-05,
+      "loss": 2.7263,
+      "step": 5358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836438575633824e-05,
+      "loss": 2.4003,
+      "step": 5359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830948981558917e-05,
+      "loss": 2.5496,
+      "step": 5360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.825460172987723e-05,
+      "loss": 2.4095,
+      "step": 5361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.81997215075435e-05,
+      "loss": 2.8696,
+      "step": 5362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.814484915692768e-05,
+      "loss": 2.555,
+      "step": 5363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.808998468636835e-05,
+      "loss": 2.2289,
+      "step": 5364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.803512810420283e-05,
+      "loss": 2.1001,
+      "step": 5365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.798027941876735e-05,
+      "loss": 2.3799,
+      "step": 5366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.792543863839685e-05,
+      "loss": 2.9873,
+      "step": 5367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.787060577142503e-05,
+      "loss": 2.3412,
+      "step": 5368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.781578082618452e-05,
+      "loss": 2.6128,
+      "step": 5369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.776096381100662e-05,
+      "loss": 2.517,
+      "step": 5370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.770615473422156e-05,
+      "loss": 2.7528,
+      "step": 5371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.765135360415822e-05,
+      "loss": 2.6414,
+      "step": 5372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.759656042914433e-05,
+      "loss": 2.3992,
+      "step": 5373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754177521750645e-05,
+      "loss": 2.8493,
+      "step": 5374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748699797756991e-05,
+      "loss": 2.4396,
+      "step": 5375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743222871765873e-05,
+      "loss": 2.3084,
+      "step": 5376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737746744609592e-05,
+      "loss": 1.1874,
+      "step": 5377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732271417120305e-05,
+      "loss": 2.4266,
+      "step": 5378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.726796890130068e-05,
+      "loss": 2.761,
+      "step": 5379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.721323164470796e-05,
+      "loss": 3.0234,
+      "step": 5380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.715850240974305e-05,
+      "loss": 2.9947,
+      "step": 5381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.710378120472258e-05,
+      "loss": 2.108,
+      "step": 5382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.704906803796233e-05,
+      "loss": 2.6046,
+      "step": 5383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.699436291777652e-05,
+      "loss": 2.7193,
+      "step": 5384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.693966585247842e-05,
+      "loss": 2.0842,
+      "step": 5385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.688497685037984e-05,
+      "loss": 2.9295,
+      "step": 5386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.683029591979154e-05,
+      "loss": 2.3342,
+      "step": 5387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.677562306902297e-05,
+      "loss": 2.7154,
+      "step": 5388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.672095830638242e-05,
+      "loss": 0.8908,
+      "step": 5389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.666630164017678e-05,
+      "loss": 2.6581,
+      "step": 5390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.661165307871199e-05,
+      "loss": 2.305,
+      "step": 5391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.655701263029246e-05,
+      "loss": 2.4114,
+      "step": 5392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.650238030322163e-05,
+      "loss": 2.5111,
+      "step": 5393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.644775610580147e-05,
+      "loss": 2.2862,
+      "step": 5394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.639314004633295e-05,
+      "loss": 2.5329,
+      "step": 5395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.633853213311554e-05,
+      "loss": 2.8815,
+      "step": 5396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.628393237444776e-05,
+      "loss": 2.8606,
+      "step": 5397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.622934077862663e-05,
+      "loss": 2.3291,
+      "step": 5398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.617475735394814e-05,
+      "loss": 2.1755,
+      "step": 5399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.612018210870687e-05,
+      "loss": 2.2519,
+      "step": 5400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.606561505119625e-05,
+      "loss": 2.2717,
+      "step": 5401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.601105618970843e-05,
+      "loss": 2.8761,
+      "step": 5402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.59565055325344e-05,
+      "loss": 2.2281,
+      "step": 5403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.590196308796372e-05,
+      "loss": 2.2993,
+      "step": 5404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.584742886428494e-05,
+      "loss": 1.7321,
+      "step": 5405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.579290286978511e-05,
+      "loss": 2.3328,
+      "step": 5406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.573838511275028e-05,
+      "loss": 2.3462,
+      "step": 5407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.5683875601465e-05,
+      "loss": 2.2613,
+      "step": 5408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.562937434421282e-05,
+      "loss": 2.3603,
+      "step": 5409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.557488134927578e-05,
+      "loss": 2.8073,
+      "step": 5410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.552039662493492e-05,
+      "loss": 2.243,
+      "step": 5411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.546592017946974e-05,
+      "loss": 2.7004,
+      "step": 5412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.541145202115878e-05,
+      "loss": 1.7778,
+      "step": 5413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.535699215827912e-05,
+      "loss": 2.9313,
+      "step": 5414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.530254059910661e-05,
+      "loss": 2.2014,
+      "step": 5415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.524809735191587e-05,
+      "loss": 2.1134,
+      "step": 5416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.519366242498034e-05,
+      "loss": 2.3757,
+      "step": 5417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.513923582657193e-05,
+      "loss": 1.1845,
+      "step": 5418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.508481756496167e-05,
+      "loss": 2.6073,
+      "step": 5419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.503040764841892e-05,
+      "loss": 2.2933,
+      "step": 5420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.497600608521213e-05,
+      "loss": 2.1417,
+      "step": 5421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.492161288360817e-05,
+      "loss": 2.5328,
+      "step": 5422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.486722805187293e-05,
+      "loss": 2.7617,
+      "step": 5423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.481285159827076e-05,
+      "loss": 2.7396,
+      "step": 5424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.475848353106495e-05,
+      "loss": 2.5748,
+      "step": 5425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.470412385851733e-05,
+      "loss": 2.4439,
+      "step": 5426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.464977258888868e-05,
+      "loss": 2.7403,
+      "step": 5427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.459542973043827e-05,
+      "loss": 2.6357,
+      "step": 5428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.454109529142425e-05,
+      "loss": 3.0699,
+      "step": 5429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.448676928010341e-05,
+      "loss": 2.7885,
+      "step": 5430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.443245170473127e-05,
+      "loss": 2.9883,
+      "step": 5431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.437814257356211e-05,
+      "loss": 2.3858,
+      "step": 5432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.432384189484893e-05,
+      "loss": 2.4845,
+      "step": 5433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.426954967684334e-05,
+      "loss": 3.0241,
+      "step": 5434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.421526592779581e-05,
+      "loss": 2.3781,
+      "step": 5435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.416099065595539e-05,
+      "loss": 2.6184,
+      "step": 5436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.410672386956998e-05,
+      "loss": 2.7942,
+      "step": 5437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.405246557688609e-05,
+      "loss": 1.9431,
+      "step": 5438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.399821578614887e-05,
+      "loss": 2.7596,
+      "step": 5439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.394397450560243e-05,
+      "loss": 2.1696,
+      "step": 5440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.388974174348927e-05,
+      "loss": 2.7479,
+      "step": 5441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.383551750805092e-05,
+      "loss": 2.1335,
+      "step": 5442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.37813018075273e-05,
+      "loss": 2.2106,
+      "step": 5443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.37270946501573e-05,
+      "loss": 2.4232,
+      "step": 5444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.36728960441783e-05,
+      "loss": 2.6553,
+      "step": 5445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.361870599782656e-05,
+      "loss": 2.0757,
+      "step": 5446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.356452451933689e-05,
+      "loss": 2.3456,
+      "step": 5447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.351035161694288e-05,
+      "loss": 2.577,
+      "step": 5448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.34561872988768e-05,
+      "loss": 2.0909,
+      "step": 5449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.340203157336966e-05,
+      "loss": 2.0144,
+      "step": 5450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.3347884448651e-05,
+      "loss": 2.66,
+      "step": 5451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.329374593294931e-05,
+      "loss": 2.2389,
+      "step": 5452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.323961603449149e-05,
+      "loss": 2.6282,
+      "step": 5453
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.318549476150344e-05,
+      "loss": 3.0383,
+      "step": 5454
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.313138212220943e-05,
+      "loss": 2.4697,
+      "step": 5455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.307727812483267e-05,
+      "loss": 2.3562,
+      "step": 5456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.302318277759488e-05,
+      "loss": 3.0295,
+      "step": 5457
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.296909608871667e-05,
+      "loss": 2.7073,
+      "step": 5458
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.291501806641704e-05,
+      "loss": 2.2944,
+      "step": 5459
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.286094871891399e-05,
+      "loss": 2.3405,
+      "step": 5460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.280688805442395e-05,
+      "loss": 2.386,
+      "step": 5461
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.275283608116218e-05,
+      "loss": 2.0649,
+      "step": 5462
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.269879280734255e-05,
+      "loss": 1.6098,
+      "step": 5463
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.264475824117768e-05,
+      "loss": 2.8694,
+      "step": 5464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.259073239087872e-05,
+      "loss": 2.5881,
+      "step": 5465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.25367152646557e-05,
+      "loss": 1.9668,
+      "step": 5466
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.248270687071711e-05,
+      "loss": 2.1339,
+      "step": 5467
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.242870721727033e-05,
+      "loss": 2.5242,
+      "step": 5468
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.237471631252119e-05,
+      "loss": 2.0761,
+      "step": 5469
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.232073416467442e-05,
+      "loss": 2.0332,
+      "step": 5470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.226676078193318e-05,
+      "loss": 2.6701,
+      "step": 5471
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.221279617249954e-05,
+      "loss": 2.1989,
+      "step": 5472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.215884034457398e-05,
+      "loss": 2.4987,
+      "step": 5473
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.210489330635591e-05,
+      "loss": 2.3074,
+      "step": 5474
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.205095506604323e-05,
+      "loss": 2.6246,
+      "step": 5475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.199702563183252e-05,
+      "loss": 2.7483,
+      "step": 5476
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.194310501191907e-05,
+      "loss": 2.4574,
+      "step": 5477
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.188919321449683e-05,
+      "loss": 1.7857,
+      "step": 5478
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.183529024775833e-05,
+      "loss": 3.3667,
+      "step": 5479
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.17813961198949e-05,
+      "loss": 2.3918,
+      "step": 5480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.172751083909636e-05,
+      "loss": 2.0555,
+      "step": 5481
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.167363441355138e-05,
+      "loss": 2.8892,
+      "step": 5482
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.161976685144704e-05,
+      "loss": 1.2904,
+      "step": 5483
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.156590816096935e-05,
+      "loss": 2.8534,
+      "step": 5484
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.151205835030271e-05,
+      "loss": 2.4922,
+      "step": 5485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.145821742763036e-05,
+      "loss": 2.0424,
+      "step": 5486
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.140438540113408e-05,
+      "loss": 2.2148,
+      "step": 5487
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.13505622789944e-05,
+      "loss": 2.3392,
+      "step": 5488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.129674806939037e-05,
+      "loss": 2.0989,
+      "step": 5489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.124294278049978e-05,
+      "loss": 2.781,
+      "step": 5490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.118914642049901e-05,
+      "loss": 2.2678,
+      "step": 5491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.113535899756314e-05,
+      "loss": 1.4303,
+      "step": 5492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.108158051986584e-05,
+      "loss": 2.173,
+      "step": 5493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.10278109955795e-05,
+      "loss": 2.5832,
+      "step": 5494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.097405043287497e-05,
+      "loss": 1.6012,
+      "step": 5495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.092029883992197e-05,
+      "loss": 2.5194,
+      "step": 5496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.086655622488865e-05,
+      "loss": 2.8741,
+      "step": 5497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.081282259594199e-05,
+      "loss": 2.2948,
+      "step": 5498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.075909796124743e-05,
+      "loss": 2.2733,
+      "step": 5499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.070538232896918e-05,
+      "loss": 2.4077,
+      "step": 5500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.065167570726995e-05,
+      "loss": 1.8673,
+      "step": 5501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.059797810431124e-05,
+      "loss": 2.1204,
+      "step": 5502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.054428952825294e-05,
+      "loss": 2.5349,
+      "step": 5503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.049060998725393e-05,
+      "loss": 2.5119,
+      "step": 5504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.043693948947135e-05,
+      "loss": 2.4034,
+      "step": 5505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.038327804306115e-05,
+      "loss": 2.3356,
+      "step": 5506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.032962565617789e-05,
+      "loss": 2.5875,
+      "step": 5507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.027598233697481e-05,
+      "loss": 2.4054,
+      "step": 5508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.022234809360353e-05,
+      "loss": 2.2554,
+      "step": 5509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.016872293421467e-05,
+      "loss": 2.0631,
+      "step": 5510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.011510686695711e-05,
+      "loss": 2.2293,
+      "step": 5511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.006149989997851e-05,
+      "loss": 2.6067,
+      "step": 5512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.000790204142525e-05,
+      "loss": 2.1005,
+      "step": 5513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.995431329944207e-05,
+      "loss": 2.9293,
+      "step": 5514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.99007336821726e-05,
+      "loss": 2.2631,
+      "step": 5515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.984716319775881e-05,
+      "loss": 2.66,
+      "step": 5516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.979360185434158e-05,
+      "loss": 2.6054,
+      "step": 5517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.97400496600601e-05,
+      "loss": 2.8558,
+      "step": 5518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.968650662305244e-05,
+      "loss": 1.6769,
+      "step": 5519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.963297275145503e-05,
+      "loss": 2.4306,
+      "step": 5520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.957944805340315e-05,
+      "loss": 2.0848,
+      "step": 5521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.952593253703049e-05,
+      "loss": 2.3117,
+      "step": 5522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.947242621046942e-05,
+      "loss": 2.3472,
+      "step": 5523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.941892908185094e-05,
+      "loss": 2.104,
+      "step": 5524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.936544115930467e-05,
+      "loss": 2.3127,
+      "step": 5525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.931196245095865e-05,
+      "loss": 2.2804,
+      "step": 5526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.925849296493982e-05,
+      "loss": 2.3195,
+      "step": 5527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.92050327093734e-05,
+      "loss": 2.3159,
+      "step": 5528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.91515816923835e-05,
+      "loss": 2.24,
+      "step": 5529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.909813992209258e-05,
+      "loss": 1.8084,
+      "step": 5530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.904470740662192e-05,
+      "loss": 2.3589,
+      "step": 5531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.899128415409114e-05,
+      "loss": 1.423,
+      "step": 5532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.893787017261871e-05,
+      "loss": 3.0858,
+      "step": 5533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.888446547032148e-05,
+      "loss": 2.4214,
+      "step": 5534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.883107005531507e-05,
+      "loss": 2.1397,
+      "step": 5535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.877768393571352e-05,
+      "loss": 2.4364,
+      "step": 5536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.872430711962957e-05,
+      "loss": 2.7268,
+      "step": 5537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.867093961517452e-05,
+      "loss": 1.7044,
+      "step": 5538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.861758143045827e-05,
+      "loss": 2.4541,
+      "step": 5539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.856423257358917e-05,
+      "loss": 1.1215,
+      "step": 5540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.851089305267442e-05,
+      "loss": 2.5902,
+      "step": 5541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.84575628758195e-05,
+      "loss": 2.7047,
+      "step": 5542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.840424205112878e-05,
+      "loss": 1.9788,
+      "step": 5543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.835093058670486e-05,
+      "loss": 2.4646,
+      "step": 5544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.829762849064927e-05,
+      "loss": 2.4711,
+      "step": 5545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.824433577106178e-05,
+      "loss": 2.5942,
+      "step": 5546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.819105243604109e-05,
+      "loss": 2.4652,
+      "step": 5547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.813777849368411e-05,
+      "loss": 2.5345,
+      "step": 5548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.808451395208664e-05,
+      "loss": 2.4313,
+      "step": 5549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.803125881934284e-05,
+      "loss": 2.4845,
+      "step": 5550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.797801310354551e-05,
+      "loss": 2.67,
+      "step": 5551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.792477681278606e-05,
+      "loss": 2.5929,
+      "step": 5552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.78715499551544e-05,
+      "loss": 2.3706,
+      "step": 5553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.781833253873901e-05,
+      "loss": 2.8011,
+      "step": 5554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.776512457162703e-05,
+      "loss": 2.4474,
+      "step": 5555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.771192606190401e-05,
+      "loss": 2.1208,
+      "step": 5556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.765873701765424e-05,
+      "loss": 2.6053,
+      "step": 5557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.760555744696036e-05,
+      "loss": 2.5988,
+      "step": 5558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.755238735790384e-05,
+      "loss": 2.575,
+      "step": 5559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.749922675856442e-05,
+      "loss": 2.4842,
+      "step": 5560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.744607565702064e-05,
+      "loss": 1.8161,
+      "step": 5561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.739293406134938e-05,
+      "loss": 2.1848,
+      "step": 5562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.733980197962633e-05,
+      "loss": 2.4131,
+      "step": 5563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.728667941992545e-05,
+      "loss": 2.5562,
+      "step": 5564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.723356639031953e-05,
+      "loss": 2.4932,
+      "step": 5565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.718046289887964e-05,
+      "loss": 2.4942,
+      "step": 5566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.712736895367566e-05,
+      "loss": 2.0986,
+      "step": 5567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.70742845627758e-05,
+      "loss": 2.5027,
+      "step": 5568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.702120973424701e-05,
+      "loss": 3.0959,
+      "step": 5569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.69681444761546e-05,
+      "loss": 2.1768,
+      "step": 5570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.691508879656261e-05,
+      "loss": 2.2877,
+      "step": 5571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.686204270353347e-05,
+      "loss": 2.2589,
+      "step": 5572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.68090062051282e-05,
+      "loss": 2.5862,
+      "step": 5573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.675597930940643e-05,
+      "loss": 2.472,
+      "step": 5574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.670296202442626e-05,
+      "loss": 2.9789,
+      "step": 5575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.66499543582443e-05,
+      "loss": 2.6977,
+      "step": 5576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.659695631891584e-05,
+      "loss": 2.6347,
+      "step": 5577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.65439679144945e-05,
+      "loss": 2.6469,
+      "step": 5578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.649098915303267e-05,
+      "loss": 2.4444,
+      "step": 5579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.643802004258101e-05,
+      "loss": 2.4615,
+      "step": 5580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.638506059118904e-05,
+      "loss": 2.4677,
+      "step": 5581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.633211080690444e-05,
+      "loss": 2.1521,
+      "step": 5582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.627917069777377e-05,
+      "loss": 1.9311,
+      "step": 5583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.622624027184189e-05,
+      "loss": 2.3776,
+      "step": 5584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.617331953715218e-05,
+      "loss": 2.6061,
+      "step": 5585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.61204085017468e-05,
+      "loss": 2.3025,
+      "step": 5586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.606750717366606e-05,
+      "loss": 2.6033,
+      "step": 5587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.601461556094918e-05,
+      "loss": 2.5697,
+      "step": 5588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.596173367163359e-05,
+      "loss": 2.873,
+      "step": 5589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.590886151375548e-05,
+      "loss": 2.5436,
+      "step": 5590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.585599909534934e-05,
+      "loss": 2.5759,
+      "step": 5591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.580314642444839e-05,
+      "loss": 2.6183,
+      "step": 5592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.575030350908421e-05,
+      "loss": 2.7577,
+      "step": 5593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.569747035728702e-05,
+      "loss": 2.6223,
+      "step": 5594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.564464697708539e-05,
+      "loss": 2.4711,
+      "step": 5595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.559183337650667e-05,
+      "loss": 1.7964,
+      "step": 5596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.55390295635764e-05,
+      "loss": 2.3852,
+      "step": 5597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.548623554631893e-05,
+      "loss": 1.0911,
+      "step": 5598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.543345133275689e-05,
+      "loss": 2.4026,
+      "step": 5599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.53806769309116e-05,
+      "loss": 2.9887,
+      "step": 5600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.532791234880274e-05,
+      "loss": 2.4648,
+      "step": 5601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.527515759444864e-05,
+      "loss": 1.4673,
+      "step": 5602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.522241267586596e-05,
+      "loss": 2.3313,
+      "step": 5603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.516967760107009e-05,
+      "loss": 2.3753,
+      "step": 5604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.511695237807475e-05,
+      "loss": 3.0295,
+      "step": 5605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.506423701489214e-05,
+      "loss": 0.0004,
+      "step": 5606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.501153151953314e-05,
+      "loss": 2.3209,
+      "step": 5607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.495883590000703e-05,
+      "loss": 2.549,
+      "step": 5608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.490615016432148e-05,
+      "loss": 2.835,
+      "step": 5609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.485347432048288e-05,
+      "loss": 2.7803,
+      "step": 5610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.480080837649588e-05,
+      "loss": 1.755,
+      "step": 5611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.474815234036391e-05,
+      "loss": 2.3665,
+      "step": 5612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.469550622008857e-05,
+      "loss": 2.692,
+      "step": 5613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.464287002367025e-05,
+      "loss": 3.0608,
+      "step": 5614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.459024375910757e-05,
+      "loss": 2.5506,
+      "step": 5615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.453762743439786e-05,
+      "loss": 2.1089,
+      "step": 5616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.448502105753679e-05,
+      "loss": 2.0644,
+      "step": 5617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.443242463651865e-05,
+      "loss": 2.7606,
+      "step": 5618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.437983817933602e-05,
+      "loss": 2.9836,
+      "step": 5619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.432726169398023e-05,
+      "loss": 2.9643,
+      "step": 5620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.427469518844084e-05,
+      "loss": 2.6452,
+      "step": 5621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.42221386707061e-05,
+      "loss": 2.3379,
+      "step": 5622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.416959214876252e-05,
+      "loss": 2.3171,
+      "step": 5623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.411705563059537e-05,
+      "loss": 2.7808,
+      "step": 5624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.406452912418812e-05,
+      "loss": 3.0109,
+      "step": 5625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.401201263752294e-05,
+      "loss": 2.3087,
+      "step": 5626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.395950617858032e-05,
+      "loss": 2.5061,
+      "step": 5627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.390700975533934e-05,
+      "loss": 2.3172,
+      "step": 5628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.385452337577746e-05,
+      "loss": 2.209,
+      "step": 5629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.380204704787073e-05,
+      "loss": 2.7856,
+      "step": 5630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.37495807795935e-05,
+      "loss": 2.1324,
+      "step": 5631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.369712457891879e-05,
+      "loss": 2.1644,
+      "step": 5632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.364467845381799e-05,
+      "loss": 1.0607,
+      "step": 5633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.359224241226086e-05,
+      "loss": 2.2772,
+      "step": 5634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.353981646221584e-05,
+      "loss": 2.0386,
+      "step": 5635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.34874006116497e-05,
+      "loss": 1.4758,
+      "step": 5636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.343499486852762e-05,
+      "loss": 2.4984,
+      "step": 5637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.338259924081343e-05,
+      "loss": 2.4946,
+      "step": 5638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.333021373646924e-05,
+      "loss": 2.6582,
+      "step": 5639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.327783836345582e-05,
+      "loss": 1.961,
+      "step": 5640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.32254731297321e-05,
+      "loss": 2.8867,
+      "step": 5641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.317311804325585e-05,
+      "loss": 2.7747,
+      "step": 5642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.312077311198292e-05,
+      "loss": 2.4603,
+      "step": 5643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.306843834386796e-05,
+      "loss": 1.563,
+      "step": 5644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.301611374686374e-05,
+      "loss": 2.4664,
+      "step": 5645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.296379932892182e-05,
+      "loss": 2.786,
+      "step": 5646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.29114950979919e-05,
+      "loss": 1.9991,
+      "step": 5647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.285920106202242e-05,
+      "loss": 2.4363,
+      "step": 5648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.280691722895997e-05,
+      "loss": 2.185,
+      "step": 5649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.275464360674993e-05,
+      "loss": 1.8819,
+      "step": 5650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.270238020333577e-05,
+      "loss": 1.8774,
+      "step": 5651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.265012702665977e-05,
+      "loss": 2.0979,
+      "step": 5652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.259788408466229e-05,
+      "loss": 3.0936,
+      "step": 5653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.254565138528247e-05,
+      "loss": 2.5986,
+      "step": 5654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.249342893645761e-05,
+      "loss": 3.0117,
+      "step": 5655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.244121674612371e-05,
+      "loss": 2.683,
+      "step": 5656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.238901482221502e-05,
+      "loss": 2.1562,
+      "step": 5657
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.233682317266422e-05,
+      "loss": 2.3749,
+      "step": 5658
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.228464180540263e-05,
+      "loss": 2.6562,
+      "step": 5659
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.223247072835977e-05,
+      "loss": 2.0183,
+      "step": 5660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.218030994946382e-05,
+      "loss": 2.3544,
+      "step": 5661
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.212815947664116e-05,
+      "loss": 2.3501,
+      "step": 5662
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.207601931781681e-05,
+      "loss": 2.2711,
+      "step": 5663
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.202388948091408e-05,
+      "loss": 2.427,
+      "step": 5664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.197176997385484e-05,
+      "loss": 2.3741,
+      "step": 5665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.191966080455929e-05,
+      "loss": 2.6004,
+      "step": 5666
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.186756198094603e-05,
+      "loss": 3.2189,
+      "step": 5667
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.181547351093222e-05,
+      "loss": 2.6936,
+      "step": 5668
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.176339540243337e-05,
+      "loss": 2.4279,
+      "step": 5669
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.17113276633633e-05,
+      "loss": 2.145,
+      "step": 5670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.165927030163452e-05,
+      "loss": 2.3737,
+      "step": 5671
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.160722332515773e-05,
+      "loss": 2.4457,
+      "step": 5672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.155518674184219e-05,
+      "loss": 2.7149,
+      "step": 5673
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.150316055959548e-05,
+      "loss": 2.7398,
+      "step": 5674
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.145114478632371e-05,
+      "loss": 2.1765,
+      "step": 5675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.139913942993124e-05,
+      "loss": 2.7125,
+      "step": 5676
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.134714449832107e-05,
+      "loss": 2.3388,
+      "step": 5677
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.12951599993944e-05,
+      "loss": 2.3129,
+      "step": 5678
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.124318594105102e-05,
+      "loss": 2.139,
+      "step": 5679
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.119122233118897e-05,
+      "loss": 2.5886,
+      "step": 5680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.113926917770492e-05,
+      "loss": 2.4224,
+      "step": 5681
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.108732648849368e-05,
+      "loss": 2.1489,
+      "step": 5682
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.103539427144871e-05,
+      "loss": 2.8693,
+      "step": 5683
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.098347253446168e-05,
+      "loss": 2.3319,
+      "step": 5684
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.093156128542289e-05,
+      "loss": 1.9428,
+      "step": 5685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.08796605322208e-05,
+      "loss": 2.6693,
+      "step": 5686
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.082777028274249e-05,
+      "loss": 2.2807,
+      "step": 5687
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.077589054487323e-05,
+      "loss": 2.3216,
+      "step": 5688
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.072402132649699e-05,
+      "loss": 1.8185,
+      "step": 5689
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.067216263549579e-05,
+      "loss": 1.0111,
+      "step": 5690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.062031447975038e-05,
+      "loss": 2.7417,
+      "step": 5691
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.056847686713959e-05,
+      "loss": 2.7778,
+      "step": 5692
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.051664980554097e-05,
+      "loss": 1.8794,
+      "step": 5693
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.046483330283023e-05,
+      "loss": 2.2318,
+      "step": 5694
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.041302736688148e-05,
+      "loss": 2.7619,
+      "step": 5695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.036123200556747e-05,
+      "loss": 1.9925,
+      "step": 5696
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.030944722675905e-05,
+      "loss": 2.3843,
+      "step": 5697
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.025767303832555e-05,
+      "loss": 2.4864,
+      "step": 5698
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.020590944813484e-05,
+      "loss": 1.0302,
+      "step": 5699
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.015415646405296e-05,
+      "loss": 2.6585,
+      "step": 5700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.010241409394452e-05,
+      "loss": 2.2506,
+      "step": 5701
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 8.005068234567236e-05,
+      "loss": 2.6513,
+      "step": 5702
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.999896122709788e-05,
+      "loss": 2.5394,
+      "step": 5703
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.994725074608066e-05,
+      "loss": 2.444,
+      "step": 5704
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.989555091047889e-05,
+      "loss": 2.5806,
+      "step": 5705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.984386172814888e-05,
+      "loss": 2.7468,
+      "step": 5706
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.979218320694562e-05,
+      "loss": 1.9394,
+      "step": 5707
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.97405153547222e-05,
+      "loss": 2.4847,
+      "step": 5708
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.968885817933033e-05,
+      "loss": 2.4205,
+      "step": 5709
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.963721168861982e-05,
+      "loss": 2.5037,
+      "step": 5710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.958557589043922e-05,
+      "loss": 2.5665,
+      "step": 5711
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.953395079263507e-05,
+      "loss": 2.3207,
+      "step": 5712
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.948233640305261e-05,
+      "loss": 2.5823,
+      "step": 5713
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.943073272953519e-05,
+      "loss": 2.7848,
+      "step": 5714
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.937913977992475e-05,
+      "loss": 2.2843,
+      "step": 5715
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.932755756206142e-05,
+      "loss": 2.883,
+      "step": 5716
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.927598608378384e-05,
+      "loss": 2.541,
+      "step": 5717
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.922442535292893e-05,
+      "loss": 2.4473,
+      "step": 5718
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.917287537733203e-05,
+      "loss": 2.2935,
+      "step": 5719
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.912133616482679e-05,
+      "loss": 2.3243,
+      "step": 5720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.906980772324529e-05,
+      "loss": 2.3073,
+      "step": 5721
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.901829006041789e-05,
+      "loss": 2.6526,
+      "step": 5722
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.896678318417341e-05,
+      "loss": 2.602,
+      "step": 5723
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.8915287102339e-05,
+      "loss": 2.5301,
+      "step": 5724
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.886380182274002e-05,
+      "loss": 2.218,
+      "step": 5725
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.88123273532005e-05,
+      "loss": 2.4015,
+      "step": 5726
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.876086370154255e-05,
+      "loss": 2.4055,
+      "step": 5727
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.870941087558668e-05,
+      "loss": 1.1905,
+      "step": 5728
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.865796888315195e-05,
+      "loss": 2.7271,
+      "step": 5729
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.860653773205552e-05,
+      "loss": 2.2991,
+      "step": 5730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.8555117430113e-05,
+      "loss": 1.8192,
+      "step": 5731
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.850370798513848e-05,
+      "loss": 2.3159,
+      "step": 5732
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.845230940494416e-05,
+      "loss": 2.3907,
+      "step": 5733
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.840092169734081e-05,
+      "loss": 2.294,
+      "step": 5734
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.834954487013739e-05,
+      "loss": 2.5648,
+      "step": 5735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.829817893114132e-05,
+      "loss": 2.6231,
+      "step": 5736
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.824682388815825e-05,
+      "loss": 2.2005,
+      "step": 5737
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.819547974899234e-05,
+      "loss": 2.2414,
+      "step": 5738
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.814414652144584e-05,
+      "loss": 2.6147,
+      "step": 5739
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.809282421331969e-05,
+      "loss": 2.6366,
+      "step": 5740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.804151283241279e-05,
+      "loss": 3.0654,
+      "step": 5741
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.799021238652268e-05,
+      "loss": 2.368,
+      "step": 5742
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.793892288344504e-05,
+      "loss": 2.0296,
+      "step": 5743
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.788764433097407e-05,
+      "loss": 2.3859,
+      "step": 5744
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.78363767369021e-05,
+      "loss": 1.9531,
+      "step": 5745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.778512010901998e-05,
+      "loss": 2.461,
+      "step": 5746
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.773387445511673e-05,
+      "loss": 2.8437,
+      "step": 5747
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.768263978297988e-05,
+      "loss": 2.7537,
+      "step": 5748
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.763141610039512e-05,
+      "loss": 2.7192,
+      "step": 5749
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.758020341514662e-05,
+      "loss": 2.6924,
+      "step": 5750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.752900173501671e-05,
+      "loss": 2.0945,
+      "step": 5751
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.747781106778626e-05,
+      "loss": 2.5963,
+      "step": 5752
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.742663142123423e-05,
+      "loss": 2.4723,
+      "step": 5753
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.737546280313815e-05,
+      "loss": 2.4076,
+      "step": 5754
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.732430522127365e-05,
+      "loss": 2.6536,
+      "step": 5755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.727315868341477e-05,
+      "loss": 2.7615,
+      "step": 5756
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.722202319733399e-05,
+      "loss": 2.4432,
+      "step": 5757
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.717089877080193e-05,
+      "loss": 1.7211,
+      "step": 5758
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.711978541158757e-05,
+      "loss": 2.3912,
+      "step": 5759
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.706868312745833e-05,
+      "loss": 2.2312,
+      "step": 5760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.701759192617977e-05,
+      "loss": 2.3754,
+      "step": 5761
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.696651181551595e-05,
+      "loss": 2.8173,
+      "step": 5762
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.691544280322904e-05,
+      "loss": 2.4287,
+      "step": 5763
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.686438489707976e-05,
+      "loss": 1.8803,
+      "step": 5764
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.681333810482689e-05,
+      "loss": 2.8771,
+      "step": 5765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.676230243422773e-05,
+      "loss": 2.3309,
+      "step": 5766
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.671127789303773e-05,
+      "loss": 2.5956,
+      "step": 5767
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.666026448901084e-05,
+      "loss": 2.9484,
+      "step": 5768
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.660926222989907e-05,
+      "loss": 2.5139,
+      "step": 5769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.6558271123453e-05,
+      "loss": 2.2771,
+      "step": 5770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.650729117742126e-05,
+      "loss": 2.3506,
+      "step": 5771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.645632239955103e-05,
+      "loss": 1.8945,
+      "step": 5772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.640536479758755e-05,
+      "loss": 2.7151,
+      "step": 5773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.635441837927457e-05,
+      "loss": 2.708,
+      "step": 5774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.630348315235396e-05,
+      "loss": 2.8712,
+      "step": 5775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.625255912456612e-05,
+      "loss": 2.6739,
+      "step": 5776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.620164630364949e-05,
+      "loss": 3.2828,
+      "step": 5777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.615074469734099e-05,
+      "loss": 1.639,
+      "step": 5778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.609985431337574e-05,
+      "loss": 2.0824,
+      "step": 5779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.604897515948722e-05,
+      "loss": 2.6473,
+      "step": 5780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.599810724340712e-05,
+      "loss": 2.4731,
+      "step": 5781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.594725057286555e-05,
+      "loss": 2.849,
+      "step": 5782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.589640515559075e-05,
+      "loss": 2.8902,
+      "step": 5783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.584557099930943e-05,
+      "loss": 2.2006,
+      "step": 5784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.579474811174642e-05,
+      "loss": 2.302,
+      "step": 5785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.57439365006249e-05,
+      "loss": 2.692,
+      "step": 5786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.569313617366641e-05,
+      "loss": 2.646,
+      "step": 5787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.564234713859069e-05,
+      "loss": 2.4789,
+      "step": 5788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.559156940311572e-05,
+      "loss": 2.4157,
+      "step": 5789
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.554080297495794e-05,
+      "loss": 2.4735,
+      "step": 5790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.549004786183183e-05,
+      "loss": 2.9277,
+      "step": 5791
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.543930407145045e-05,
+      "loss": 1.8291,
+      "step": 5792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.538857161152481e-05,
+      "loss": 3.0151,
+      "step": 5793
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.533785048976448e-05,
+      "loss": 2.8394,
+      "step": 5794
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.528714071387707e-05,
+      "loss": 2.4837,
+      "step": 5795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.523644229156871e-05,
+      "loss": 2.143,
+      "step": 5796
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.518575523054356e-05,
+      "loss": 2.8832,
+      "step": 5797
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.513507953850427e-05,
+      "loss": 2.7139,
+      "step": 5798
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.508441522315157e-05,
+      "loss": 1.0987,
+      "step": 5799
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.503376229218465e-05,
+      "loss": 2.636,
+      "step": 5800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.498312075330084e-05,
+      "loss": 1.7517,
+      "step": 5801
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.493249061419568e-05,
+      "loss": 2.2312,
+      "step": 5802
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.488187188256321e-05,
+      "loss": 2.9464,
+      "step": 5803
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.483126456609546e-05,
+      "loss": 2.4515,
+      "step": 5804
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.478066867248302e-05,
+      "loss": 2.1449,
+      "step": 5805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.473008420941441e-05,
+      "loss": 2.0099,
+      "step": 5806
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.467951118457676e-05,
+      "loss": 1.6816,
+      "step": 5807
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.462894960565514e-05,
+      "loss": 2.3262,
+      "step": 5808
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.457839948033319e-05,
+      "loss": 2.3528,
+      "step": 5809
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.452786081629247e-05,
+      "loss": 2.5447,
+      "step": 5810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.447733362121315e-05,
+      "loss": 1.9559,
+      "step": 5811
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.442681790277335e-05,
+      "loss": 2.3232,
+      "step": 5812
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.43763136686497e-05,
+      "loss": 2.0764,
+      "step": 5813
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.432582092651685e-05,
+      "loss": 2.7677,
+      "step": 5814
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.427533968404793e-05,
+      "loss": 2.1957,
+      "step": 5815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.422486994891418e-05,
+      "loss": 2.1555,
+      "step": 5816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.417441172878503e-05,
+      "loss": 1.4338,
+      "step": 5817
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.41239650313284e-05,
+      "loss": 2.3889,
+      "step": 5818
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.407352986421025e-05,
+      "loss": 2.6587,
+      "step": 5819
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.40231062350948e-05,
+      "loss": 2.5066,
+      "step": 5820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.397269415164467e-05,
+      "loss": 2.5111,
+      "step": 5821
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.39222936215205e-05,
+      "loss": 2.4913,
+      "step": 5822
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.387190465238145e-05,
+      "loss": 2.6913,
+      "step": 5823
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.382152725188462e-05,
+      "loss": 2.2032,
+      "step": 5824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.377116142768563e-05,
+      "loss": 2.5095,
+      "step": 5825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.37208071874381e-05,
+      "loss": 2.6815,
+      "step": 5826
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.367046453879413e-05,
+      "loss": 2.8674,
+      "step": 5827
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.362013348940379e-05,
+      "loss": 2.7337,
+      "step": 5828
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.356981404691566e-05,
+      "loss": 2.4403,
+      "step": 5829
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.351950621897634e-05,
+      "loss": 2.4523,
+      "step": 5830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.346921001323082e-05,
+      "loss": 2.1567,
+      "step": 5831
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.341892543732215e-05,
+      "loss": 2.4329,
+      "step": 5832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.336865249889184e-05,
+      "loss": 2.7128,
+      "step": 5833
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.331839120557941e-05,
+      "loss": 2.7735,
+      "step": 5834
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.326814156502281e-05,
+      "loss": 2.5875,
+      "step": 5835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.321790358485799e-05,
+      "loss": 2.4201,
+      "step": 5836
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.31676772727194e-05,
+      "loss": 2.518,
+      "step": 5837
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.311746263623943e-05,
+      "loss": 2.2623,
+      "step": 5838
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.306725968304897e-05,
+      "loss": 2.1452,
+      "step": 5839
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.301706842077689e-05,
+      "loss": 2.4648,
+      "step": 5840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.296688885705048e-05,
+      "loss": 2.5897,
+      "step": 5841
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.291672099949512e-05,
+      "loss": 2.2317,
+      "step": 5842
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.286656485573452e-05,
+      "loss": 2.3452,
+      "step": 5843
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.281642043339046e-05,
+      "loss": 2.8708,
+      "step": 5844
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.276628774008312e-05,
+      "loss": 2.4762,
+      "step": 5845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.271616678343079e-05,
+      "loss": 2.8585,
+      "step": 5846
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.26660575710499e-05,
+      "loss": 2.0197,
+      "step": 5847
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.261596011055532e-05,
+      "loss": 2.1485,
+      "step": 5848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.256587440955994e-05,
+      "loss": 2.4475,
+      "step": 5849
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.251580047567487e-05,
+      "loss": 2.4419,
+      "step": 5850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.246573831650962e-05,
+      "loss": 2.4507,
+      "step": 5851
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.241568793967162e-05,
+      "loss": 2.7204,
+      "step": 5852
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.236564935276682e-05,
+      "loss": 1.777,
+      "step": 5853
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.231562256339911e-05,
+      "loss": 2.0101,
+      "step": 5854
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.22656075791708e-05,
+      "loss": 2.6309,
+      "step": 5855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.22156044076822e-05,
+      "loss": 1.904,
+      "step": 5856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.216561305653206e-05,
+      "loss": 2.449,
+      "step": 5857
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.211563353331708e-05,
+      "loss": 2.6812,
+      "step": 5858
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.206566584563241e-05,
+      "loss": 2.6856,
+      "step": 5859
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.20157100010712e-05,
+      "loss": 2.8276,
+      "step": 5860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.196576600722493e-05,
+      "loss": 2.4215,
+      "step": 5861
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.191583387168315e-05,
+      "loss": 1.988,
+      "step": 5862
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.186591360203383e-05,
+      "loss": 2.7325,
+      "step": 5863
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.181600520586286e-05,
+      "loss": 1.6945,
+      "step": 5864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.176610869075457e-05,
+      "loss": 2.8402,
+      "step": 5865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.171622406429127e-05,
+      "loss": 2.4861,
+      "step": 5866
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.166635133405366e-05,
+      "loss": 2.0563,
+      "step": 5867
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.161649050762048e-05,
+      "loss": 2.8707,
+      "step": 5868
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.156664159256879e-05,
+      "loss": 2.0877,
+      "step": 5869
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.151680459647369e-05,
+      "loss": 2.9443,
+      "step": 5870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.146697952690864e-05,
+      "loss": 1.6752,
+      "step": 5871
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.141716639144511e-05,
+      "loss": 2.5893,
+      "step": 5872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.136736519765294e-05,
+      "loss": 2.5214,
+      "step": 5873
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.131757595310003e-05,
+      "loss": 2.2063,
+      "step": 5874
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.12677986653524e-05,
+      "loss": 2.2598,
+      "step": 5875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.12180333419745e-05,
+      "loss": 2.8333,
+      "step": 5876
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.116827999052872e-05,
+      "loss": 2.6069,
+      "step": 5877
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.111853861857568e-05,
+      "loss": 1.1844,
+      "step": 5878
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.106880923367435e-05,
+      "loss": 2.4165,
+      "step": 5879
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.101909184338167e-05,
+      "loss": 1.9084,
+      "step": 5880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.09693864552528e-05,
+      "loss": 2.2695,
+      "step": 5881
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.09196930768412e-05,
+      "loss": 2.3765,
+      "step": 5882
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.087001171569832e-05,
+      "loss": 2.3404,
+      "step": 5883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.082034237937399e-05,
+      "loss": 2.4938,
+      "step": 5884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.077068507541601e-05,
+      "loss": 2.6501,
+      "step": 5885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.072103981137053e-05,
+      "loss": 2.6501,
+      "step": 5886
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.067140659478168e-05,
+      "loss": 2.0051,
+      "step": 5887
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.062178543319197e-05,
+      "loss": 2.0066,
+      "step": 5888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.057217633414188e-05,
+      "loss": 2.5465,
+      "step": 5889
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.052257930517025e-05,
+      "loss": 2.2172,
+      "step": 5890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.04729943538139e-05,
+      "loss": 3.0374,
+      "step": 5891
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.042342148760797e-05,
+      "loss": 2.723,
+      "step": 5892
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.037386071408561e-05,
+      "loss": 2.4347,
+      "step": 5893
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.032431204077832e-05,
+      "loss": 2.9347,
+      "step": 5894
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.027477547521555e-05,
+      "loss": 2.5993,
+      "step": 5895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.022525102492512e-05,
+      "loss": 2.5296,
+      "step": 5896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.01757386974328e-05,
+      "loss": 2.6198,
+      "step": 5897
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.012623850026274e-05,
+      "loss": 1.6614,
+      "step": 5898
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.007675044093702e-05,
+      "loss": 2.5352,
+      "step": 5899
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.002727452697609e-05,
+      "loss": 2.1343,
+      "step": 5900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.997781076589834e-05,
+      "loss": 2.5605,
+      "step": 5901
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.992835916522055e-05,
+      "loss": 3.0306,
+      "step": 5902
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.98789197324574e-05,
+      "loss": 2.9792,
+      "step": 5903
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.982949247512197e-05,
+      "loss": 2.4445,
+      "step": 5904
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.97800774007253e-05,
+      "loss": 2.7111,
+      "step": 5905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.973067451677659e-05,
+      "loss": 2.5195,
+      "step": 5906
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.968128383078335e-05,
+      "loss": 2.3151,
+      "step": 5907
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.963190535025109e-05,
+      "loss": 2.7269,
+      "step": 5908
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.958253908268345e-05,
+      "loss": 2.6539,
+      "step": 5909
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.953318503558236e-05,
+      "loss": 2.2482,
+      "step": 5910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.948384321644769e-05,
+      "loss": 2.6024,
+      "step": 5911
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.943451363277771e-05,
+      "loss": 2.5418,
+      "step": 5912
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.938519629206852e-05,
+      "loss": 2.3901,
+      "step": 5913
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.933589120181468e-05,
+      "loss": 2.3938,
+      "step": 5914
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.92865983695086e-05,
+      "loss": 2.3585,
+      "step": 5915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.92373178026411e-05,
+      "loss": 1.7509,
+      "step": 5916
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.918804950870083e-05,
+      "loss": 2.0287,
+      "step": 5917
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.913879349517488e-05,
+      "loss": 2.2627,
+      "step": 5918
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.908954976954825e-05,
+      "loss": 2.8687,
+      "step": 5919
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.904031833930421e-05,
+      "loss": 2.7475,
+      "step": 5920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.899109921192404e-05,
+      "loss": 2.871,
+      "step": 5921
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.894189239488732e-05,
+      "loss": 2.4407,
+      "step": 5922
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.889269789567153e-05,
+      "loss": 1.9867,
+      "step": 5923
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.884351572175257e-05,
+      "loss": 2.3266,
+      "step": 5924
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.879434588060411e-05,
+      "loss": 2.4235,
+      "step": 5925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.874518837969832e-05,
+      "loss": 2.371,
+      "step": 5926
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.869604322650519e-05,
+      "loss": 2.3762,
+      "step": 5927
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.864691042849303e-05,
+      "loss": 2.5712,
+      "step": 5928
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.859778999312813e-05,
+      "loss": 2.8315,
+      "step": 5929
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.854868192787507e-05,
+      "loss": 2.413,
+      "step": 5930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.849958624019634e-05,
+      "loss": 2.8131,
+      "step": 5931
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.845050293755277e-05,
+      "loss": 2.4684,
+      "step": 5932
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.840143202740308e-05,
+      "loss": 2.4998,
+      "step": 5933
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.835237351720433e-05,
+      "loss": 2.3198,
+      "step": 5934
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.830332741441158e-05,
+      "loss": 2.1478,
+      "step": 5935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.825429372647791e-05,
+      "loss": 2.356,
+      "step": 5936
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.820527246085477e-05,
+      "loss": 2.4549,
+      "step": 5937
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.815626362499147e-05,
+      "loss": 2.4776,
+      "step": 5938
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.810726722633553e-05,
+      "loss": 2.5702,
+      "step": 5939
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.805828327233264e-05,
+      "loss": 2.4326,
+      "step": 5940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.800931177042647e-05,
+      "loss": 2.1169,
+      "step": 5941
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.796035272805895e-05,
+      "loss": 1.8356,
+      "step": 5942
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.791140615266996e-05,
+      "loss": 2.7486,
+      "step": 5943
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.786247205169764e-05,
+      "loss": 2.9572,
+      "step": 5944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.781355043257805e-05,
+      "loss": 2.7754,
+      "step": 5945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.776464130274559e-05,
+      "loss": 2.6674,
+      "step": 5946
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.771574466963256e-05,
+      "loss": 2.086,
+      "step": 5947
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.766686054066939e-05,
+      "loss": 2.4878,
+      "step": 5948
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.761798892328472e-05,
+      "loss": 2.3169,
+      "step": 5949
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.756912982490516e-05,
+      "loss": 2.7175,
+      "step": 5950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.752028325295556e-05,
+      "loss": 2.207,
+      "step": 5951
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.747144921485869e-05,
+      "loss": 2.3963,
+      "step": 5952
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.742262771803561e-05,
+      "loss": 2.5754,
+      "step": 5953
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.737381876990526e-05,
+      "loss": 2.3031,
+      "step": 5954
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.73250223778849e-05,
+      "loss": 2.7988,
+      "step": 5955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.727623854938966e-05,
+      "loss": 2.1824,
+      "step": 5956
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.7227467291833e-05,
+      "loss": 2.612,
+      "step": 5957
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.717870861262619e-05,
+      "loss": 2.3858,
+      "step": 5958
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.712996251917888e-05,
+      "loss": 2.5036,
+      "step": 5959
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.708122901889854e-05,
+      "loss": 2.4837,
+      "step": 5960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.703250811919096e-05,
+      "loss": 2.2867,
+      "step": 5961
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.698379982745979e-05,
+      "loss": 2.1039,
+      "step": 5962
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.693510415110702e-05,
+      "loss": 2.0415,
+      "step": 5963
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.688642109753245e-05,
+      "loss": 2.3339,
+      "step": 5964
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.68377506741342e-05,
+      "loss": 2.5558,
+      "step": 5965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.678909288830836e-05,
+      "loss": 2.6157,
+      "step": 5966
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.674044774744898e-05,
+      "loss": 2.6091,
+      "step": 5967
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.669181525894846e-05,
+      "loss": 2.0301,
+      "step": 5968
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.66431954301971e-05,
+      "loss": 1.81,
+      "step": 5969
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.65945882685832e-05,
+      "loss": 2.2413,
+      "step": 5970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.654599378149338e-05,
+      "loss": 2.8193,
+      "step": 5971
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.64974119763121e-05,
+      "loss": 2.5642,
+      "step": 5972
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.644884286042208e-05,
+      "loss": 2.2523,
+      "step": 5973
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.640028644120391e-05,
+      "loss": 2.3325,
+      "step": 5974
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.635174272603649e-05,
+      "loss": 2.9004,
+      "step": 5975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.630321172229653e-05,
+      "loss": 2.1686,
+      "step": 5976
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.625469343735907e-05,
+      "loss": 2.412,
+      "step": 5977
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.620618787859697e-05,
+      "loss": 2.0981,
+      "step": 5978
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.615769505338137e-05,
+      "loss": 2.5439,
+      "step": 5979
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.610921496908127e-05,
+      "loss": 2.4032,
+      "step": 5980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.606074763306397e-05,
+      "loss": 2.4924,
+      "step": 5981
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.601229305269457e-05,
+      "loss": 2.6007,
+      "step": 5982
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.596385123533651e-05,
+      "loss": 1.7672,
+      "step": 5983
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.5915422188351e-05,
+      "loss": 2.3671,
+      "step": 5984
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.586700591909761e-05,
+      "loss": 2.3022,
+      "step": 5985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.581860243493364e-05,
+      "loss": 2.0868,
+      "step": 5986
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.577021174321479e-05,
+      "loss": 2.1347,
+      "step": 5987
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.572183385129452e-05,
+      "loss": 2.5206,
+      "step": 5988
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.567346876652457e-05,
+      "loss": 2.6736,
+      "step": 5989
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.562511649625453e-05,
+      "loss": 2.1991,
+      "step": 5990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.557677704783227e-05,
+      "loss": 2.6747,
+      "step": 5991
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.552845042860346e-05,
+      "loss": 1.6769,
+      "step": 5992
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.548013664591208e-05,
+      "loss": 2.0362,
+      "step": 5993
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.543183570709992e-05,
+      "loss": 2.5124,
+      "step": 5994
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.538354761950703e-05,
+      "loss": 2.9674,
+      "step": 5995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.533527239047135e-05,
+      "loss": 2.4457,
+      "step": 5996
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.528701002732888e-05,
+      "loss": 2.083,
+      "step": 5997
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.523876053741378e-05,
+      "loss": 1.8841,
+      "step": 5998
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.519052392805817e-05,
+      "loss": 2.4671,
+      "step": 5999
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.514230020659215e-05,
+      "loss": 2.7667,
+      "step": 6000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.509408938034402e-05,
+      "loss": 2.929,
+      "step": 6001
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.504589145663996e-05,
+      "loss": 1.8538,
+      "step": 6002
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.499770644280436e-05,
+      "loss": 2.2486,
+      "step": 6003
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.494953434615945e-05,
+      "loss": 2.0094,
+      "step": 6004
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.49013751740257e-05,
+      "loss": 2.1131,
+      "step": 6005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.48532289337214e-05,
+      "loss": 1.899,
+      "step": 6006
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.480509563256311e-05,
+      "loss": 2.0198,
+      "step": 6007
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.475697527786517e-05,
+      "loss": 2.4667,
+      "step": 6008
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.470886787694023e-05,
+      "loss": 2.2596,
+      "step": 6009
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.46607734370987e-05,
+      "loss": 2.6143,
+      "step": 6010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.461269196564926e-05,
+      "loss": 2.161,
+      "step": 6011
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.456462346989842e-05,
+      "loss": 1.4429,
+      "step": 6012
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.451656795715088e-05,
+      "loss": 1.2206,
+      "step": 6013
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.44685254347092e-05,
+      "loss": 2.2661,
+      "step": 6014
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.442049590987416e-05,
+      "loss": 2.6712,
+      "step": 6015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.437247938994437e-05,
+      "loss": 2.7689,
+      "step": 6016
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.432447588221665e-05,
+      "loss": 2.8455,
+      "step": 6017
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.427648539398564e-05,
+      "loss": 2.3681,
+      "step": 6018
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.422850793254423e-05,
+      "loss": 2.4921,
+      "step": 6019
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.418054350518316e-05,
+      "loss": 2.6913,
+      "step": 6020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.413259211919118e-05,
+      "loss": 2.8998,
+      "step": 6021
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.408465378185522e-05,
+      "loss": 2.6882,
+      "step": 6022
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.403672850046005e-05,
+      "loss": 2.5689,
+      "step": 6023
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.398881628228859e-05,
+      "loss": 2.5639,
+      "step": 6024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.394091713462167e-05,
+      "loss": 2.6069,
+      "step": 6025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.389303106473822e-05,
+      "loss": 2.2116,
+      "step": 6026
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.384515807991513e-05,
+      "loss": 2.2369,
+      "step": 6027
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.379729818742728e-05,
+      "loss": 2.5654,
+      "step": 6028
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.374945139454766e-05,
+      "loss": 2.3327,
+      "step": 6029
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.370161770854717e-05,
+      "loss": 2.5961,
+      "step": 6030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.365379713669471e-05,
+      "loss": 2.2561,
+      "step": 6031
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.360598968625733e-05,
+      "loss": 2.5111,
+      "step": 6032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.355819536449986e-05,
+      "loss": 2.1366,
+      "step": 6033
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.351041417868543e-05,
+      "loss": 1.7692,
+      "step": 6034
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.34626461360748e-05,
+      "loss": 2.186,
+      "step": 6035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.341489124392714e-05,
+      "loss": 1.3287,
+      "step": 6036
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.336714950949925e-05,
+      "loss": 2.3193,
+      "step": 6037
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.331942094004623e-05,
+      "loss": 2.2147,
+      "step": 6038
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.327170554282094e-05,
+      "loss": 3.0163,
+      "step": 6039
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.322400332507447e-05,
+      "loss": 2.0093,
+      "step": 6040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.317631429405563e-05,
+      "loss": 2.4214,
+      "step": 6041
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.312863845701155e-05,
+      "loss": 2.8127,
+      "step": 6042
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.308097582118705e-05,
+      "loss": 2.3786,
+      "step": 6043
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.303332639382518e-05,
+      "loss": 3.2575,
+      "step": 6044
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.298569018216678e-05,
+      "loss": 2.0754,
+      "step": 6045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.29380671934509e-05,
+      "loss": 2.7308,
+      "step": 6046
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.289045743491436e-05,
+      "loss": 2.7811,
+      "step": 6047
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.284286091379217e-05,
+      "loss": 3.0464,
+      "step": 6048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.27952776373171e-05,
+      "loss": 1.9426,
+      "step": 6049
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.274770761272022e-05,
+      "loss": 2.2223,
+      "step": 6050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.270015084723023e-05,
+      "loss": 3.0851,
+      "step": 6051
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.265260734807415e-05,
+      "loss": 2.5676,
+      "step": 6052
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.260507712247669e-05,
+      "loss": 2.2769,
+      "step": 6053
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.255756017766077e-05,
+      "loss": 2.468,
+      "step": 6054
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.251005652084712e-05,
+      "loss": 2.5833,
+      "step": 6055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.246256615925464e-05,
+      "loss": 2.4614,
+      "step": 6056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.241508910010002e-05,
+      "loss": 2.3307,
+      "step": 6057
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.2367625350598e-05,
+      "loss": 2.5236,
+      "step": 6058
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.232017491796137e-05,
+      "loss": 2.1979,
+      "step": 6059
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.22727378094008e-05,
+      "loss": 2.3562,
+      "step": 6060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.222531403212492e-05,
+      "loss": 1.8247,
+      "step": 6061
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.217790359334048e-05,
+      "loss": 2.6294,
+      "step": 6062
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.213050650025198e-05,
+      "loss": 2.0019,
+      "step": 6063
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.208312276006217e-05,
+      "loss": 2.752,
+      "step": 6064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.203575237997146e-05,
+      "loss": 2.3449,
+      "step": 6065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.198839536717855e-05,
+      "loss": 1.84,
+      "step": 6066
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.194105172887979e-05,
+      "loss": 2.769,
+      "step": 6067
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.189372147226977e-05,
+      "loss": 2.4133,
+      "step": 6068
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.184640460454086e-05,
+      "loss": 2.4861,
+      "step": 6069
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.179910113288353e-05,
+      "loss": 2.2385,
+      "step": 6070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.175181106448607e-05,
+      "loss": 2.6257,
+      "step": 6071
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.17045344065349e-05,
+      "loss": 1.7967,
+      "step": 6072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.165727116621423e-05,
+      "loss": 2.5518,
+      "step": 6073
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.16100213507064e-05,
+      "loss": 2.5288,
+      "step": 6074
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.156278496719154e-05,
+      "loss": 2.635,
+      "step": 6075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.151556202284792e-05,
+      "loss": 2.6437,
+      "step": 6076
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.14683525248516e-05,
+      "loss": 2.4823,
+      "step": 6077
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.142115648037672e-05,
+      "loss": 2.3802,
+      "step": 6078
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.137397389659526e-05,
+      "loss": 2.688,
+      "step": 6079
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.132680478067731e-05,
+      "loss": 2.1557,
+      "step": 6080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.127964913979072e-05,
+      "loss": 2.2612,
+      "step": 6081
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.12325069811015e-05,
+      "loss": 2.8345,
+      "step": 6082
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.11853783117734e-05,
+      "loss": 2.6915,
+      "step": 6083
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.113826313896833e-05,
+      "loss": 2.1871,
+      "step": 6084
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.109116146984603e-05,
+      "loss": 2.572,
+      "step": 6085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.10440733115641e-05,
+      "loss": 2.3655,
+      "step": 6086
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.09969986712783e-05,
+      "loss": 3.1026,
+      "step": 6087
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.094993755614221e-05,
+      "loss": 2.9078,
+      "step": 6088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.09028899733073e-05,
+      "loss": 2.0003,
+      "step": 6089
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0855855929923156e-05,
+      "loss": 2.263,
+      "step": 6090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0808835433137096e-05,
+      "loss": 2.9063,
+      "step": 6091
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.07618284900946e-05,
+      "loss": 2.6023,
+      "step": 6092
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.071483510793893e-05,
+      "loss": 2.5014,
+      "step": 6093
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.066785529381126e-05,
+      "loss": 2.5269,
+      "step": 6094
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.062088905485092e-05,
+      "loss": 1.5871,
+      "step": 6095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.05739363981949e-05,
+      "loss": 2.0469,
+      "step": 6096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0526997330978365e-05,
+      "loss": 2.1797,
+      "step": 6097
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.048007186033421e-05,
+      "loss": 2.4733,
+      "step": 6098
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.043315999339349e-05,
+      "loss": 2.0601,
+      "step": 6099
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.038626173728491e-05,
+      "loss": 2.2347,
+      "step": 6100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0339377099135405e-05,
+      "loss": 2.5666,
+      "step": 6101
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0292506086069595e-05,
+      "loss": 2.7474,
+      "step": 6102
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.024564870521022e-05,
+      "loss": 2.3743,
+      "step": 6103
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.019880496367776e-05,
+      "loss": 2.5512,
+      "step": 6104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.015197486859084e-05,
+      "loss": 2.0959,
+      "step": 6105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0105158427065786e-05,
+      "loss": 2.6087,
+      "step": 6106
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.005835564621705e-05,
+      "loss": 2.1343,
+      "step": 6107
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 6.0011566533156825e-05,
+      "loss": 2.243,
+      "step": 6108
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9964791094995424e-05,
+      "loss": 2.0576,
+      "step": 6109
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.991802933884085e-05,
+      "loss": 2.9201,
+      "step": 6110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9871281271799285e-05,
+      "loss": 2.4298,
+      "step": 6111
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.982454690097457e-05,
+      "loss": 1.9562,
+      "step": 6112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9777826233468714e-05,
+      "loss": 2.3586,
+      "step": 6113
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.973111927638142e-05,
+      "loss": 2.7082,
+      "step": 6114
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.968442603681049e-05,
+      "loss": 2.211,
+      "step": 6115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.963774652185149e-05,
+      "loss": 2.3306,
+      "step": 6116
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.959108073859806e-05,
+      "loss": 1.3536,
+      "step": 6117
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.954442869414161e-05,
+      "loss": 2.536,
+      "step": 6118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.949779039557148e-05,
+      "loss": 1.632,
+      "step": 6119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9451165849975054e-05,
+      "loss": 3.1021,
+      "step": 6120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.940455506443747e-05,
+      "loss": 1.9783,
+      "step": 6121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.935795804604178e-05,
+      "loss": 2.4591,
+      "step": 6122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9311374801869135e-05,
+      "loss": 2.0544,
+      "step": 6123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.926480533899833e-05,
+      "loss": 2.4848,
+      "step": 6124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.921824966450631e-05,
+      "loss": 2.587,
+      "step": 6125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.917170778546768e-05,
+      "loss": 2.1443,
+      "step": 6126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9125179708955196e-05,
+      "loss": 0.9859,
+      "step": 6127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9078665442039294e-05,
+      "loss": 2.3591,
+      "step": 6128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.9032164991788515e-05,
+      "loss": 2.4959,
+      "step": 6129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8985678365269114e-05,
+      "loss": 2.7232,
+      "step": 6130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8939205569545415e-05,
+      "loss": 2.4773,
+      "step": 6131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.889274661167945e-05,
+      "loss": 2.3852,
+      "step": 6132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.884630149873136e-05,
+      "loss": 1.4233,
+      "step": 6133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.879987023775898e-05,
+      "loss": 2.2992,
+      "step": 6134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8753452835818246e-05,
+      "loss": 2.6086,
+      "step": 6135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8707049299962765e-05,
+      "loss": 2.5682,
+      "step": 6136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.866065963724427e-05,
+      "loss": 2.3569,
+      "step": 6137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.861428385471213e-05,
+      "loss": 2.5727,
+      "step": 6138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.856792195941388e-05,
+      "loss": 1.9718,
+      "step": 6139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.852157395839468e-05,
+      "loss": 2.4118,
+      "step": 6140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.847523985869782e-05,
+      "loss": 2.0446,
+      "step": 6141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.842891966736424e-05,
+      "loss": 2.0953,
+      "step": 6142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8382613391433e-05,
+      "loss": 1.9115,
+      "step": 6143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.833632103794085e-05,
+      "loss": 2.5159,
+      "step": 6144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8290042613922596e-05,
+      "loss": 2.7102,
+      "step": 6145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.824377812641078e-05,
+      "loss": 2.7833,
+      "step": 6146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.819752758243585e-05,
+      "loss": 1.5696,
+      "step": 6147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.815129098902627e-05,
+      "loss": 2.0746,
+      "step": 6148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.810506835320822e-05,
+      "loss": 2.4794,
+      "step": 6149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.8058859682005764e-05,
+      "loss": 2.2327,
+      "step": 6150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.801266498244104e-05,
+      "loss": 2.5324,
+      "step": 6151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.79664842615338e-05,
+      "loss": 2.6401,
+      "step": 6152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.792031752630191e-05,
+      "loss": 2.4873,
+      "step": 6153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.78741647837609e-05,
+      "loss": 2.6931,
+      "step": 6154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.782802604092434e-05,
+      "loss": 2.0989,
+      "step": 6155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.778190130480356e-05,
+      "loss": 2.7355,
+      "step": 6156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.773579058240786e-05,
+      "loss": 2.0627,
+      "step": 6157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.7689693880744274e-05,
+      "loss": 1.8609,
+      "step": 6158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.764361120681789e-05,
+      "loss": 1.7418,
+      "step": 6159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.759754256763145e-05,
+      "loss": 2.3429,
+      "step": 6160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.75514879701858e-05,
+      "loss": 1.8838,
+      "step": 6161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.750544742147941e-05,
+      "loss": 2.3187,
+      "step": 6162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.745942092850884e-05,
+      "loss": 2.4651,
+      "step": 6163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.741340849826835e-05,
+      "loss": 2.651,
+      "step": 6164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.736741013775008e-05,
+      "loss": 2.5953,
+      "step": 6165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.732142585394416e-05,
+      "loss": 2.7948,
+      "step": 6166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.7275455653838394e-05,
+      "loss": 2.5566,
+      "step": 6167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.722949954441866e-05,
+      "loss": 2.225,
+      "step": 6168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.718355753266846e-05,
+      "loss": 2.331,
+      "step": 6169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.713762962556937e-05,
+      "loss": 2.8067,
+      "step": 6170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.709171583010065e-05,
+      "loss": 2.2019,
+      "step": 6171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.704581615323958e-05,
+      "loss": 2.0716,
+      "step": 6172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.699993060196111e-05,
+      "loss": 2.503,
+      "step": 6173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.695405918323822e-05,
+      "loss": 2.3734,
+      "step": 6174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.690820190404158e-05,
+      "loss": 2.3459,
+      "step": 6175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.686235877133989e-05,
+      "loss": 2.6399,
+      "step": 6176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.68165297920995e-05,
+      "loss": 1.9292,
+      "step": 6177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.67707149732848e-05,
+      "loss": 2.1222,
+      "step": 6178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.672491432185791e-05,
+      "loss": 2.8753,
+      "step": 6179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.6679127844778774e-05,
+      "loss": 2.2276,
+      "step": 6180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.663335554900534e-05,
+      "loss": 2.4084,
+      "step": 6181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.6587597441493235e-05,
+      "loss": 2.3772,
+      "step": 6182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.654185352919592e-05,
+      "loss": 2.1614,
+      "step": 6183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.649612381906493e-05,
+      "loss": 2.2813,
+      "step": 6184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.645040831804933e-05,
+      "loss": 2.4491,
+      "step": 6185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.640470703309629e-05,
+      "loss": 2.8133,
+      "step": 6186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.635901997115063e-05,
+      "loss": 2.5146,
+      "step": 6187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.6313347139155143e-05,
+      "loss": 2.4765,
+      "step": 6188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.626768854405035e-05,
+      "loss": 2.1877,
+      "step": 6189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.6222044192774736e-05,
+      "loss": 2.155,
+      "step": 6190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.617641409226446e-05,
+      "loss": 2.6764,
+      "step": 6191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.6130798249453665e-05,
+      "loss": 2.6359,
+      "step": 6192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.608519667127421e-05,
+      "loss": 2.8273,
+      "step": 6193
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.603960936465591e-05,
+      "loss": 1.0064,
+      "step": 6194
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.599403633652627e-05,
+      "loss": 2.0384,
+      "step": 6195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.594847759381078e-05,
+      "loss": 2.426,
+      "step": 6196
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.590293314343257e-05,
+      "loss": 2.4934,
+      "step": 6197
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5857402992312823e-05,
+      "loss": 2.4837,
+      "step": 6198
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5811887147370314e-05,
+      "loss": 2.6095,
+      "step": 6199
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.576638561552187e-05,
+      "loss": 2.3913,
+      "step": 6200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.572089840368192e-05,
+      "loss": 2.1154,
+      "step": 6201
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.567542551876295e-05,
+      "loss": 2.5948,
+      "step": 6202
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5629966967675025e-05,
+      "loss": 2.9488,
+      "step": 6203
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5584522757326286e-05,
+      "loss": 1.7684,
+      "step": 6204
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5539092894622446e-05,
+      "loss": 2.574,
+      "step": 6205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.549367738646725e-05,
+      "loss": 2.533,
+      "step": 6206
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.544827623976213e-05,
+      "loss": 3.3376,
+      "step": 6207
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.540288946140633e-05,
+      "loss": 2.4464,
+      "step": 6208
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.535751705829705e-05,
+      "loss": 2.6082,
+      "step": 6209
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.531215903732915e-05,
+      "loss": 2.4335,
+      "step": 6210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5266815405395314e-05,
+      "loss": 2.1927,
+      "step": 6211
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5221486169386194e-05,
+      "loss": 2.8779,
+      "step": 6212
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.5176171336190056e-05,
+      "loss": 2.5123,
+      "step": 6213
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.513087091269316e-05,
+      "loss": 2.2153,
+      "step": 6214
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.50855849057794e-05,
+      "loss": 2.9786,
+      "step": 6215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.504031332233066e-05,
+      "loss": 2.3162,
+      "step": 6216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.499505616922644e-05,
+      "loss": 2.4712,
+      "step": 6217
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.494981345334423e-05,
+      "loss": 2.1635,
+      "step": 6218
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.490458518155916e-05,
+      "loss": 2.6793,
+      "step": 6219
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4859371360744345e-05,
+      "loss": 2.4787,
+      "step": 6220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4814171997770526e-05,
+      "loss": 2.2319,
+      "step": 6221
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4768987099506376e-05,
+      "loss": 2.3664,
+      "step": 6222
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.472381667281827e-05,
+      "loss": 2.6504,
+      "step": 6223
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.467866072457051e-05,
+      "loss": 2.6853,
+      "step": 6224
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.463351926162503e-05,
+      "loss": 2.6265,
+      "step": 6225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4588392290841745e-05,
+      "loss": 3.0414,
+      "step": 6226
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4543279819078196e-05,
+      "loss": 2.1763,
+      "step": 6227
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.449818185318988e-05,
+      "loss": 2.6777,
+      "step": 6228
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.445309840002994e-05,
+      "loss": 2.597,
+      "step": 6229
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.440802946644945e-05,
+      "loss": 2.77,
+      "step": 6230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.436297505929715e-05,
+      "loss": 2.872,
+      "step": 6231
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4317935185419727e-05,
+      "loss": 2.135,
+      "step": 6232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.427290985166146e-05,
+      "loss": 2.6711,
+      "step": 6233
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4227899064864625e-05,
+      "loss": 2.3971,
+      "step": 6234
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.4182902831869084e-05,
+      "loss": 2.2825,
+      "step": 6235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.413792115951273e-05,
+      "loss": 2.2815,
+      "step": 6236
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.409295405463101e-05,
+      "loss": 2.3651,
+      "step": 6237
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.404800152405722e-05,
+      "loss": 2.0686,
+      "step": 6238
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.400306357462258e-05,
+      "loss": 2.4664,
+      "step": 6239
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.395814021315594e-05,
+      "loss": 2.1353,
+      "step": 6240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.3913231446483917e-05,
+      "loss": 1.6847,
+      "step": 6241
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.3868337281431096e-05,
+      "loss": 2.1552,
+      "step": 6242
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.3823457724819665e-05,
+      "loss": 2.9051,
+      "step": 6243
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.37785927834696e-05,
+      "loss": 2.9923,
+      "step": 6244
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.373374246419878e-05,
+      "loss": 2.5868,
+      "step": 6245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.368890677382273e-05,
+      "loss": 2.5745,
+      "step": 6246
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.364408571915488e-05,
+      "loss": 2.4146,
+      "step": 6247
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.359927930700627e-05,
+      "loss": 2.3705,
+      "step": 6248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.355448754418593e-05,
+      "loss": 2.8138,
+      "step": 6249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.350971043750042e-05,
+      "loss": 2.4531,
+      "step": 6250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.346494799375429e-05,
+      "loss": 2.9687,
+      "step": 6251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.34202002197497e-05,
+      "loss": 2.1503,
+      "step": 6252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.337546712228673e-05,
+      "loss": 2.8385,
+      "step": 6253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.333074870816303e-05,
+      "loss": 2.3172,
+      "step": 6254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.328604498417428e-05,
+      "loss": 1.6517,
+      "step": 6255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.3241355957113666e-05,
+      "loss": 2.5944,
+      "step": 6256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.3196681633772345e-05,
+      "loss": 2.1004,
+      "step": 6257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.315202202093907e-05,
+      "loss": 2.907,
+      "step": 6258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.310737712540055e-05,
+      "loss": 2.7351,
+      "step": 6259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.306274695394104e-05,
+      "loss": 2.2582,
+      "step": 6260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.301813151334276e-05,
+      "loss": 2.3867,
+      "step": 6261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.29735308103855e-05,
+      "loss": 2.1254,
+      "step": 6262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.292894485184704e-05,
+      "loss": 2.5416,
+      "step": 6263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2884373644502664e-05,
+      "loss": 2.5521,
+      "step": 6264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2839817195125635e-05,
+      "loss": 2.7046,
+      "step": 6265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2795275510486785e-05,
+      "loss": 2.1512,
+      "step": 6266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.27507485973549e-05,
+      "loss": 2.3015,
+      "step": 6267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.270623646249636e-05,
+      "loss": 2.185,
+      "step": 6268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.266173911267531e-05,
+      "loss": 2.906,
+      "step": 6269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.26172565546538e-05,
+      "loss": 2.4063,
+      "step": 6270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.257278879519146e-05,
+      "loss": 2.501,
+      "step": 6271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.25283358410457e-05,
+      "loss": 1.9819,
+      "step": 6272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2483897698971824e-05,
+      "loss": 2.1027,
+      "step": 6273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.243947437572266e-05,
+      "loss": 2.652,
+      "step": 6274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2395065878049004e-05,
+      "loss": 2.0521,
+      "step": 6275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.235067221269921e-05,
+      "loss": 1.8444,
+      "step": 6276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.230629338641958e-05,
+      "loss": 2.6536,
+      "step": 6277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2261929405953904e-05,
+      "loss": 2.7427,
+      "step": 6278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2217580278044004e-05,
+      "loss": 2.3114,
+      "step": 6279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.217324600942916e-05,
+      "loss": 2.3852,
+      "step": 6280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.212892660684666e-05,
+      "loss": 2.5575,
+      "step": 6281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.2084622077031303e-05,
+      "loss": 1.7891,
+      "step": 6282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.204033242671583e-05,
+      "loss": 2.6275,
+      "step": 6283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.199605766263051e-05,
+      "loss": 2.6189,
+      "step": 6284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1951797791503586e-05,
+      "loss": 2.3211,
+      "step": 6285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.19075528200608e-05,
+      "loss": 2.762,
+      "step": 6286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.186332275502584e-05,
+      "loss": 2.2423,
+      "step": 6287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.181910760311992e-05,
+      "loss": 2.61,
+      "step": 6288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1774907371062255e-05,
+      "loss": 2.1813,
+      "step": 6289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.173072206556948e-05,
+      "loss": 2.7315,
+      "step": 6290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1686551693356234e-05,
+      "loss": 2.2547,
+      "step": 6291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.164239626113469e-05,
+      "loss": 2.3359,
+      "step": 6292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.159825577561493e-05,
+      "loss": 2.4293,
+      "step": 6293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1554130243504554e-05,
+      "loss": 2.6749,
+      "step": 6294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1510019671509105e-05,
+      "loss": 2.7837,
+      "step": 6295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.146592406633166e-05,
+      "loss": 2.4102,
+      "step": 6296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.142184343467322e-05,
+      "loss": 2.3302,
+      "step": 6297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.137777778323233e-05,
+      "loss": 2.6801,
+      "step": 6298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1333727118705305e-05,
+      "loss": 2.8221,
+      "step": 6299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.12896914477863e-05,
+      "loss": 2.6223,
+      "step": 6300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1245670777167055e-05,
+      "loss": 1.8476,
+      "step": 6301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.120166511353701e-05,
+      "loss": 2.7637,
+      "step": 6302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.1157674463583506e-05,
+      "loss": 1.9676,
+      "step": 6303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.111369883399139e-05,
+      "loss": 1.7854,
+      "step": 6304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.106973823144342e-05,
+      "loss": 2.3222,
+      "step": 6305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.102579266261986e-05,
+      "loss": 2.3766,
+      "step": 6306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.098186213419893e-05,
+      "loss": 2.8824,
+      "step": 6307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0937946652856334e-05,
+      "loss": 2.4937,
+      "step": 6308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.089404622526566e-05,
+      "loss": 2.3467,
+      "step": 6309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.085016085809813e-05,
+      "loss": 2.6013,
+      "step": 6310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.080629055802263e-05,
+      "loss": 2.115,
+      "step": 6311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.076243533170591e-05,
+      "loss": 2.426,
+      "step": 6312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.071859518581223e-05,
+      "loss": 2.1891,
+      "step": 6313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.067477012700379e-05,
+      "loss": 1.8092,
+      "step": 6314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.063096016194024e-05,
+      "loss": 2.1185,
+      "step": 6315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.058716529727916e-05,
+      "loss": 1.7978,
+      "step": 6316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.05433855396757e-05,
+      "loss": 3.094,
+      "step": 6317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.049962089578279e-05,
+      "loss": 2.6297,
+      "step": 6318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.045587137225097e-05,
+      "loss": 2.1724,
+      "step": 6319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.041213697572863e-05,
+      "loss": 2.6368,
+      "step": 6320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0368417712861665e-05,
+      "loss": 2.5765,
+      "step": 6321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.03247135902939e-05,
+      "loss": 2.4262,
+      "step": 6322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0281024614666617e-05,
+      "loss": 2.311,
+      "step": 6323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.023735079261903e-05,
+      "loss": 1.9914,
+      "step": 6324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0193692130787815e-05,
+      "loss": 2.5269,
+      "step": 6325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.01500486358076e-05,
+      "loss": 2.3388,
+      "step": 6326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.010642031431047e-05,
+      "loss": 2.3952,
+      "step": 6327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.006280717292639e-05,
+      "loss": 1.5928,
+      "step": 6328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.001920921828289e-05,
+      "loss": 2.4907,
+      "step": 6329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.997562645700521e-05,
+      "loss": 2.5298,
+      "step": 6330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.993205889571638e-05,
+      "loss": 2.1119,
+      "step": 6331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9888506541037035e-05,
+      "loss": 1.9048,
+      "step": 6332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.984496939958543e-05,
+      "loss": 2.6417,
+      "step": 6333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.980144747797772e-05,
+      "loss": 2.6492,
+      "step": 6334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.97579407828275e-05,
+      "loss": 1.4554,
+      "step": 6335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9714449320746285e-05,
+      "loss": 2.3842,
+      "step": 6336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.967097309834306e-05,
+      "loss": 2.6449,
+      "step": 6337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.96275121222247e-05,
+      "loss": 2.5245,
+      "step": 6338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958406639899554e-05,
+      "loss": 2.7246,
+      "step": 6339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954063593525784e-05,
+      "loss": 2.5815,
+      "step": 6340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949722073761129e-05,
+      "loss": 2.3249,
+      "step": 6341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945382081265351e-05,
+      "loss": 2.7316,
+      "step": 6342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9410436166979585e-05,
+      "loss": 2.3411,
+      "step": 6343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9367066807182436e-05,
+      "loss": 2.2718,
+      "step": 6344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.932371273985252e-05,
+      "loss": 2.0308,
+      "step": 6345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.928037397157812e-05,
+      "loss": 2.3566,
+      "step": 6346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.923705050894504e-05,
+      "loss": 3.0893,
+      "step": 6347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9193742358536935e-05,
+      "loss": 2.3766,
+      "step": 6348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.915044952693494e-05,
+      "loss": 1.225,
+      "step": 6349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.910717202071805e-05,
+      "loss": 2.4306,
+      "step": 6350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9063909846462726e-05,
+      "loss": 2.6447,
+      "step": 6351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9020663010743314e-05,
+      "loss": 2.5637,
+      "step": 6352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8977431520131655e-05,
+      "loss": 2.5863,
+      "step": 6353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8934215381197404e-05,
+      "loss": 2.9627,
+      "step": 6354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.88910146005077e-05,
+      "loss": 2.1983,
+      "step": 6355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8847829184627585e-05,
+      "loss": 2.5507,
+      "step": 6356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.880465914011956e-05,
+      "loss": 2.575,
+      "step": 6357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.876150447354385e-05,
+      "loss": 2.5365,
+      "step": 6358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.871836519145844e-05,
+      "loss": 2.4667,
+      "step": 6359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.867524130041878e-05,
+      "loss": 2.3857,
+      "step": 6360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.863213280697823e-05,
+      "loss": 2.4407,
+      "step": 6361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.858903971768762e-05,
+      "loss": 2.621,
+      "step": 6362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.854596203909545e-05,
+      "loss": 2.6113,
+      "step": 6363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.850289977774802e-05,
+      "loss": 2.23,
+      "step": 6364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.845985294018911e-05,
+      "loss": 2.3068,
+      "step": 6365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8416821532960325e-05,
+      "loss": 2.03,
+      "step": 6366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.837380556260074e-05,
+      "loss": 2.6504,
+      "step": 6367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.833080503564728e-05,
+      "loss": 2.3239,
+      "step": 6368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8287819958634363e-05,
+      "loss": 2.6224,
+      "step": 6369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8244850338094176e-05,
+      "loss": 2.0819,
+      "step": 6370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.820189618055642e-05,
+      "loss": 1.9467,
+      "step": 6371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8158957492548656e-05,
+      "loss": 2.9282,
+      "step": 6372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.811603428059583e-05,
+      "loss": 2.2049,
+      "step": 6373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.80731265512208e-05,
+      "loss": 2.6635,
+      "step": 6374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.803023431094386e-05,
+      "loss": 2.2991,
+      "step": 6375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.79873575662831e-05,
+      "loss": 2.634,
+      "step": 6376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7944496323754136e-05,
+      "loss": 2.8663,
+      "step": 6377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7901650589870335e-05,
+      "loss": 2.5726,
+      "step": 6378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7858820371142595e-05,
+      "loss": 2.9243,
+      "step": 6379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7816005674079605e-05,
+      "loss": 2.8852,
+      "step": 6380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7773206505187514e-05,
+      "loss": 2.0784,
+      "step": 6381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7730422870970295e-05,
+      "loss": 2.7033,
+      "step": 6382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7687654777929426e-05,
+      "loss": 2.6122,
+      "step": 6383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7644902232564016e-05,
+      "loss": 2.5577,
+      "step": 6384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.760216524137096e-05,
+      "loss": 2.4183,
+      "step": 6385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7559443810844625e-05,
+      "loss": 2.6285,
+      "step": 6386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.751673794747714e-05,
+      "loss": 2.5636,
+      "step": 6387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.747404765775813e-05,
+      "loss": 2.3908,
+      "step": 6388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7431372948175015e-05,
+      "loss": 1.9613,
+      "step": 6389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.738871382521273e-05,
+      "loss": 0.9032,
+      "step": 6390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.734607029535384e-05,
+      "loss": 2.4689,
+      "step": 6391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7303442365078654e-05,
+      "loss": 2.7094,
+      "step": 6392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.726083004086499e-05,
+      "loss": 2.4763,
+      "step": 6393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7218233329188315e-05,
+      "loss": 2.3462,
+      "step": 6394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7175652236521806e-05,
+      "loss": 2.1495,
+      "step": 6395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.713308676933612e-05,
+      "loss": 2.6697,
+      "step": 6396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.709053693409972e-05,
+      "loss": 2.4025,
+      "step": 6397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.704800273727853e-05,
+      "loss": 2.2201,
+      "step": 6398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.700548418533624e-05,
+      "loss": 2.5354,
+      "step": 6399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6962981284734e-05,
+      "loss": 2.6761,
+      "step": 6400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.692049404193079e-05,
+      "loss": 2.1331,
+      "step": 6401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6878022463382954e-05,
+      "loss": 1.7664,
+      "step": 6402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6835566555544715e-05,
+      "loss": 2.2306,
+      "step": 6403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.679312632486769e-05,
+      "loss": 2.1214,
+      "step": 6404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6750701777801315e-05,
+      "loss": 2.3047,
+      "step": 6405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6708292920792454e-05,
+      "loss": 2.6856,
+      "step": 6406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.666589976028578e-05,
+      "loss": 2.3805,
+      "step": 6407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.662352230272337e-05,
+      "loss": 2.4569,
+      "step": 6408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.658116055454513e-05,
+      "loss": 2.1135,
+      "step": 6409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.653881452218835e-05,
+      "loss": 2.2634,
+      "step": 6410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.649648421208819e-05,
+      "loss": 1.7307,
+      "step": 6411
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.645416963067716e-05,
+      "loss": 1.0599,
+      "step": 6412
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6411870784385604e-05,
+      "loss": 2.4059,
+      "step": 6413
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.636958767964128e-05,
+      "loss": 2.5187,
+      "step": 6414
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.632732032286975e-05,
+      "loss": 2.1396,
+      "step": 6415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.628506872049399e-05,
+      "loss": 2.7141,
+      "step": 6416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.624283287893475e-05,
+      "loss": 2.3965,
+      "step": 6417
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.620061280461028e-05,
+      "loss": 2.3708,
+      "step": 6418
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.615840850393642e-05,
+      "loss": 2.253,
+      "step": 6419
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6116219983326724e-05,
+      "loss": 2.2374,
+      "step": 6420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.607404724919227e-05,
+      "loss": 2.0635,
+      "step": 6421
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.603189030794167e-05,
+      "loss": 1.519,
+      "step": 6422
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.59897491659813e-05,
+      "loss": 2.1156,
+      "step": 6423
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.594762382971498e-05,
+      "loss": 2.5098,
+      "step": 6424
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.590551430554429e-05,
+      "loss": 2.2436,
+      "step": 6425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.58634205998682e-05,
+      "loss": 3.2505,
+      "step": 6426
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.58213427190835e-05,
+      "loss": 2.9418,
+      "step": 6427
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.577928066958435e-05,
+      "loss": 2.2886,
+      "step": 6428
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.573723445776274e-05,
+      "loss": 2.9171,
+      "step": 6429
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.569520409000802e-05,
+      "loss": 2.2326,
+      "step": 6430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.565318957270733e-05,
+      "loss": 2.7086,
+      "step": 6431
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5611190912245256e-05,
+      "loss": 2.0145,
+      "step": 6432
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5569208115004095e-05,
+      "loss": 2.6591,
+      "step": 6433
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.552724118736359e-05,
+      "loss": 1.7991,
+      "step": 6434
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.548529013570125e-05,
+      "loss": 2.6032,
+      "step": 6435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.544335496639197e-05,
+      "loss": 2.5255,
+      "step": 6436
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5401435685808463e-05,
+      "loss": 2.2237,
+      "step": 6437
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5359532300320775e-05,
+      "loss": 2.9032,
+      "step": 6438
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5317644816296786e-05,
+      "loss": 2.5624,
+      "step": 6439
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.527577324010172e-05,
+      "loss": 2.5869,
+      "step": 6440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.52339175780986e-05,
+      "loss": 1.6994,
+      "step": 6441
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5192077836647836e-05,
+      "loss": 2.3481,
+      "step": 6442
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5150254022107636e-05,
+      "loss": 2.1536,
+      "step": 6443
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5108446140833535e-05,
+      "loss": 2.1766,
+      "step": 6444
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5066654199178884e-05,
+      "loss": 2.5974,
+      "step": 6445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.502487820349442e-05,
+      "loss": 2.4551,
+      "step": 6446
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.498311816012862e-05,
+      "loss": 2.5577,
+      "step": 6447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.4941374075427435e-05,
+      "loss": 2.7537,
+      "step": 6448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.489964595573436e-05,
+      "loss": 2.7185,
+      "step": 6449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.485793380739059e-05,
+      "loss": 2.2691,
+      "step": 6450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.481623763673482e-05,
+      "loss": 2.1702,
+      "step": 6451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.477455745010322e-05,
+      "loss": 1.6195,
+      "step": 6452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.473289325382977e-05,
+      "loss": 2.2019,
+      "step": 6453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.469124505424574e-05,
+      "loss": 2.4426,
+      "step": 6454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.464961285768025e-05,
+      "loss": 2.1288,
+      "step": 6455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.4607996670459774e-05,
+      "loss": 2.4107,
+      "step": 6456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.456639649890838e-05,
+      "loss": 1.0873,
+      "step": 6457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.4524812349347854e-05,
+      "loss": 1.8394,
+      "step": 6458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.448324422809734e-05,
+      "loss": 2.6183,
+      "step": 6459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.4441692141473736e-05,
+      "loss": 1.8717,
+      "step": 6460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.440015609579133e-05,
+      "loss": 2.3864,
+      "step": 6461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.4358636097362145e-05,
+      "loss": 2.1667,
+      "step": 6462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.431713215249561e-05,
+      "loss": 2.463,
+      "step": 6463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.427564426749883e-05,
+      "loss": 2.4291,
+      "step": 6464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.423417244867637e-05,
+      "loss": 2.3542,
+      "step": 6465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.419271670233047e-05,
+      "loss": 2.4421,
+      "step": 6466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.41512770347608e-05,
+      "loss": 2.3686,
+      "step": 6467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.410985345226472e-05,
+      "loss": 2.5585,
+      "step": 6468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.4068445961136996e-05,
+      "loss": 2.6431,
+      "step": 6469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.402705456767011e-05,
+      "loss": 2.7574,
+      "step": 6470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.398567927815392e-05,
+      "loss": 2.3786,
+      "step": 6471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.394432009887605e-05,
+      "loss": 2.568,
+      "step": 6472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.390297703612144e-05,
+      "loss": 2.4313,
+      "step": 6473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3861650096172795e-05,
+      "loss": 2.5566,
+      "step": 6474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.38203392853102e-05,
+      "loss": 2.5635,
+      "step": 6475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.377904460981144e-05,
+      "loss": 2.5867,
+      "step": 6476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3737766075951696e-05,
+      "loss": 2.0907,
+      "step": 6477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3737766075951696e-05,
+      "loss": 2.6916,
+      "step": 6478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.369650369000384e-05,
+      "loss": 2.4195,
+      "step": 6479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.365525745823818e-05,
+      "loss": 2.7716,
+      "step": 6480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.36140273869226e-05,
+      "loss": 1.972,
+      "step": 6481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3572813482322576e-05,
+      "loss": 1.9402,
+      "step": 6482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.353161575070108e-05,
+      "loss": 2.5253,
+      "step": 6483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.349043419831859e-05,
+      "loss": 2.1656,
+      "step": 6484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3449268831433247e-05,
+      "loss": 2.2709,
+      "step": 6485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.340811965630058e-05,
+      "loss": 2.3835,
+      "step": 6486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.336698667917382e-05,
+      "loss": 2.0778,
+      "step": 6487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.332586990630356e-05,
+      "loss": 3.0857,
+      "step": 6488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3284769343938126e-05,
+      "loss": 2.5519,
+      "step": 6489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.3243684998323166e-05,
+      "loss": 2.2083,
+      "step": 6490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.320261687570208e-05,
+      "loss": 2.0139,
+      "step": 6491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.316156498231559e-05,
+      "loss": 2.8097,
+      "step": 6492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.312052932440217e-05,
+      "loss": 2.3017,
+      "step": 6493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.307950990819761e-05,
+      "loss": 2.088,
+      "step": 6494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.303850673993544e-05,
+      "loss": 2.2761,
+      "step": 6495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.299751982584651e-05,
+      "loss": 2.8538,
+      "step": 6496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.29565491721594e-05,
+      "loss": 2.4218,
+      "step": 6497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2915594785100056e-05,
+      "loss": 2.2795,
+      "step": 6498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.287465667089211e-05,
+      "loss": 2.8221,
+      "step": 6499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.283373483575652e-05,
+      "loss": 1.7484,
+      "step": 6500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2792829285912e-05,
+      "loss": 2.5196,
+      "step": 6501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.275194002757459e-05,
+      "loss": 2.8179,
+      "step": 6502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2711067066958e-05,
+      "loss": 2.3386,
+      "step": 6503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.267021041027333e-05,
+      "loss": 2.9963,
+      "step": 6504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.262937006372935e-05,
+      "loss": 2.0952,
+      "step": 6505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.258854603353221e-05,
+      "loss": 2.6442,
+      "step": 6506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.254773832588572e-05,
+      "loss": 2.987,
+      "step": 6507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.250694694699104e-05,
+      "loss": 2.1521,
+      "step": 6508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.246617190304704e-05,
+      "loss": 1.9539,
+      "step": 6509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.242541320024997e-05,
+      "loss": 2.1199,
+      "step": 6510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.238467084479358e-05,
+      "loss": 2.3019,
+      "step": 6511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2343944842869326e-05,
+      "loss": 2.6558,
+      "step": 6512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.230323520066595e-05,
+      "loss": 2.4342,
+      "step": 6513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.226254192436978e-05,
+      "loss": 2.4534,
+      "step": 6514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.222186502016478e-05,
+      "loss": 2.7388,
+      "step": 6515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.218120449423223e-05,
+      "loss": 2.5018,
+      "step": 6516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.214056035275112e-05,
+      "loss": 2.2489,
+      "step": 6517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.209993260189776e-05,
+      "loss": 2.5533,
+      "step": 6518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2059321247846126e-05,
+      "loss": 2.8817,
+      "step": 6519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2018726296767564e-05,
+      "loss": 2.6946,
+      "step": 6520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.197814775483108e-05,
+      "loss": 2.8126,
+      "step": 6521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.193758562820302e-05,
+      "loss": 2.7202,
+      "step": 6522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.189703992304741e-05,
+      "loss": 2.4975,
+      "step": 6523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.185651064552561e-05,
+      "loss": 2.1505,
+      "step": 6524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.181599780179662e-05,
+      "loss": 2.4949,
+      "step": 6525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.177550139801683e-05,
+      "loss": 2.8232,
+      "step": 6526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.173502144034028e-05,
+      "loss": 2.4244,
+      "step": 6527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.169455793491838e-05,
+      "loss": 2.6154,
+      "step": 6528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1654110887899995e-05,
+      "loss": 2.9002,
+      "step": 6529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.16136803054317e-05,
+      "loss": 2.3682,
+      "step": 6530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.157326619365734e-05,
+      "loss": 2.4203,
+      "step": 6531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1532868558718445e-05,
+      "loss": 2.5197,
+      "step": 6532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.149248740675386e-05,
+      "loss": 2.2962,
+      "step": 6533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1452122743900136e-05,
+      "loss": 2.8497,
+      "step": 6534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.141177457629108e-05,
+      "loss": 2.4036,
+      "step": 6535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.137144291005822e-05,
+      "loss": 2.7757,
+      "step": 6536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1331127751330396e-05,
+      "loss": 2.4866,
+      "step": 6537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1290829106234066e-05,
+      "loss": 2.3879,
+      "step": 6538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.125054698089308e-05,
+      "loss": 2.0878,
+      "step": 6539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.121028138142888e-05,
+      "loss": 2.348,
+      "step": 6540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.117003231396032e-05,
+      "loss": 2.6056,
+      "step": 6541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.11297997846037e-05,
+      "loss": 2.4928,
+      "step": 6542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1089583799472976e-05,
+      "loss": 1.7061,
+      "step": 6543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1049384364679414e-05,
+      "loss": 2.278,
+      "step": 6544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.1009201486331823e-05,
+      "loss": 2.8864,
+      "step": 6545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0969035170536575e-05,
+      "loss": 2.5273,
+      "step": 6546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0928885423397373e-05,
+      "loss": 1.9667,
+      "step": 6547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0888752251015574e-05,
+      "loss": 2.5577,
+      "step": 6548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.084863565948984e-05,
+      "loss": 2.3868,
+      "step": 6549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0808535654916506e-05,
+      "loss": 2.1973,
+      "step": 6550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0768452243389174e-05,
+      "loss": 2.7871,
+      "step": 6551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.072838543099914e-05,
+      "loss": 2.0543,
+      "step": 6552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.068833522383497e-05,
+      "loss": 1.6387,
+      "step": 6553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0648301627982907e-05,
+      "loss": 2.4433,
+      "step": 6554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.060828464952646e-05,
+      "loss": 2.6892,
+      "step": 6555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.056828429454683e-05,
+      "loss": 2.7267,
+      "step": 6556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0528300569122504e-05,
+      "loss": 2.2591,
+      "step": 6557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.048833347932959e-05,
+      "loss": 2.5691,
+      "step": 6558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.044838303124151e-05,
+      "loss": 2.5025,
+      "step": 6559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.040844923092937e-05,
+      "loss": 2.3123,
+      "step": 6560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.036853208446151e-05,
+      "loss": 2.3663,
+      "step": 6561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0328631597903946e-05,
+      "loss": 2.2397,
+      "step": 6562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0288747777319975e-05,
+      "loss": 1.7514,
+      "step": 6563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.024888062877055e-05,
+      "loss": 2.5353,
+      "step": 6564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.02090301583139e-05,
+      "loss": 3.1644,
+      "step": 6565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.016919637200594e-05,
+      "loss": 2.4952,
+      "step": 6566
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.0129379275899785e-05,
+      "loss": 2.4382,
+      "step": 6567
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.008957887604629e-05,
+      "loss": 2.145,
+      "step": 6568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.004979517849351e-05,
+      "loss": 2.143,
+      "step": 6569
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.001002818928719e-05,
+      "loss": 2.1131,
+      "step": 6570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9970277914470413e-05,
+      "loss": 2.1013,
+      "step": 6571
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9930544360083674e-05,
+      "loss": 1.795,
+      "step": 6572
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.989082753216509e-05,
+      "loss": 2.5591,
+      "step": 6573
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.985112743675011e-05,
+      "loss": 2.4974,
+      "step": 6574
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9811444079871605e-05,
+      "loss": 2.7291,
+      "step": 6575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.97717774675601e-05,
+      "loss": 2.6305,
+      "step": 6576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.973212760584332e-05,
+      "loss": 2.581,
+      "step": 6577
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9692494500746685e-05,
+      "loss": 1.9151,
+      "step": 6578
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.965287815829284e-05,
+      "loss": 2.193,
+      "step": 6579
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9613278584502114e-05,
+      "loss": 2.4369,
+      "step": 6580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.957369578539206e-05,
+      "loss": 1.9374,
+      "step": 6581
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9534129766977896e-05,
+      "loss": 2.381,
+      "step": 6582
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.949458053527209e-05,
+      "loss": 2.5381,
+      "step": 6583
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9455048096284755e-05,
+      "loss": 2.3682,
+      "step": 6584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.941553245602324e-05,
+      "loss": 2.3602,
+      "step": 6585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.937603362049257e-05,
+      "loss": 2.1674,
+      "step": 6586
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.933655159569501e-05,
+      "loss": 2.3585,
+      "step": 6587
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.929708638763043e-05,
+      "loss": 2.5967,
+      "step": 6588
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.925763800229601e-05,
+      "loss": 2.1987,
+      "step": 6589
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9218206445686504e-05,
+      "loss": 2.279,
+      "step": 6590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.917879172379397e-05,
+      "loss": 2.4759,
+      "step": 6591
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.913939384260807e-05,
+      "loss": 2.1484,
+      "step": 6592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9100012808115735e-05,
+      "loss": 2.4106,
+      "step": 6593
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9060648626301495e-05,
+      "loss": 2.5835,
+      "step": 6594
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.902130130314715e-05,
+      "loss": 2.6717,
+      "step": 6595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.898197084463214e-05,
+      "loss": 2.4364,
+      "step": 6596
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.894265725673312e-05,
+      "loss": 2.2499,
+      "step": 6597
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.890336054542441e-05,
+      "loss": 2.4265,
+      "step": 6598
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.88640807166776e-05,
+      "loss": 2.0482,
+      "step": 6599
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.882481777646171e-05,
+      "loss": 2.0695,
+      "step": 6600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.878557173074334e-05,
+      "loss": 2.5599,
+      "step": 6601
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.874634258548638e-05,
+      "loss": 2.7976,
+      "step": 6602
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.870713034665217e-05,
+      "loss": 2.5883,
+      "step": 6603
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.86679350201996e-05,
+      "loss": 2.3261,
+      "step": 6604
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.862875661208486e-05,
+      "loss": 2.8931,
+      "step": 6605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.858959512826156e-05,
+      "loss": 2.3342,
+      "step": 6606
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.855045057468088e-05,
+      "loss": 2.2659,
+      "step": 6607
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.8511322957291265e-05,
+      "loss": 1.3468,
+      "step": 6608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.8472212282038736e-05,
+      "loss": 2.5646,
+      "step": 6609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.843311855486657e-05,
+      "loss": 2.2604,
+      "step": 6610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.8394041781715646e-05,
+      "loss": 1.9283,
+      "step": 6611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.8354981968524126e-05,
+      "loss": 2.4039,
+      "step": 6612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.831593912122769e-05,
+      "loss": 2.1361,
+      "step": 6613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.827691324575935e-05,
+      "loss": 2.9748,
+      "step": 6614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.823790434804966e-05,
+      "loss": 2.7901,
+      "step": 6615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.819891243402644e-05,
+      "loss": 1.9831,
+      "step": 6616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.8159937509615104e-05,
+      "loss": 1.9131,
+      "step": 6617
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.81209795807383e-05,
+      "loss": 2.0752,
+      "step": 6618
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.808203865331628e-05,
+      "loss": 2.6559,
+      "step": 6619
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.804311473326651e-05,
+      "loss": 2.4239,
+      "step": 6620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.800420782650409e-05,
+      "loss": 2.5583,
+      "step": 6621
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.7965317938941335e-05,
+      "loss": 2.7336,
+      "step": 6622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.792644507648814e-05,
+      "loss": 2.5746,
+      "step": 6623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.7887589245051656e-05,
+      "loss": 2.6096,
+      "step": 6624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.784875045053661e-05,
+      "loss": 2.7762,
+      "step": 6625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.780992869884495e-05,
+      "loss": 2.748,
+      "step": 6626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.777112399587627e-05,
+      "loss": 2.6201,
+      "step": 6627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.773233634752733e-05,
+      "loss": 2.0884,
+      "step": 6628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.769356575969249e-05,
+      "loss": 2.4267,
+      "step": 6629
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.765481223826337e-05,
+      "loss": 2.4039,
+      "step": 6630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.7616075789129144e-05,
+      "loss": 2.1071,
+      "step": 6631
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.757735641817627e-05,
+      "loss": 2.5849,
+      "step": 6632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.753865413128862e-05,
+      "loss": 2.5139,
+      "step": 6633
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.749996893434758e-05,
+      "loss": 2.0823,
+      "step": 6634
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.746130083323182e-05,
+      "loss": 2.5183,
+      "step": 6635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.742264983381741e-05,
+      "loss": 2.2129,
+      "step": 6636
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.7384015941977955e-05,
+      "loss": 1.7421,
+      "step": 6637
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.734539916358428e-05,
+      "loss": 1.8549,
+      "step": 6638
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.730679950450481e-05,
+      "loss": 2.8919,
+      "step": 6639
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.726821697060514e-05,
+      "loss": 2.849,
+      "step": 6640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.722965156774849e-05,
+      "loss": 2.5616,
+      "step": 6641
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.719110330179527e-05,
+      "loss": 2.5349,
+      "step": 6642
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.7152572178603494e-05,
+      "loss": 3.0305,
+      "step": 6643
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.7114058204028346e-05,
+      "loss": 2.5757,
+      "step": 6644
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.707556138392261e-05,
+      "loss": 2.6032,
+      "step": 6645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.70370817241363e-05,
+      "loss": 2.9353,
+      "step": 6646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.699861923051697e-05,
+      "loss": 2.4361,
+      "step": 6647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6960173908909414e-05,
+      "loss": 2.3558,
+      "step": 6648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.692174576515598e-05,
+      "loss": 2.4771,
+      "step": 6649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.688333480509622e-05,
+      "loss": 2.2143,
+      "step": 6650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.684494103456725e-05,
+      "loss": 1.9682,
+      "step": 6651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6806564459403446e-05,
+      "loss": 2.2871,
+      "step": 6652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.676820508543666e-05,
+      "loss": 2.3476,
+      "step": 6653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6729862918496065e-05,
+      "loss": 2.7909,
+      "step": 6654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.669153796440828e-05,
+      "loss": 2.5643,
+      "step": 6655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.665323022899721e-05,
+      "loss": 2.4752,
+      "step": 6656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6614939718084284e-05,
+      "loss": 2.1904,
+      "step": 6657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6576666437488166e-05,
+      "loss": 2.5813,
+      "step": 6658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.653841039302505e-05,
+      "loss": 2.3041,
+      "step": 6659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.650017159050841e-05,
+      "loss": 2.2399,
+      "step": 6660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.646195003574904e-05,
+      "loss": 2.5978,
+      "step": 6661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.642374573455531e-05,
+      "loss": 2.3149,
+      "step": 6662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6385558692732795e-05,
+      "loss": 2.0405,
+      "step": 6663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6347388916084494e-05,
+      "loss": 2.4766,
+      "step": 6664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6309236410410854e-05,
+      "loss": 2.6183,
+      "step": 6665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.627110118150956e-05,
+      "loss": 2.5352,
+      "step": 6666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.623298323517583e-05,
+      "loss": 2.2244,
+      "step": 6667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.61948825772021e-05,
+      "loss": 3.3138,
+      "step": 6668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.615679921337833e-05,
+      "loss": 2.1699,
+      "step": 6669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.611873314949172e-05,
+      "loss": 2.5697,
+      "step": 6670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.608068439132693e-05,
+      "loss": 2.4719,
+      "step": 6671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6042652944665924e-05,
+      "loss": 2.8259,
+      "step": 6672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.6004638815288126e-05,
+      "loss": 3.0403,
+      "step": 6673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.5966642008970225e-05,
+      "loss": 2.0942,
+      "step": 6674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.592866253148631e-05,
+      "loss": 2.4396,
+      "step": 6675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.58907003886079e-05,
+      "loss": 1.1382,
+      "step": 6676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.585275558610378e-05,
+      "loss": 2.7692,
+      "step": 6677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.581482812974021e-05,
+      "loss": 2.2853,
+      "step": 6678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.577691802528067e-05,
+      "loss": 2.4251,
+      "step": 6679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.573902527848619e-05,
+      "loss": 2.4713,
+      "step": 6680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.570114989511497e-05,
+      "loss": 2.3407,
+      "step": 6681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.566329188092273e-05,
+      "loss": 2.1648,
+      "step": 6682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.562545124166241e-05,
+      "loss": 2.8843,
+      "step": 6683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.558762798308447e-05,
+      "loss": 1.9996,
+      "step": 6684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.554982211093655e-05,
+      "loss": 2.5277,
+      "step": 6685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.551203363096381e-05,
+      "loss": 2.3404,
+      "step": 6686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.5474262548908625e-05,
+      "loss": 1.7352,
+      "step": 6687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.543650887051089e-05,
+      "loss": 2.7158,
+      "step": 6688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.5398772601507655e-05,
+      "loss": 2.2021,
+      "step": 6689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.536105374763353e-05,
+      "loss": 2.2207,
+      "step": 6690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.53233523146203e-05,
+      "loss": 2.406,
+      "step": 6691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.528566830819724e-05,
+      "loss": 2.6061,
+      "step": 6692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.5248001734090904e-05,
+      "loss": 2.5409,
+      "step": 6693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.521035259802516e-05,
+      "loss": 2.839,
+      "step": 6694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.517272090572137e-05,
+      "loss": 1.2808,
+      "step": 6695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.5135106662898086e-05,
+      "loss": 2.2602,
+      "step": 6696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.509750987527126e-05,
+      "loss": 2.2291,
+      "step": 6697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.5059930548554286e-05,
+      "loss": 2.6033,
+      "step": 6698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.502236868845773e-05,
+      "loss": 2.3532,
+      "step": 6699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.498482430068969e-05,
+      "loss": 1.2123,
+      "step": 6700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.494729739095545e-05,
+      "loss": 2.2084,
+      "step": 6701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4909787964957755e-05,
+      "loss": 2.4453,
+      "step": 6702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4872296028396594e-05,
+      "loss": 2.7734,
+      "step": 6703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.483482158696942e-05,
+      "loss": 2.3899,
+      "step": 6704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.479736464637087e-05,
+      "loss": 2.2888,
+      "step": 6705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.475992521229309e-05,
+      "loss": 2.2722,
+      "step": 6706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4722503290425405e-05,
+      "loss": 0.7705,
+      "step": 6707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.468509888645462e-05,
+      "loss": 2.323,
+      "step": 6708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4647712006064766e-05,
+      "loss": 2.2369,
+      "step": 6709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4610342654937325e-05,
+      "loss": 2.3049,
+      "step": 6710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4572990838750946e-05,
+      "loss": 2.282,
+      "step": 6711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.453565656318185e-05,
+      "loss": 2.7632,
+      "step": 6712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4498339833903335e-05,
+      "loss": 2.4101,
+      "step": 6713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.446104065658625e-05,
+      "loss": 2.2511,
+      "step": 6714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.442375903689861e-05,
+      "loss": 2.6302,
+      "step": 6715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.438649498050593e-05,
+      "loss": 2.2881,
+      "step": 6716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.434924849307084e-05,
+      "loss": 2.7332,
+      "step": 6717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.431201958025355e-05,
+      "loss": 2.6127,
+      "step": 6718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.427480824771136e-05,
+      "loss": 2.6692,
+      "step": 6719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.423761450109909e-05,
+      "loss": 2.6337,
+      "step": 6720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4200438346068785e-05,
+      "loss": 1.9342,
+      "step": 6721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.416327978826979e-05,
+      "loss": 2.4661,
+      "step": 6722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.412613883334891e-05,
+      "loss": 2.4522,
+      "step": 6723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4089015486950154e-05,
+      "loss": 2.3462,
+      "step": 6724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.405190975471483e-05,
+      "loss": 2.6545,
+      "step": 6725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4014821642281735e-05,
+      "loss": 1.9779,
+      "step": 6726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3977751155286785e-05,
+      "loss": 2.695,
+      "step": 6727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3940698299363425e-05,
+      "loss": 2.583,
+      "step": 6728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.390366308014221e-05,
+      "loss": 2.193,
+      "step": 6729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.386664550325119e-05,
+      "loss": 2.4888,
+      "step": 6730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.38296455743156e-05,
+      "loss": 2.397,
+      "step": 6731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.379266329895813e-05,
+      "loss": 2.8021,
+      "step": 6732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3755698682798626e-05,
+      "loss": 2.5627,
+      "step": 6733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.371875173145442e-05,
+      "loss": 2.493,
+      "step": 6734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3681822450539996e-05,
+      "loss": 2.6301,
+      "step": 6735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.364491084566732e-05,
+      "loss": 2.6696,
+      "step": 6736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.360801692244549e-05,
+      "loss": 2.2864,
+      "step": 6737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.35711406864811e-05,
+      "loss": 2.5458,
+      "step": 6738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.353428214337789e-05,
+      "loss": 2.3814,
+      "step": 6739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.349744129873707e-05,
+      "loss": 2.2078,
+      "step": 6740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.346061815815699e-05,
+      "loss": 2.3918,
+      "step": 6741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3423812727233475e-05,
+      "loss": 2.3865,
+      "step": 6742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.338702501155954e-05,
+      "loss": 2.5809,
+      "step": 6743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.335025501672559e-05,
+      "loss": 2.2149,
+      "step": 6744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.331350274831924e-05,
+      "loss": 2.1588,
+      "step": 6745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.327676821192553e-05,
+      "loss": 2.5988,
+      "step": 6746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.324005141312672e-05,
+      "loss": 2.2353,
+      "step": 6747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.320335235750237e-05,
+      "loss": 1.6124,
+      "step": 6748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3166671050629425e-05,
+      "loss": 2.7598,
+      "step": 6749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.313000749808202e-05,
+      "loss": 2.1169,
+      "step": 6750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.309336170543172e-05,
+      "loss": 2.7398,
+      "step": 6751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.305673367824727e-05,
+      "loss": 2.7215,
+      "step": 6752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.3020123422094816e-05,
+      "loss": 2.6441,
+      "step": 6753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.298353094253775e-05,
+      "loss": 2.6469,
+      "step": 6754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2946956245136696e-05,
+      "loss": 2.7785,
+      "step": 6755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.291039933544976e-05,
+      "loss": 2.1767,
+      "step": 6756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2873860219032164e-05,
+      "loss": 2.4288,
+      "step": 6757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2837338901436486e-05,
+      "loss": 2.5278,
+      "step": 6758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.280083538821267e-05,
+      "loss": 2.8509,
+      "step": 6759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.276434968490783e-05,
+      "loss": 1.1245,
+      "step": 6760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2727881797066484e-05,
+      "loss": 2.6524,
+      "step": 6761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.269143173023035e-05,
+      "loss": 2.2213,
+      "step": 6762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.265499948993856e-05,
+      "loss": 2.3208,
+      "step": 6763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2618585081727355e-05,
+      "loss": 2.4662,
+      "step": 6764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2582188511130494e-05,
+      "loss": 2.7512,
+      "step": 6765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2545809783678787e-05,
+      "loss": 2.0765,
+      "step": 6766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.250944890490054e-05,
+      "loss": 2.2474,
+      "step": 6767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.247310588032118e-05,
+      "loss": 2.5387,
+      "step": 6768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.243678071546359e-05,
+      "loss": 1.9648,
+      "step": 6769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.240047341584773e-05,
+      "loss": 2.7663,
+      "step": 6770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2364183986991086e-05,
+      "loss": 2.2317,
+      "step": 6771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.232791243440818e-05,
+      "loss": 2.9644,
+      "step": 6772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.229165876361104e-05,
+      "loss": 2.7288,
+      "step": 6773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.22554229801088e-05,
+      "loss": 2.5677,
+      "step": 6774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2219205089408026e-05,
+      "loss": 2.4259,
+      "step": 6775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.218300509701242e-05,
+      "loss": 2.6487,
+      "step": 6776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2146823008423104e-05,
+      "loss": 1.7546,
+      "step": 6777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2110658829138356e-05,
+      "loss": 1.9636,
+      "step": 6778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2074512564653825e-05,
+      "loss": 2.7429,
+      "step": 6779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2038384220462365e-05,
+      "loss": 2.197,
+      "step": 6780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.200227380205418e-05,
+      "loss": 1.1807,
+      "step": 6781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.196618131491671e-05,
+      "loss": 2.2756,
+      "step": 6782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1930106764534626e-05,
+      "loss": 2.2053,
+      "step": 6783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.189405015638996e-05,
+      "loss": 2.3811,
+      "step": 6784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1858011495962005e-05,
+      "loss": 2.7422,
+      "step": 6785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.182199078872719e-05,
+      "loss": 2.2408,
+      "step": 6786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.178598804015945e-05,
+      "loss": 0.9945,
+      "step": 6787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1750003255729764e-05,
+      "loss": 2.211,
+      "step": 6788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.171403644090657e-05,
+      "loss": 2.388,
+      "step": 6789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1678087601155405e-05,
+      "loss": 2.4301,
+      "step": 6790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.164215674193923e-05,
+      "loss": 1.9776,
+      "step": 6791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1606243868718156e-05,
+      "loss": 2.6236,
+      "step": 6792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.157034898694964e-05,
+      "loss": 2.3665,
+      "step": 6793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1534472102088305e-05,
+      "loss": 3.021,
+      "step": 6794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.149861321958619e-05,
+      "loss": 2.4105,
+      "step": 6795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.146277234489243e-05,
+      "loss": 2.4127,
+      "step": 6796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.14269494834536e-05,
+      "loss": 2.4433,
+      "step": 6797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.139114464071335e-05,
+      "loss": 2.1339,
+      "step": 6798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.135535782211277e-05,
+      "loss": 2.571,
+      "step": 6799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.131958903309003e-05,
+      "loss": 2.6657,
+      "step": 6800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.128383827908075e-05,
+      "loss": 2.6641,
+      "step": 6801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.124810556551765e-05,
+      "loss": 2.3555,
+      "step": 6802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.121239089783086e-05,
+      "loss": 2.1989,
+      "step": 6803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.117669428144757e-05,
+      "loss": 2.2758,
+      "step": 6804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1141015721792424e-05,
+      "loss": 2.1964,
+      "step": 6805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.110535522428719e-05,
+      "loss": 1.6769,
+      "step": 6806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.106971279435098e-05,
+      "loss": 1.5396,
+      "step": 6807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.103408843740007e-05,
+      "loss": 2.335,
+      "step": 6808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0998482158848096e-05,
+      "loss": 2.136,
+      "step": 6809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.096289396410587e-05,
+      "loss": 2.4395,
+      "step": 6810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.092732385858142e-05,
+      "loss": 2.4781,
+      "step": 6811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.089177184768017e-05,
+      "loss": 2.8134,
+      "step": 6812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0856237936804615e-05,
+      "loss": 1.7532,
+      "step": 6813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.082072213135468e-05,
+      "loss": 2.8308,
+      "step": 6814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.078522443672739e-05,
+      "loss": 1.7145,
+      "step": 6815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.074974485831706e-05,
+      "loss": 2.4015,
+      "step": 6816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.071428340151535e-05,
+      "loss": 2.8597,
+      "step": 6817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0678840071710966e-05,
+      "loss": 1.977,
+      "step": 6818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.06434148742901e-05,
+      "loss": 1.195,
+      "step": 6819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0608007814636e-05,
+      "loss": 2.8094,
+      "step": 6820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.057261889812919e-05,
+      "loss": 2.4653,
+      "step": 6821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0537248130147555e-05,
+      "loss": 2.5107,
+      "step": 6822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0501895516066044e-05,
+      "loss": 2.5047,
+      "step": 6823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.046656106125705e-05,
+      "loss": 2.3205,
+      "step": 6824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0431244771089997e-05,
+      "loss": 2.253,
+      "step": 6825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0395946650931725e-05,
+      "loss": 2.4168,
+      "step": 6826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.036066670614616e-05,
+      "loss": 2.3377,
+      "step": 6827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.032540494209463e-05,
+      "loss": 1.926,
+      "step": 6828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.029016136413553e-05,
+      "loss": 2.4041,
+      "step": 6829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.025493597762466e-05,
+      "loss": 3.018,
+      "step": 6830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0219728787914877e-05,
+      "loss": 2.4185,
+      "step": 6831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0184539800356443e-05,
+      "loss": 2.3301,
+      "step": 6832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.014936902029672e-05,
+      "loss": 2.1623,
+      "step": 6833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.011421645308043e-05,
+      "loss": 2.8355,
+      "step": 6834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0079082104049367e-05,
+      "loss": 2.454,
+      "step": 6835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.0043965978542744e-05,
+      "loss": 2.208,
+      "step": 6836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.000886808189681e-05,
+      "loss": 2.8992,
+      "step": 6837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9973788419445228e-05,
+      "loss": 1.9198,
+      "step": 6838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.993872699651872e-05,
+      "loss": 2.1637,
+      "step": 6839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9903683818445406e-05,
+      "loss": 2.0875,
+      "step": 6840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9868658890550464e-05,
+      "loss": 2.2975,
+      "step": 6841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9833652218156464e-05,
+      "loss": 2.1055,
+      "step": 6842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9798663806583084e-05,
+      "loss": 2.3662,
+      "step": 6843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.976369366114721e-05,
+      "loss": 2.3156,
+      "step": 6844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.97287417871631e-05,
+      "loss": 2.7117,
+      "step": 6845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96938081899421e-05,
+      "loss": 2.0557,
+      "step": 6846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.965889287479275e-05,
+      "loss": 2.3694,
+      "step": 6847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9623995847020997e-05,
+      "loss": 2.5813,
+      "step": 6848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.95891171119298e-05,
+      "loss": 2.5519,
+      "step": 6849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.95542566748195e-05,
+      "loss": 2.6539,
+      "step": 6850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9519414540987508e-05,
+      "loss": 2.4974,
+      "step": 6851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.948459071572864e-05,
+      "loss": 2.3238,
+      "step": 6852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.944978520433473e-05,
+      "loss": 2.7183,
+      "step": 6853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.941499801209498e-05,
+      "loss": 2.0064,
+      "step": 6854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9380229144295715e-05,
+      "loss": 2.4314,
+      "step": 6855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.934547860622055e-05,
+      "loss": 2.2654,
+      "step": 6856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.931074640315022e-05,
+      "loss": 2.5429,
+      "step": 6857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9276032540362798e-05,
+      "loss": 2.8386,
+      "step": 6858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.924133702313343e-05,
+      "loss": 2.5944,
+      "step": 6859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9206659856734623e-05,
+      "loss": 2.3927,
+      "step": 6860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9172001046435932e-05,
+      "loss": 2.3498,
+      "step": 6861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.913736059750429e-05,
+      "loss": 2.279,
+      "step": 6862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9102738515203693e-05,
+      "loss": 2.1942,
+      "step": 6863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9068134804795473e-05,
+      "loss": 2.6728,
+      "step": 6864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9033549471538037e-05,
+      "loss": 2.7289,
+      "step": 6865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.899898252068713e-05,
+      "loss": 2.5757,
+      "step": 6866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8964433957495602e-05,
+      "loss": 2.5045,
+      "step": 6867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.892990378721362e-05,
+      "loss": 2.8327,
+      "step": 6868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8895392015088387e-05,
+      "loss": 2.2375,
+      "step": 6869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8860898646364504e-05,
+      "loss": 2.1801,
+      "step": 6870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8826423686283633e-05,
+      "loss": 2.4767,
+      "step": 6871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8791967140084665e-05,
+      "loss": 1.6883,
+      "step": 6872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8757529013003778e-05,
+      "loss": 2.554,
+      "step": 6873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8723109310274268e-05,
+      "loss": 2.4369,
+      "step": 6874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8688708037126583e-05,
+      "loss": 1.6112,
+      "step": 6875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.865432519878853e-05,
+      "loss": 2.5007,
+      "step": 6876
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8619960800484957e-05,
+      "loss": 2.4298,
+      "step": 6877
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8585614847438044e-05,
+      "loss": 2.4823,
+      "step": 6878
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8551287344867008e-05,
+      "loss": 2.5428,
+      "step": 6879
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8516978297988462e-05,
+      "loss": 2.5466,
+      "step": 6880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8482687712016006e-05,
+      "loss": 2.7698,
+      "step": 6881
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8448415592160616e-05,
+      "loss": 2.2113,
+      "step": 6882
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.84141619436303e-05,
+      "loss": 2.3958,
+      "step": 6883
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8379926771630434e-05,
+      "loss": 2.6956,
+      "step": 6884
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8345710081363416e-05,
+      "loss": 2.216,
+      "step": 6885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8311511878028964e-05,
+      "loss": 2.0712,
+      "step": 6886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8277332166823873e-05,
+      "loss": 2.283,
+      "step": 6887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8243170952942256e-05,
+      "loss": 2.1866,
+      "step": 6888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8209028241575294e-05,
+      "loss": 3.1118,
+      "step": 6889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8174904037911468e-05,
+      "loss": 2.6423,
+      "step": 6890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8140798347136363e-05,
+      "loss": 2.6999,
+      "step": 6891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8106711174432734e-05,
+      "loss": 2.2581,
+      "step": 6892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.807264252498062e-05,
+      "loss": 2.1751,
+      "step": 6893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.803859240395715e-05,
+      "loss": 2.1532,
+      "step": 6894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8004560816536732e-05,
+      "loss": 2.4281,
+      "step": 6895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7970547767890845e-05,
+      "loss": 2.5277,
+      "step": 6896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7936553263188255e-05,
+      "loss": 1.4489,
+      "step": 6897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7902577307594797e-05,
+      "loss": 2.8071,
+      "step": 6898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.786861990627366e-05,
+      "loss": 2.4728,
+      "step": 6899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7834681064384995e-05,
+      "loss": 2.7574,
+      "step": 6900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7800760787086346e-05,
+      "loss": 1.8835,
+      "step": 6901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7766859079532238e-05,
+      "loss": 2.648,
+      "step": 6902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7732975946874552e-05,
+      "loss": 2.8233,
+      "step": 6903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7699111394262253e-05,
+      "loss": 1.5589,
+      "step": 6904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7665265426841422e-05,
+      "loss": 2.5924,
+      "step": 6905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.763143804975548e-05,
+      "loss": 2.3252,
+      "step": 6906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7597629268144896e-05,
+      "loss": 2.4387,
+      "step": 6907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7563839087147295e-05,
+      "loss": 2.1236,
+      "step": 6908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7530067511897635e-05,
+      "loss": 2.4614,
+      "step": 6909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.749631454752783e-05,
+      "loss": 2.5223,
+      "step": 6910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7462580199167196e-05,
+      "loss": 2.1846,
+      "step": 6911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7428864471941973e-05,
+      "loss": 2.3904,
+      "step": 6912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7395167370975807e-05,
+      "loss": 1.8743,
+      "step": 6913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7361488901389328e-05,
+      "loss": 2.5377,
+      "step": 6914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7327829068300494e-05,
+      "loss": 2.4085,
+      "step": 6915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7294187876824246e-05,
+      "loss": 1.7516,
+      "step": 6916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.72605653320729e-05,
+      "loss": 1.4734,
+      "step": 6917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7226961439155726e-05,
+      "loss": 2.5566,
+      "step": 6918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.719337620317938e-05,
+      "loss": 2.8302,
+      "step": 6919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7159809629247476e-05,
+      "loss": 2.3024,
+      "step": 6920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.712626172246095e-05,
+      "loss": 2.7464,
+      "step": 6921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7092732487917795e-05,
+      "loss": 2.5962,
+      "step": 6922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7059221930713247e-05,
+      "loss": 2.26,
+      "step": 6923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.702573005593962e-05,
+      "loss": 2.402,
+      "step": 6924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.699225686868649e-05,
+      "loss": 2.7266,
+      "step": 6925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.695880237404049e-05,
+      "loss": 2.776,
+      "step": 6926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.69253665770855e-05,
+      "loss": 1.8676,
+      "step": 6927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6891949482902458e-05,
+      "loss": 2.113,
+      "step": 6928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6858551096569603e-05,
+      "loss": 2.6171,
+      "step": 6929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6825171423162156e-05,
+      "loss": 1.6048,
+      "step": 6930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6791810467752695e-05,
+      "loss": 2.5464,
+      "step": 6931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.675846823541078e-05,
+      "loss": 2.5085,
+      "step": 6932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6725144731203157e-05,
+      "loss": 2.2085,
+      "step": 6933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6691839960193846e-05,
+      "loss": 1.8149,
+      "step": 6934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6658553927443893e-05,
+      "loss": 2.2478,
+      "step": 6935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6625286638011507e-05,
+      "loss": 2.0152,
+      "step": 6936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.659203809695214e-05,
+      "loss": 2.4609,
+      "step": 6937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6558808309318268e-05,
+      "loss": 2.6478,
+      "step": 6938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6525597280159655e-05,
+      "loss": 2.1339,
+      "step": 6939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6492405014523092e-05,
+      "loss": 1.9375,
+      "step": 6940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6459231517452618e-05,
+      "loss": 2.3931,
+      "step": 6941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6426076793989293e-05,
+      "loss": 2.0827,
+      "step": 6942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6392940849171508e-05,
+      "loss": 2.493,
+      "step": 6943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6359823688034598e-05,
+      "loss": 2.7741,
+      "step": 6944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6326725315611224e-05,
+      "loss": 2.5407,
+      "step": 6945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.629364573693103e-05,
+      "loss": 2.2204,
+      "step": 6946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6260584957020958e-05,
+      "loss": 2.2843,
+      "step": 6947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6227542980904943e-05,
+      "loss": 2.4332,
+      "step": 6948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.619451981360423e-05,
+      "loss": 2.888,
+      "step": 6949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6161515460137013e-05,
+      "loss": 2.6035,
+      "step": 6950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6128529925518805e-05,
+      "loss": 1.8768,
+      "step": 6951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6095563214762134e-05,
+      "loss": 2.345,
+      "step": 6952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6062615332876756e-05,
+      "loss": 2.7689,
+      "step": 6953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6029686284869473e-05,
+      "loss": 2.4774,
+      "step": 6954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.599677607574434e-05,
+      "loss": 2.2977,
+      "step": 6955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5963884710502414e-05,
+      "loss": 2.3655,
+      "step": 6956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.593101219414204e-05,
+      "loss": 2.6995,
+      "step": 6957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5898158531658546e-05,
+      "loss": 2.7901,
+      "step": 6958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5865323728044534e-05,
+      "loss": 2.2286,
+      "step": 6959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5832507788289613e-05,
+      "loss": 2.0124,
+      "step": 6960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5799710717380643e-05,
+      "loss": 2.7163,
+      "step": 6961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5766932520301524e-05,
+      "loss": 1.9954,
+      "step": 6962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.573417320203332e-05,
+      "loss": 2.2541,
+      "step": 6963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5701432767554265e-05,
+      "loss": 2.2534,
+      "step": 6964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.566871122183966e-05,
+      "loss": 2.3947,
+      "step": 6965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.563600856986195e-05,
+      "loss": 2.1521,
+      "step": 6966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.560332481659077e-05,
+      "loss": 2.6366,
+      "step": 6967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.557065996699282e-05,
+      "loss": 2.3201,
+      "step": 6968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5538014026031887e-05,
+      "loss": 2.7861,
+      "step": 6969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.550538699866903e-05,
+      "loss": 2.4565,
+      "step": 6970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.547277888986226e-05,
+      "loss": 2.4952,
+      "step": 6971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5440189704566882e-05,
+      "loss": 2.4057,
+      "step": 6972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5407619447735167e-05,
+      "loss": 2.7698,
+      "step": 6973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5375068124316657e-05,
+      "loss": 1.5319,
+      "step": 6974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5342535739257873e-05,
+      "loss": 2.4071,
+      "step": 6975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5310022297502586e-05,
+      "loss": 2.3822,
+      "step": 6976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.527752780399157e-05,
+      "loss": 2.3964,
+      "step": 6977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5245052263662863e-05,
+      "loss": 2.6754,
+      "step": 6978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5212595681451446e-05,
+      "loss": 2.5151,
+      "step": 6979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5180158062289612e-05,
+      "loss": 2.5541,
+      "step": 6980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5147739411106584e-05,
+      "loss": 2.6346,
+      "step": 6981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5115339732828864e-05,
+      "loss": 2.3713,
+      "step": 6982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5082959032379923e-05,
+      "loss": 2.6665,
+      "step": 6983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.505059731468052e-05,
+      "loss": 2.2561,
+      "step": 6984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5018254584648356e-05,
+      "loss": 1.9056,
+      "step": 6985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4985930847198365e-05,
+      "loss": 2.4888,
+      "step": 6986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.495362610724252e-05,
+      "loss": 2.7297,
+      "step": 6987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4921340369690007e-05,
+      "loss": 2.613,
+      "step": 6988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4889073639446956e-05,
+      "loss": 2.3208,
+      "step": 6989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4856825921416823e-05,
+      "loss": 2.589,
+      "step": 6990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.482459722049996e-05,
+      "loss": 2.0888,
+      "step": 6991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4792387541594026e-05,
+      "loss": 2.5422,
+      "step": 6992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4760196889593666e-05,
+      "loss": 3.0577,
+      "step": 6993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.472802526939059e-05,
+      "loss": 2.2283,
+      "step": 6994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.469587268587379e-05,
+      "loss": 2.3657,
+      "step": 6995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.466373914392923e-05,
+      "loss": 2.611,
+      "step": 6996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4631624648439963e-05,
+      "loss": 2.3796,
+      "step": 6997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4599529204286288e-05,
+      "loss": 2.0568,
+      "step": 6998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.456745281634542e-05,
+      "loss": 2.4683,
+      "step": 6999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4535395489491882e-05,
+      "loss": 2.0809,
+      "step": 7000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.450335722859712e-05,
+      "loss": 2.737,
+      "step": 7001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.447133803852981e-05,
+      "loss": 2.3121,
+      "step": 7002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4439337924155622e-05,
+      "loss": 2.4435,
+      "step": 7003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4407356890337472e-05,
+      "loss": 2.8995,
+      "step": 7004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4375394941935183e-05,
+      "loss": 2.3224,
+      "step": 7005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4343452083805877e-05,
+      "loss": 2.2366,
+      "step": 7006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4311528320803614e-05,
+      "loss": 2.3705,
+      "step": 7007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4279623657779668e-05,
+      "loss": 2.6636,
+      "step": 7008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4247738099582325e-05,
+      "loss": 2.5624,
+      "step": 7009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4215871651057045e-05,
+      "loss": 2.1785,
+      "step": 7010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4184024317046285e-05,
+      "loss": 2.1762,
+      "step": 7011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4152196102389744e-05,
+      "loss": 2.2735,
+      "step": 7012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.412038701192403e-05,
+      "loss": 2.4941,
+      "step": 7013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4088597050483032e-05,
+      "loss": 2.5258,
+      "step": 7014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4056826222897568e-05,
+      "loss": 3.0232,
+      "step": 7015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4025074533995676e-05,
+      "loss": 2.5031,
+      "step": 7016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3993341988602393e-05,
+      "loss": 2.6389,
+      "step": 7017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.396162859153995e-05,
+      "loss": 2.4827,
+      "step": 7018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.392993434762752e-05,
+      "loss": 2.8577,
+      "step": 7019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.389825926168154e-05,
+      "loss": 2.6713,
+      "step": 7020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3866603338515356e-05,
+      "loss": 2.5138,
+      "step": 7021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.383496658293957e-05,
+      "loss": 2.7235,
+      "step": 7022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3803348999761767e-05,
+      "loss": 2.2877,
+      "step": 7023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.377175059378659e-05,
+      "loss": 2.1443,
+      "step": 7024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.374017136981592e-05,
+      "loss": 2.1618,
+      "step": 7025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3708611332648573e-05,
+      "loss": 2.2389,
+      "step": 7026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3677070487080474e-05,
+      "loss": 2.755,
+      "step": 7027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.364554883790473e-05,
+      "loss": 2.9082,
+      "step": 7028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3614046389911384e-05,
+      "loss": 2.1199,
+      "step": 7029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3582563147887716e-05,
+      "loss": 2.3737,
+      "step": 7030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3551099116617924e-05,
+      "loss": 2.3218,
+      "step": 7031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3519654300883455e-05,
+      "loss": 2.5771,
+      "step": 7032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.348822870546268e-05,
+      "loss": 2.1112,
+      "step": 7033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.345682233513121e-05,
+      "loss": 1.9491,
+      "step": 7034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.342543519466154e-05,
+      "loss": 2.0976,
+      "step": 7035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3394067288823442e-05,
+      "loss": 1.7987,
+      "step": 7036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3362718622383636e-05,
+      "loss": 2.44,
+      "step": 7037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.333138920010591e-05,
+      "loss": 1.9865,
+      "step": 7038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3300079026751255e-05,
+      "loss": 2.469,
+      "step": 7039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3268788107077586e-05,
+      "loss": 2.0729,
+      "step": 7040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3237516445840026e-05,
+      "loss": 2.1905,
+      "step": 7041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3206264047790612e-05,
+      "loss": 2.5021,
+      "step": 7042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.317503091767866e-05,
+      "loss": 2.5171,
+      "step": 7043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.314381706025034e-05,
+      "loss": 1.9795,
+      "step": 7044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3112622480249094e-05,
+      "loss": 1.8585,
+      "step": 7045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.308144718241524e-05,
+      "loss": 2.3884,
+      "step": 7046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.305029117148636e-05,
+      "loss": 2.5728,
+      "step": 7047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3019154452196937e-05,
+      "loss": 2.8683,
+      "step": 7048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.298803702927866e-05,
+      "loss": 2.6659,
+      "step": 7049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2956938907460153e-05,
+      "loss": 2.3764,
+      "step": 7050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2925860091467253e-05,
+      "loss": 3.3824,
+      "step": 7051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2894800586022692e-05,
+      "loss": 2.312,
+      "step": 7052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2863760395846452e-05,
+      "loss": 1.6894,
+      "step": 7053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.283273952565546e-05,
+      "loss": 2.3621,
+      "step": 7054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.280173798016367e-05,
+      "loss": 2.3291,
+      "step": 7055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.277075576408227e-05,
+      "loss": 2.8111,
+      "step": 7056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2739792882119346e-05,
+      "loss": 2.2506,
+      "step": 7057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2708849338980077e-05,
+      "loss": 1.8549,
+      "step": 7058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.267792513936681e-05,
+      "loss": 1.7843,
+      "step": 7059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.264702028797881e-05,
+      "loss": 1.5042,
+      "step": 7060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2616134789512513e-05,
+      "loss": 2.4013,
+      "step": 7061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.258526864866131e-05,
+      "loss": 2.6295,
+      "step": 7062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.255442187011578e-05,
+      "loss": 2.9696,
+      "step": 7063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.252359445856341e-05,
+      "loss": 2.3207,
+      "step": 7064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2492786418688902e-05,
+      "loss": 2.5844,
+      "step": 7065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2461997755173862e-05,
+      "loss": 2.2803,
+      "step": 7066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2431228472697076e-05,
+      "loss": 2.4081,
+      "step": 7067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.240047857593428e-05,
+      "loss": 2.2999,
+      "step": 7068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.236974806955839e-05,
+      "loss": 2.2973,
+      "step": 7069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2339036958239214e-05,
+      "loss": 2.9786,
+      "step": 7070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2308345246643767e-05,
+      "loss": 2.2232,
+      "step": 7071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2277672939435997e-05,
+      "loss": 2.3855,
+      "step": 7072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.224702004127703e-05,
+      "loss": 2.6251,
+      "step": 7073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.221638655682487e-05,
+      "loss": 2.6212,
+      "step": 7074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2185772490734766e-05,
+      "loss": 1.8469,
+      "step": 7075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2155177847658838e-05,
+      "loss": 2.5179,
+      "step": 7076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2124602632246392e-05,
+      "loss": 2.5818,
+      "step": 7077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2094046849143687e-05,
+      "loss": 2.1295,
+      "step": 7078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2063510502994097e-05,
+      "loss": 2.1026,
+      "step": 7079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2032993598437974e-05,
+      "loss": 2.5848,
+      "step": 7080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2002496140112796e-05,
+      "loss": 2.3367,
+      "step": 7081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1972018132653e-05,
+      "loss": 2.4535,
+      "step": 7082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1941559580690187e-05,
+      "loss": 2.5138,
+      "step": 7083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.191112048885285e-05,
+      "loss": 1.8338,
+      "step": 7084
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1880700861766593e-05,
+      "loss": 2.7675,
+      "step": 7085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1850300704054148e-05,
+      "loss": 2.5901,
+      "step": 7086
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1819920020335153e-05,
+      "loss": 2.7725,
+      "step": 7087
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1789558815226315e-05,
+      "loss": 2.7247,
+      "step": 7088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.175921709334148e-05,
+      "loss": 2.1771,
+      "step": 7089
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.172889485929139e-05,
+      "loss": 2.5124,
+      "step": 7090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1698592117683977e-05,
+      "loss": 2.2725,
+      "step": 7091
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1668308873124075e-05,
+      "loss": 2.1863,
+      "step": 7092
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.163804513021363e-05,
+      "loss": 2.5511,
+      "step": 7093
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.160780089355159e-05,
+      "loss": 2.3662,
+      "step": 7094
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1577576167734007e-05,
+      "loss": 2.6758,
+      "step": 7095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1547370957353836e-05,
+      "loss": 2.6311,
+      "step": 7096
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1517185267001248e-05,
+      "loss": 2.5104,
+      "step": 7097
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.148701910126323e-05,
+      "loss": 2.4976,
+      "step": 7098
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1456872464724025e-05,
+      "loss": 2.3162,
+      "step": 7099
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.142674536196471e-05,
+      "loss": 2.0096,
+      "step": 7100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1396637797563576e-05,
+      "loss": 2.6899,
+      "step": 7101
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1366549776095764e-05,
+      "loss": 2.3963,
+      "step": 7102
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1336481302133633e-05,
+      "loss": 2.2462,
+      "step": 7103
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1306432380246362e-05,
+      "loss": 1.7301,
+      "step": 7104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.127640301500038e-05,
+      "loss": 2.336,
+      "step": 7105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1246393210958957e-05,
+      "loss": 2.1379,
+      "step": 7106
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1216402972682524e-05,
+      "loss": 2.5561,
+      "step": 7107
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.118643230472842e-05,
+      "loss": 2.5885,
+      "step": 7108
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1156481211651132e-05,
+      "loss": 1.1664,
+      "step": 7109
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1126549698002094e-05,
+      "loss": 2.1764,
+      "step": 7110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1096637768329744e-05,
+      "loss": 2.6076,
+      "step": 7111
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1066745427179667e-05,
+      "loss": 2.4815,
+      "step": 7112
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.103687267909431e-05,
+      "loss": 2.2164,
+      "step": 7113
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.100701952861327e-05,
+      "loss": 2.3097,
+      "step": 7114
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.09771859802731e-05,
+      "loss": 2.2104,
+      "step": 7115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0947372038607357e-05,
+      "loss": 2.3759,
+      "step": 7116
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0917577708146725e-05,
+      "loss": 1.9499,
+      "step": 7117
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.08878029934188e-05,
+      "loss": 2.2754,
+      "step": 7118
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0858047898948194e-05,
+      "loss": 2.9283,
+      "step": 7119
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0828312429256638e-05,
+      "loss": 2.0876,
+      "step": 7120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0798596588862753e-05,
+      "loss": 2.0541,
+      "step": 7121
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0768900382282327e-05,
+      "loss": 2.6642,
+      "step": 7122
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0739223814027982e-05,
+      "loss": 1.169,
+      "step": 7123
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0709566888609545e-05,
+      "loss": 2.2133,
+      "step": 7124
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0679929610533713e-05,
+      "loss": 2.6891,
+      "step": 7125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.065031198430428e-05,
+      "loss": 2.1647,
+      "step": 7126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0620714014421978e-05,
+      "loss": 2.2894,
+      "step": 7127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.059113570538468e-05,
+      "loss": 1.0987,
+      "step": 7128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.056157706168709e-05,
+      "loss": 2.3064,
+      "step": 7129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0532038087821128e-05,
+      "loss": 2.406,
+      "step": 7130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0502518788275523e-05,
+      "loss": 2.707,
+      "step": 7131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0473019167536185e-05,
+      "loss": 2.6515,
+      "step": 7132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0443539230085914e-05,
+      "loss": 2.1682,
+      "step": 7133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.041407898040461e-05,
+      "loss": 2.76,
+      "step": 7134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.038463842296907e-05,
+      "loss": 2.2899,
+      "step": 7135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.035521756225325e-05,
+      "loss": 2.1064,
+      "step": 7136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0325816402727945e-05,
+      "loss": 2.7903,
+      "step": 7137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.029643494886111e-05,
+      "loss": 2.3952,
+      "step": 7138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0267073205117564e-05,
+      "loss": 2.4566,
+      "step": 7139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0237731175959276e-05,
+      "loss": 2.3303,
+      "step": 7140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.020840886584507e-05,
+      "loss": 2.6213,
+      "step": 7141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0179106279230934e-05,
+      "loss": 2.1901,
+      "step": 7142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0149823420569667e-05,
+      "loss": 2.3326,
+      "step": 7143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0120560294311276e-05,
+      "loss": 2.7215,
+      "step": 7144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0091316904902627e-05,
+      "loss": 2.8725,
+      "step": 7145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0062093256787592e-05,
+      "loss": 1.1841,
+      "step": 7146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0032889354407156e-05,
+      "loss": 2.2415,
+      "step": 7147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0003705202199182e-05,
+      "loss": 2.4192,
+      "step": 7148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9974540804598544e-05,
+      "loss": 2.2048,
+      "step": 7149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9945396166037207e-05,
+      "loss": 2.5445,
+      "step": 7150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9916271290944035e-05,
+      "loss": 2.2209,
+      "step": 7151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.988716618374497e-05,
+      "loss": 2.6295,
+      "step": 7152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9858080848862834e-05,
+      "loss": 2.4363,
+      "step": 7153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9829015290717615e-05,
+      "loss": 2.5181,
+      "step": 7154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9799969513726094e-05,
+      "loss": 1.9313,
+      "step": 7155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.977094352230224e-05,
+      "loss": 2.601,
+      "step": 7156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.974193732085686e-05,
+      "loss": 2.2297,
+      "step": 7157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9712950913797876e-05,
+      "loss": 2.1634,
+      "step": 7158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683984305530083e-05,
+      "loss": 2.4069,
+      "step": 7159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655037500455406e-05,
+      "loss": 2.7034,
+      "step": 7160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962611050297258e-05,
+      "loss": 2.5191,
+      "step": 7161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9597203317477555e-05,
+      "loss": 2.6281,
+      "step": 7162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9568315948363033e-05,
+      "loss": 2.1765,
+      "step": 7163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.953944840001892e-05,
+      "loss": 2.4574,
+      "step": 7164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.951060067683192e-05,
+      "loss": 2.5746,
+      "step": 7165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9481772783185918e-05,
+      "loss": 2.344,
+      "step": 7166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9452964723461567e-05,
+      "loss": 2.6315,
+      "step": 7167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9424176502036737e-05,
+      "loss": 2.5144,
+      "step": 7168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.939540812328606e-05,
+      "loss": 2.6616,
+      "step": 7169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9366659591581356e-05,
+      "loss": 2.1878,
+      "step": 7170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9337930911291267e-05,
+      "loss": 1.7431,
+      "step": 7171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930922208678153e-05,
+      "loss": 2.2731,
+      "step": 7172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9280533122414804e-05,
+      "loss": 2.7453,
+      "step": 7173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9251864022550707e-05,
+      "loss": 2.8126,
+      "step": 7174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9223214791545934e-05,
+      "loss": 2.1627,
+      "step": 7175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.91945854337541e-05,
+      "loss": 1.5155,
+      "step": 7176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.916597595352573e-05,
+      "loss": 2.2305,
+      "step": 7177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.913738635520851e-05,
+      "loss": 2.4619,
+      "step": 7178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9108816643146902e-05,
+      "loss": 2.5572,
+      "step": 7179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.908026682168251e-05,
+      "loss": 2.0342,
+      "step": 7180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9051736895153785e-05,
+      "loss": 2.6872,
+      "step": 7181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.902322686789629e-05,
+      "loss": 2.3117,
+      "step": 7182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8994736744242444e-05,
+      "loss": 2.4108,
+      "step": 7183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8966266528521666e-05,
+      "loss": 1.8712,
+      "step": 7184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8937816225060416e-05,
+      "loss": 2.7304,
+      "step": 7185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8909385838182044e-05,
+      "loss": 2.4338,
+      "step": 7186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8880975372206968e-05,
+      "loss": 2.1599,
+      "step": 7187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.885258483145245e-05,
+      "loss": 2.7723,
+      "step": 7188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8824214220232886e-05,
+      "loss": 2.576,
+      "step": 7189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8795863542859445e-05,
+      "loss": 1.9811,
+      "step": 7190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.87675328036405e-05,
+      "loss": 2.6424,
+      "step": 7191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8739222006881155e-05,
+      "loss": 2.2785,
+      "step": 7192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8710931156883684e-05,
+      "loss": 2.514,
+      "step": 7193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8682660257947184e-05,
+      "loss": 2.259,
+      "step": 7194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8654409314367852e-05,
+      "loss": 2.4828,
+      "step": 7195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.86261783304387e-05,
+      "loss": 2.5282,
+      "step": 7196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8597967310449868e-05,
+      "loss": 2.8712,
+      "step": 7197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8569776258688308e-05,
+      "loss": 2.3942,
+      "step": 7198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8541605179438074e-05,
+      "loss": 2.8526,
+      "step": 7199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8513454076980078e-05,
+      "loss": 2.2049,
+      "step": 7200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8485322955592296e-05,
+      "loss": 2.1737,
+      "step": 7201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8457211819549544e-05,
+      "loss": 2.5228,
+      "step": 7202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8429120673123738e-05,
+      "loss": 2.6613,
+      "step": 7203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.840104952058361e-05,
+      "loss": 2.1678,
+      "step": 7204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8372998366195036e-05,
+      "loss": 2.1677,
+      "step": 7205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8344967214220668e-05,
+      "loss": 2.4689,
+      "step": 7206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8316956068920202e-05,
+      "loss": 2.7928,
+      "step": 7207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.828896493455035e-05,
+      "loss": 2.2264,
+      "step": 7208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8260993815364687e-05,
+      "loss": 2.099,
+      "step": 7209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8233042715613754e-05,
+      "loss": 2.4711,
+      "step": 7210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8205111639545135e-05,
+      "loss": 2.3969,
+      "step": 7211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8177200591403262e-05,
+      "loss": 2.4352,
+      "step": 7212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8149309575429648e-05,
+      "loss": 2.5211,
+      "step": 7213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.812143859586262e-05,
+      "loss": 2.3105,
+      "step": 7214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8093587656937585e-05,
+      "loss": 2.8828,
+      "step": 7215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8065756762886796e-05,
+      "loss": 2.3044,
+      "step": 7216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8037945917939585e-05,
+      "loss": 2.6132,
+      "step": 7217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8010155126322085e-05,
+      "loss": 2.2857,
+      "step": 7218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7982384392257538e-05,
+      "loss": 2.7834,
+      "step": 7219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7954633719966017e-05,
+      "loss": 2.3525,
+      "step": 7220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7926903113664633e-05,
+      "loss": 1.3272,
+      "step": 7221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7899192577567363e-05,
+      "loss": 2.4373,
+      "step": 7222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7871502115885222e-05,
+      "loss": 2.1388,
+      "step": 7223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.784383173282608e-05,
+      "loss": 2.1532,
+      "step": 7224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7816181432594873e-05,
+      "loss": 2.1561,
+      "step": 7225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7788551219393355e-05,
+      "loss": 2.4018,
+      "step": 7226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7760941097420332e-05,
+      "loss": 2.4667,
+      "step": 7227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7733351070871506e-05,
+      "loss": 2.2693,
+      "step": 7228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7705781143939545e-05,
+      "loss": 2.4815,
+      "step": 7229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7678231320814025e-05,
+      "loss": 2.7624,
+      "step": 7230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7650701605681537e-05,
+      "loss": 2.54,
+      "step": 7231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7623192002725527e-05,
+      "loss": 1.1977,
+      "step": 7232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7595702516126496e-05,
+      "loss": 2.8215,
+      "step": 7233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7568233150061772e-05,
+      "loss": 2.4158,
+      "step": 7234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.754078390870568e-05,
+      "loss": 2.3607,
+      "step": 7235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.751335479622952e-05,
+      "loss": 1.9894,
+      "step": 7236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7485945816801482e-05,
+      "loss": 2.7933,
+      "step": 7237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7458556974586662e-05,
+      "loss": 2.7417,
+      "step": 7238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7431188273747243e-05,
+      "loss": 2.4311,
+      "step": 7239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.740383971844215e-05,
+      "loss": 2.3136,
+      "step": 7240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7376511312827433e-05,
+      "loss": 2.3218,
+      "step": 7241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7349203061055932e-05,
+      "loss": 2.6603,
+      "step": 7242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7321914967277557e-05,
+      "loss": 2.8085,
+      "step": 7243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7294647035638986e-05,
+      "loss": 2.0467,
+      "step": 7244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7267399270284026e-05,
+      "loss": 1.9803,
+      "step": 7245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7240171675353266e-05,
+      "loss": 2.835,
+      "step": 7246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7212964254984325e-05,
+      "loss": 2.2531,
+      "step": 7247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7185777013311663e-05,
+      "loss": 2.2314,
+      "step": 7248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.715860995446683e-05,
+      "loss": 2.369,
+      "step": 7249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.71314630825781e-05,
+      "loss": 2.2841,
+      "step": 7250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7104336401770864e-05,
+      "loss": 2.4947,
+      "step": 7251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.7077229916167335e-05,
+      "loss": 2.7624,
+      "step": 7252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.705014362988672e-05,
+      "loss": 2.3551,
+      "step": 7253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.702307754704511e-05,
+      "loss": 2.8705,
+      "step": 7254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6996031671755518e-05,
+      "loss": 2.4944,
+      "step": 7255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.696900600812797e-05,
+      "loss": 2.3829,
+      "step": 7256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6942000560269307e-05,
+      "loss": 2.4641,
+      "step": 7257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6915015332283393e-05,
+      "loss": 2.6032,
+      "step": 7258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.688805032827094e-05,
+      "loss": 1.5871,
+      "step": 7259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.686110555232969e-05,
+      "loss": 2.1864,
+      "step": 7260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6834181008554183e-05,
+      "loss": 1.1369,
+      "step": 7261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6807276701035992e-05,
+      "loss": 2.447,
+      "step": 7262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6780392633863538e-05,
+      "loss": 2.2361,
+      "step": 7263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6753528811122248e-05,
+      "loss": 2.5042,
+      "step": 7264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6726685236894372e-05,
+      "loss": 2.43,
+      "step": 7265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6699861915259184e-05,
+      "loss": 0.6763,
+      "step": 7266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6673058850292804e-05,
+      "loss": 2.2582,
+      "step": 7267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6646276046068273e-05,
+      "loss": 0.007,
+      "step": 7268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.661951350665566e-05,
+      "loss": 2.0574,
+      "step": 7269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6592771236121834e-05,
+      "loss": 2.2309,
+      "step": 7270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.656604923853058e-05,
+      "loss": 2.072,
+      "step": 7271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6539347517942742e-05,
+      "loss": 2.1711,
+      "step": 7272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6512666078415897e-05,
+      "loss": 2.7937,
+      "step": 7273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6486004924004724e-05,
+      "loss": 2.9217,
+      "step": 7274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6459364058760632e-05,
+      "loss": 2.2733,
+      "step": 7275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.643274348673215e-05,
+      "loss": 2.3247,
+      "step": 7276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6406143211964503e-05,
+      "loss": 1.9915,
+      "step": 7277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6379563238500043e-05,
+      "loss": 2.222,
+      "step": 7278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.635300357037786e-05,
+      "loss": 2.8661,
+      "step": 7279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6326464211634087e-05,
+      "loss": 2.4959,
+      "step": 7280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6299945166301705e-05,
+      "loss": 2.4472,
+      "step": 7281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.627344643841063e-05,
+      "loss": 2.0119,
+      "step": 7282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.624696803198767e-05,
+      "loss": 2.6469,
+      "step": 7283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6220509951056575e-05,
+      "loss": 2.3645,
+      "step": 7284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6194072199637976e-05,
+      "loss": 2.267,
+      "step": 7285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.616765478174945e-05,
+      "loss": 2.4902,
+      "step": 7286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6141257701405425e-05,
+      "loss": 2.7721,
+      "step": 7287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6114880962617338e-05,
+      "loss": 2.7999,
+      "step": 7288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6088524569393402e-05,
+      "loss": 2.7677,
+      "step": 7289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.606218852573886e-05,
+      "loss": 2.4109,
+      "step": 7290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6035872835655784e-05,
+      "loss": 0.9357,
+      "step": 7291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.6009577503143216e-05,
+      "loss": 2.4885,
+      "step": 7292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5983302532197017e-05,
+      "loss": 2.5987,
+      "step": 7293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5957047926810073e-05,
+      "loss": 2.6964,
+      "step": 7294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5930813690972062e-05,
+      "loss": 2.7614,
+      "step": 7295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5904599828669606e-05,
+      "loss": 2.3505,
+      "step": 7296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5878406343886273e-05,
+      "loss": 2.3341,
+      "step": 7297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5852233240602498e-05,
+      "loss": 2.4461,
+      "step": 7298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5826080522795558e-05,
+      "loss": 2.2502,
+      "step": 7299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5799948194439766e-05,
+      "loss": 2.5023,
+      "step": 7300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5773836259506222e-05,
+      "loss": 2.1946,
+      "step": 7301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5747744721963022e-05,
+      "loss": 2.6035,
+      "step": 7302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5721673585775045e-05,
+      "loss": 2.8645,
+      "step": 7303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.569562285490419e-05,
+      "loss": 2.748,
+      "step": 7304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.566959253330915e-05,
+      "loss": 2.4036,
+      "step": 7305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5643582624945616e-05,
+      "loss": 2.5459,
+      "step": 7306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5617593133766094e-05,
+      "loss": 2.3386,
+      "step": 7307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5591624063720053e-05,
+      "loss": 2.2691,
+      "step": 7308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5565675418753777e-05,
+      "loss": 2.4295,
+      "step": 7309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5539747202810555e-05,
+      "loss": 2.7015,
+      "step": 7310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5513839419830454e-05,
+      "loss": 2.6759,
+      "step": 7311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5487952073750542e-05,
+      "loss": 1.26,
+      "step": 7312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5462085168504707e-05,
+      "loss": 2.3491,
+      "step": 7313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5436238708023785e-05,
+      "loss": 2.3056,
+      "step": 7314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5410412696235418e-05,
+      "loss": 2.8425,
+      "step": 7315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5384607137064287e-05,
+      "loss": 1.8135,
+      "step": 7316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.53588220344318e-05,
+      "loss": 2.8481,
+      "step": 7317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5333057392256394e-05,
+      "loss": 2.1477,
+      "step": 7318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5307313214453283e-05,
+      "loss": 2.3592,
+      "step": 7319
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.528158950493469e-05,
+      "loss": 2.2711,
+      "step": 7320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5255886267609584e-05,
+      "loss": 2.0703,
+      "step": 7321
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.523020350638398e-05,
+      "loss": 2.4021,
+      "step": 7322
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5204541225160672e-05,
+      "loss": 2.5606,
+      "step": 7323
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5178899427839335e-05,
+      "loss": 2.4354,
+      "step": 7324
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5153278118316641e-05,
+      "loss": 2.4903,
+      "step": 7325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5127677300486024e-05,
+      "loss": 2.5726,
+      "step": 7326
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5102096978237854e-05,
+      "loss": 2.2525,
+      "step": 7327
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.507653715545944e-05,
+      "loss": 2.4909,
+      "step": 7328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5050997836034872e-05,
+      "loss": 2.0401,
+      "step": 7329
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5025479023845215e-05,
+      "loss": 2.6988,
+      "step": 7330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4999980722768385e-05,
+      "loss": 2.5257,
+      "step": 7331
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4974502936679112e-05,
+      "loss": 1.917,
+      "step": 7332
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.494904566944915e-05,
+      "loss": 2.2072,
+      "step": 7333
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4923608924947006e-05,
+      "loss": 2.0311,
+      "step": 7334
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.489819270703817e-05,
+      "loss": 2.7395,
+      "step": 7335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4872797019584904e-05,
+      "loss": 2.2528,
+      "step": 7336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4847421866446451e-05,
+      "loss": 2.4399,
+      "step": 7337
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4822067251478864e-05,
+      "loss": 2.0747,
+      "step": 7338
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4796733178535142e-05,
+      "loss": 2.5212,
+      "step": 7339
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4771419651465072e-05,
+      "loss": 2.7204,
+      "step": 7340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4746126674115427e-05,
+      "loss": 2.4489,
+      "step": 7341
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.472085425032975e-05,
+      "loss": 2.2767,
+      "step": 7342
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4695602383948535e-05,
+      "loss": 2.8824,
+      "step": 7343
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4670371078809113e-05,
+      "loss": 2.5136,
+      "step": 7344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4645160338745732e-05,
+      "loss": 1.7921,
+      "step": 7345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4619970167589445e-05,
+      "loss": 2.0757,
+      "step": 7346
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4594800569168274e-05,
+      "loss": 2.2071,
+      "step": 7347
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.456965154730701e-05,
+      "loss": 2.568,
+      "step": 7348
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4544523105827426e-05,
+      "loss": 2.4953,
+      "step": 7349
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4519415248548055e-05,
+      "loss": 1.4449,
+      "step": 7350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.449432797928441e-05,
+      "loss": 2.0149,
+      "step": 7351
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4469261301848772e-05,
+      "loss": 2.1204,
+      "step": 7352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4444215220050399e-05,
+      "loss": 2.7496,
+      "step": 7353
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4419189737695308e-05,
+      "loss": 2.2001,
+      "step": 7354
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.439418485858651e-05,
+      "loss": 1.5836,
+      "step": 7355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4369200586523766e-05,
+      "loss": 2.2143,
+      "step": 7356
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4344236925303754e-05,
+      "loss": 2.2361,
+      "step": 7357
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4319293878720056e-05,
+      "loss": 2.6416,
+      "step": 7358
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4294371450563092e-05,
+      "loss": 2.0326,
+      "step": 7359
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4269469644620079e-05,
+      "loss": 2.2828,
+      "step": 7360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4244588464675261e-05,
+      "loss": 2.2557,
+      "step": 7361
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.421972791450956e-05,
+      "loss": 2.2283,
+      "step": 7362
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.419488799790094e-05,
+      "loss": 2.3606,
+      "step": 7363
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4170068718624078e-05,
+      "loss": 2.0339,
+      "step": 7364
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4145270080450643e-05,
+      "loss": 1.716,
+      "step": 7365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4120492087149038e-05,
+      "loss": 1.7349,
+      "step": 7366
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4095734742484666e-05,
+      "loss": 2.2815,
+      "step": 7367
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4070998050219668e-05,
+      "loss": 2.4924,
+      "step": 7368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4046282014113153e-05,
+      "loss": 2.6465,
+      "step": 7369
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4021586637920983e-05,
+      "loss": 1.8414,
+      "step": 7370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3996911925395994e-05,
+      "loss": 2.3596,
+      "step": 7371
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.397225788028778e-05,
+      "loss": 3.0392,
+      "step": 7372
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.394762450634288e-05,
+      "loss": 2.3389,
+      "step": 7373
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3923011807304602e-05,
+      "loss": 2.0142,
+      "step": 7374
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3898419786913235e-05,
+      "loss": 2.2513,
+      "step": 7375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3873848448905772e-05,
+      "loss": 2.8238,
+      "step": 7376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3849297797016212e-05,
+      "loss": 2.5879,
+      "step": 7377
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.382476783497527e-05,
+      "loss": 2.7783,
+      "step": 7378
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3800258566510664e-05,
+      "loss": 2.3408,
+      "step": 7379
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.377576999534683e-05,
+      "loss": 2.5673,
+      "step": 7380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3751302125205194e-05,
+      "loss": 2.5392,
+      "step": 7381
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.372685495980388e-05,
+      "loss": 2.1692,
+      "step": 7382
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3702428502857998e-05,
+      "loss": 2.3548,
+      "step": 7383
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3678022758079465e-05,
+      "loss": 1.75,
+      "step": 7384
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3653637729177014e-05,
+      "loss": 2.7791,
+      "step": 7385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3629273419856296e-05,
+      "loss": 2.4264,
+      "step": 7386
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3604929833819784e-05,
+      "loss": 2.6183,
+      "step": 7387
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3580606974766734e-05,
+      "loss": 2.5725,
+      "step": 7388
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3556304846393407e-05,
+      "loss": 2.2777,
+      "step": 7389
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3532023452392748e-05,
+      "loss": 1.7073,
+      "step": 7390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3507762796454708e-05,
+      "loss": 2.0224,
+      "step": 7391
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3483522882265902e-05,
+      "loss": 2.4153,
+      "step": 7392
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.345930371351e-05,
+      "loss": 2.7485,
+      "step": 7393
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3435105293867344e-05,
+      "loss": 2.2143,
+      "step": 7394
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3410927627015239e-05,
+      "loss": 2.5007,
+      "step": 7395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3386770716627748e-05,
+      "loss": 1.8498,
+      "step": 7396
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3362634566375867e-05,
+      "loss": 1.8288,
+      "step": 7397
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3338519179927365e-05,
+      "loss": 2.4006,
+      "step": 7398
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3314424560946907e-05,
+      "loss": 2.0033,
+      "step": 7399
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3290350713095982e-05,
+      "loss": 2.1455,
+      "step": 7400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3266297640032864e-05,
+      "loss": 2.4533,
+      "step": 7401
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3242265345412812e-05,
+      "loss": 2.4446,
+      "step": 7402
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3218253832887771e-05,
+      "loss": 2.9904,
+      "step": 7403
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3194263106106639e-05,
+      "loss": 2.2711,
+      "step": 7404
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3170293168715101e-05,
+      "loss": 2.4933,
+      "step": 7405
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.314634402435571e-05,
+      "loss": 2.24,
+      "step": 7406
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3122415676667819e-05,
+      "loss": 2.3923,
+      "step": 7407
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3098508129287687e-05,
+      "loss": 2.5476,
+      "step": 7408
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3074621385848326e-05,
+      "loss": 3.0452,
+      "step": 7409
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3050755449979683e-05,
+      "loss": 2.3691,
+      "step": 7410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3026910325308438e-05,
+      "loss": 2.8241,
+      "step": 7411
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3003086015458225e-05,
+      "loss": 3.0005,
+      "step": 7412
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.29792825240494e-05,
+      "loss": 2.3667,
+      "step": 7413
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.295549985469927e-05,
+      "loss": 2.2246,
+      "step": 7414
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2931738011021841e-05,
+      "loss": 2.2845,
+      "step": 7415
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2907996996628095e-05,
+      "loss": 3.0171,
+      "step": 7416
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.288427681512576e-05,
+      "loss": 3.0117,
+      "step": 7417
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2860577470119404e-05,
+      "loss": 2.5221,
+      "step": 7418
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2836898965210496e-05,
+      "loss": 1.9051,
+      "step": 7419
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.281324130399724e-05,
+      "loss": 2.5752,
+      "step": 7420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2789604490074728e-05,
+      "loss": 2.5708,
+      "step": 7421
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2765988527034904e-05,
+      "loss": 2.5503,
+      "step": 7422
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2742393418466495e-05,
+      "loss": 2.4984,
+      "step": 7423
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2718819167955101e-05,
+      "loss": 2.3993,
+      "step": 7424
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2695265779083091e-05,
+      "loss": 2.4069,
+      "step": 7425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2671733255429766e-05,
+      "loss": 2.3651,
+      "step": 7426
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2648221600571135e-05,
+      "loss": 2.4581,
+      "step": 7427
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2624730818080136e-05,
+      "loss": 2.1977,
+      "step": 7428
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2601260911526468e-05,
+      "loss": 2.447,
+      "step": 7429
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.257781188447674e-05,
+      "loss": 2.2345,
+      "step": 7430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2554383740494256e-05,
+      "loss": 2.4382,
+      "step": 7431
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2530976483139282e-05,
+      "loss": 2.3286,
+      "step": 7432
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2507590115968807e-05,
+      "loss": 2.3543,
+      "step": 7433
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.248422464253674e-05,
+      "loss": 2.7099,
+      "step": 7434
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2460880066393703e-05,
+      "loss": 2.4773,
+      "step": 7435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2437556391087277e-05,
+      "loss": 2.4977,
+      "step": 7436
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2414253620161729e-05,
+      "loss": 2.5753,
+      "step": 7437
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2390971757158275e-05,
+      "loss": 2.4363,
+      "step": 7438
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2367710805614817e-05,
+      "loss": 2.7426,
+      "step": 7439
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2344470769066229e-05,
+      "loss": 1.6562,
+      "step": 7440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2321251651044083e-05,
+      "loss": 2.0269,
+      "step": 7441
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2298053455076862e-05,
+      "loss": 2.46,
+      "step": 7442
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.227487618468979e-05,
+      "loss": 1.6846,
+      "step": 7443
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2251719843405005e-05,
+      "loss": 2.3872,
+      "step": 7444
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2228584434741356e-05,
+      "loss": 2.0659,
+      "step": 7445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2205469962214582e-05,
+      "loss": 2.5297,
+      "step": 7446
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2182376429337236e-05,
+      "loss": 2.5386,
+      "step": 7447
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2159303839618684e-05,
+      "loss": 2.8215,
+      "step": 7448
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2136252196565049e-05,
+      "loss": 2.4373,
+      "step": 7449
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.21132215036794e-05,
+      "loss": 2.7501,
+      "step": 7450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2090211764461482e-05,
+      "loss": 1.7233,
+      "step": 7451
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2067222982407971e-05,
+      "loss": 1.0426,
+      "step": 7452
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2044255161012267e-05,
+      "loss": 2.2988,
+      "step": 7453
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2021308303764648e-05,
+      "loss": 1.0444,
+      "step": 7454
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1998382414152175e-05,
+      "loss": 2.1226,
+      "step": 7455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1975477495658759e-05,
+      "loss": 2.4442,
+      "step": 7456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.195259355176505e-05,
+      "loss": 2.3001,
+      "step": 7457
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.19297305859486e-05,
+      "loss": 1.6624,
+      "step": 7458
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1906888601683694e-05,
+      "loss": 2.5947,
+      "step": 7459
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1884067602441511e-05,
+      "loss": 2.6361,
+      "step": 7460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1861267591689921e-05,
+      "loss": 2.3429,
+      "step": 7461
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1838488572893772e-05,
+      "loss": 2.0671,
+      "step": 7462
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1815730549514545e-05,
+      "loss": 1.8534,
+      "step": 7463
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1792993525010675e-05,
+      "loss": 2.632,
+      "step": 7464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1770277502837294e-05,
+      "loss": 2.3324,
+      "step": 7465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1747582486446461e-05,
+      "loss": 2.4358,
+      "step": 7466
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1724908479286899e-05,
+      "loss": 2.0833,
+      "step": 7467
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1702255484804268e-05,
+      "loss": 2.1394,
+      "step": 7468
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1679623506440933e-05,
+      "loss": 2.7506,
+      "step": 7469
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.165701254763619e-05,
+      "loss": 1.7621,
+      "step": 7470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1634422611825977e-05,
+      "loss": 2.3007,
+      "step": 7471
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1611853702443196e-05,
+      "loss": 2.2853,
+      "step": 7472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1589305822917455e-05,
+      "loss": 2.4658,
+      "step": 7473
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.156677897667518e-05,
+      "loss": 2.6855,
+      "step": 7474
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1544273167139628e-05,
+      "loss": 2.2231,
+      "step": 7475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1521788397730835e-05,
+      "loss": 2.5249,
+      "step": 7476
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1499324671865701e-05,
+      "loss": 2.0601,
+      "step": 7477
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.147688199295781e-05,
+      "loss": 2.5342,
+      "step": 7478
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1454460364417634e-05,
+      "loss": 2.1606,
+      "step": 7479
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1432059789652465e-05,
+      "loss": 1.4228,
+      "step": 7480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1409680272066312e-05,
+      "loss": 2.9237,
+      "step": 7481
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1387321815060041e-05,
+      "loss": 2.0503,
+      "step": 7482
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1364984422031332e-05,
+      "loss": 2.3861,
+      "step": 7483
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1342668096374602e-05,
+      "loss": 2.4873,
+      "step": 7484
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.132037284148114e-05,
+      "loss": 1.974,
+      "step": 7485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1298098660738952e-05,
+      "loss": 2.2571,
+      "step": 7486
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1275845557532947e-05,
+      "loss": 2.1573,
+      "step": 7487
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1253613535244704e-05,
+      "loss": 2.5695,
+      "step": 7488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.123140259725272e-05,
+      "loss": 2.6696,
+      "step": 7489
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1209212746932195e-05,
+      "loss": 2.0859,
+      "step": 7490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1187043987655198e-05,
+      "loss": 2.5536,
+      "step": 7491
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1164896322790502e-05,
+      "loss": 2.2051,
+      "step": 7492
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.114276975570378e-05,
+      "loss": 1.9511,
+      "step": 7493
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1120664289757425e-05,
+      "loss": 2.666,
+      "step": 7494
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.109857992831067e-05,
+      "loss": 2.7342,
+      "step": 7495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.107651667471946e-05,
+      "loss": 2.0744,
+      "step": 7496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1054474532336666e-05,
+      "loss": 2.7707,
+      "step": 7497
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1032453504511807e-05,
+      "loss": 2.5736,
+      "step": 7498
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1010453594591324e-05,
+      "loss": 2.3201,
+      "step": 7499
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0988474805918324e-05,
+      "loss": 2.4243,
+      "step": 7500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.096651714183282e-05,
+      "loss": 2.8417,
+      "step": 7501
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0944580605671526e-05,
+      "loss": 2.5556,
+      "step": 7502
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0922665200767993e-05,
+      "loss": 2.3518,
+      "step": 7503
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.090077093045254e-05,
+      "loss": 2.4125,
+      "step": 7504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0878897798052322e-05,
+      "loss": 2.1248,
+      "step": 7505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0857045806891196e-05,
+      "loss": 2.3963,
+      "step": 7506
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.083521496028984e-05,
+      "loss": 2.195,
+      "step": 7507
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.08134052615658e-05,
+      "loss": 2.5188,
+      "step": 7508
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0791616714033274e-05,
+      "loss": 1.3541,
+      "step": 7509
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0769849321003333e-05,
+      "loss": 2.5556,
+      "step": 7510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0748103085783827e-05,
+      "loss": 2.4432,
+      "step": 7511
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0726378011679349e-05,
+      "loss": 2.3204,
+      "step": 7512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.070467410199134e-05,
+      "loss": 1.9566,
+      "step": 7513
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0682991360017945e-05,
+      "loss": 2.1044,
+      "step": 7514
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0661329789054179e-05,
+      "loss": 2.8164,
+      "step": 7515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0639689392391759e-05,
+      "loss": 2.2187,
+      "step": 7516
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0618070173319254e-05,
+      "loss": 2.5122,
+      "step": 7517
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0596472135121936e-05,
+      "loss": 2.5428,
+      "step": 7518
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0574895281081963e-05,
+      "loss": 2.4658,
+      "step": 7519
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0553339614478162e-05,
+      "loss": 1.9115,
+      "step": 7520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.053180513858623e-05,
+      "loss": 2.5488,
+      "step": 7521
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0510291856678583e-05,
+      "loss": 2.4259,
+      "step": 7522
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.048879977202447e-05,
+      "loss": 2.3903,
+      "step": 7523
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0467328887889848e-05,
+      "loss": 2.6026,
+      "step": 7524
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0445879207537538e-05,
+      "loss": 2.1781,
+      "step": 7525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0424450734227052e-05,
+      "loss": 2.7208,
+      "step": 7526
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0403043471214767e-05,
+      "loss": 2.3346,
+      "step": 7527
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0381657421753747e-05,
+      "loss": 2.201,
+      "step": 7528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.036029258909391e-05,
+      "loss": 2.0985,
+      "step": 7529
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0338948976481892e-05,
+      "loss": 2.4943,
+      "step": 7530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0317626587161166e-05,
+      "loss": 1.9153,
+      "step": 7531
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0296325424371887e-05,
+      "loss": 2.6631,
+      "step": 7532
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0275045491351103e-05,
+      "loss": 2.7561,
+      "step": 7533
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0253786791332525e-05,
+      "loss": 2.254,
+      "step": 7534
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0232549327546719e-05,
+      "loss": 2.2788,
+      "step": 7535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0211333103220987e-05,
+      "loss": 2.1014,
+      "step": 7536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.019013812157935e-05,
+      "loss": 3.0353,
+      "step": 7537
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0168964385842732e-05,
+      "loss": 2.1705,
+      "step": 7538
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0147811899228742e-05,
+      "loss": 2.1532,
+      "step": 7539
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0126680664951709e-05,
+      "loss": 2.4996,
+      "step": 7540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.010557068622288e-05,
+      "loss": 2.691,
+      "step": 7541
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0084481966250124e-05,
+      "loss": 2.483,
+      "step": 7542
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0063414508238177e-05,
+      "loss": 2.2744,
+      "step": 7543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0042368315388494e-05,
+      "loss": 2.4516,
+      "step": 7544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0021343390899333e-05,
+      "loss": 2.4218,
+      "step": 7545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0000339737965707e-05,
+      "loss": 2.2476,
+      "step": 7546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.979357359779327e-06,
+      "loss": 2.7166,
+      "step": 7547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.958396259528828e-06,
+      "loss": 2.1274,
+      "step": 7548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.937456440399427e-06,
+      "loss": 2.7283,
+      "step": 7549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.91653790557328e-06,
+      "loss": 2.2631,
+      "step": 7550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.895640658229176e-06,
+      "loss": 2.0315,
+      "step": 7551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.874764701542742e-06,
+      "loss": 2.4121,
+      "step": 7552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.853910038686325e-06,
+      "loss": 2.1916,
+      "step": 7553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833076672829089e-06,
+      "loss": 2.0726,
+      "step": 7554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812264607136915e-06,
+      "loss": 2.0821,
+      "step": 7555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.79147384477248e-06,
+      "loss": 2.2474,
+      "step": 7556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.770704388895168e-06,
+      "loss": 2.2419,
+      "step": 7557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.749956242661205e-06,
+      "loss": 1.7203,
+      "step": 7558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729229409223505e-06,
+      "loss": 2.8693,
+      "step": 7559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708523891731828e-06,
+      "loss": 1.3894,
+      "step": 7560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687839693332594e-06,
+      "loss": 2.9266,
+      "step": 7561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.667176817169053e-06,
+      "loss": 2.351,
+      "step": 7562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.646535266381178e-06,
+      "loss": 2.659,
+      "step": 7563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.62591504410576e-06,
+      "loss": 1.9793,
+      "step": 7564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.605316153476256e-06,
+      "loss": 2.3437,
+      "step": 7565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.584738597622965e-06,
+      "loss": 2.4494,
+      "step": 7566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.564182379672903e-06,
+      "loss": 2.6268,
+      "step": 7567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.543647502749834e-06,
+      "loss": 2.4324,
+      "step": 7568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.523133969974334e-06,
+      "loss": 2.2419,
+      "step": 7569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.502641784463672e-06,
+      "loss": 2.4855,
+      "step": 7570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.482170949331874e-06,
+      "loss": 2.3954,
+      "step": 7571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.461721467689803e-06,
+      "loss": 2.3665,
+      "step": 7572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.441293342644956e-06,
+      "loss": 2.3993,
+      "step": 7573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.420886577301734e-06,
+      "loss": 1.4985,
+      "step": 7574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.400501174761122e-06,
+      "loss": 2.375,
+      "step": 7575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.380137138120992e-06,
+      "loss": 1.2473,
+      "step": 7576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.3597944704759e-06,
+      "loss": 2.2771,
+      "step": 7577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.339473174917195e-06,
+      "loss": 2.3347,
+      "step": 7578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.319173254532936e-06,
+      "loss": 2.277,
+      "step": 7579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.298894712407989e-06,
+      "loss": 2.2885,
+      "step": 7580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.278637551623907e-06,
+      "loss": 2.3886,
+      "step": 7581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.258401775259045e-06,
+      "loss": 2.3887,
+      "step": 7582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.238187386388474e-06,
+      "loss": 2.3454,
+      "step": 7583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.217994388084055e-06,
+      "loss": 2.1586,
+      "step": 7584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.197822783414333e-06,
+      "loss": 2.0318,
+      "step": 7585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.17767257544469e-06,
+      "loss": 2.4581,
+      "step": 7586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.157543767237157e-06,
+      "loss": 2.6411,
+      "step": 7587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.137436361850625e-06,
+      "loss": 2.637,
+      "step": 7588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.117350362340613e-06,
+      "loss": 2.1595,
+      "step": 7589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.0972857717595e-06,
+      "loss": 2.0345,
+      "step": 7590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.077242593156326e-06,
+      "loss": 2.6502,
+      "step": 7591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.057220829576923e-06,
+      "loss": 2.1336,
+      "step": 7592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.037220484063824e-06,
+      "loss": 1.1572,
+      "step": 7593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.017241559656396e-06,
+      "loss": 2.3961,
+      "step": 7594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.99728405939063e-06,
+      "loss": 2.3963,
+      "step": 7595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.977347986299399e-06,
+      "loss": 2.3744,
+      "step": 7596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.957433343412178e-06,
+      "loss": 2.4003,
+      "step": 7597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.937540133755267e-06,
+      "loss": 2.3362,
+      "step": 7598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.917668360351726e-06,
+      "loss": 2.4525,
+      "step": 7599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.89781802622131e-06,
+      "loss": 2.1502,
+      "step": 7600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.877989134380486e-06,
+      "loss": 2.361,
+      "step": 7601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.858181687842597e-06,
+      "loss": 1.8664,
+      "step": 7602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.83839568961755e-06,
+      "loss": 2.7035,
+      "step": 7603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.818631142712157e-06,
+      "loss": 2.357,
+      "step": 7604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.798888050129848e-06,
+      "loss": 2.4818,
+      "step": 7605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.779166414870858e-06,
+      "loss": 2.5327,
+      "step": 7606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.759466239932123e-06,
+      "loss": 2.4435,
+      "step": 7607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.73978752830738e-06,
+      "loss": 2.0588,
+      "step": 7608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.72013028298702e-06,
+      "loss": 2.3101,
+      "step": 7609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.70049450695824e-06,
+      "loss": 2.1145,
+      "step": 7610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.680880203204916e-06,
+      "loss": 2.0726,
+      "step": 7611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.661287374707732e-06,
+      "loss": 2.1915,
+      "step": 7612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.64171602444404e-06,
+      "loss": 2.4406,
+      "step": 7613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.622166155387993e-06,
+      "loss": 2.5007,
+      "step": 7614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.602637770510401e-06,
+      "loss": 1.8372,
+      "step": 7615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.583130872778905e-06,
+      "loss": 2.6703,
+      "step": 7616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.56364546515778e-06,
+      "loss": 2.3147,
+      "step": 7617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.544181550608097e-06,
+      "loss": 2.395,
+      "step": 7618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.524739132087671e-06,
+      "loss": 1.2589,
+      "step": 7619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.505318212550987e-06,
+      "loss": 1.86,
+      "step": 7620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.485918794949354e-06,
+      "loss": 2.2249,
+      "step": 7621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.46654088223071e-06,
+      "loss": 1.9726,
+      "step": 7622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.447184477339836e-06,
+      "loss": 1.3857,
+      "step": 7623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.42784958321811e-06,
+      "loss": 2.2159,
+      "step": 7624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.4085362028038e-06,
+      "loss": 2.5185,
+      "step": 7625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.389244339031759e-06,
+      "loss": 2.1695,
+      "step": 7626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.369973994833673e-06,
+      "loss": 2.2468,
+      "step": 7627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.350725173137901e-06,
+      "loss": 2.5516,
+      "step": 7628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.331497876869537e-06,
+      "loss": 2.8958,
+      "step": 7629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.312292108950458e-06,
+      "loss": 1.2138,
+      "step": 7630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.2931078722992e-06,
+      "loss": 2.5114,
+      "step": 7631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.273945169831026e-06,
+      "loss": 2.6536,
+      "step": 7632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.254804004457994e-06,
+      "loss": 2.298,
+      "step": 7633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.235684379088826e-06,
+      "loss": 2.561,
+      "step": 7634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.21658629662903e-06,
+      "loss": 2.3826,
+      "step": 7635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.19750975998077e-06,
+      "loss": 1.6908,
+      "step": 7636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.178454772042992e-06,
+      "loss": 2.3503,
+      "step": 7637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.159421335711313e-06,
+      "loss": 2.6783,
+      "step": 7638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.140409453878155e-06,
+      "loss": 2.9319,
+      "step": 7639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.121419129432587e-06,
+      "loss": 1.9289,
+      "step": 7640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.102450365260455e-06,
+      "loss": 2.2281,
+      "step": 7641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.083503164244265e-06,
+      "loss": 2.6086,
+      "step": 7642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.06457752926335e-06,
+      "loss": 1.3282,
+      "step": 7643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.045673463193641e-06,
+      "loss": 2.3323,
+      "step": 7644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.026790968907892e-06,
+      "loss": 2.8649,
+      "step": 7645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.007930049275524e-06,
+      "loss": 2.2821,
+      "step": 7646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.989090707162726e-06,
+      "loss": 1.5101,
+      "step": 7647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.970272945432327e-06,
+      "loss": 2.118,
+      "step": 7648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.95147676694397e-06,
+      "loss": 2.6938,
+      "step": 7649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.932702174553939e-06,
+      "loss": 2.2666,
+      "step": 7650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.913949171115336e-06,
+      "loss": 2.4929,
+      "step": 7651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.895217759477845e-06,
+      "loss": 1.8608,
+      "step": 7652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.876507942487991e-06,
+      "loss": 2.2432,
+      "step": 7653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.857819722988967e-06,
+      "loss": 2.105,
+      "step": 7654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.839153103820684e-06,
+      "loss": 1.8111,
+      "step": 7655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.820508087819755e-06,
+      "loss": 2.1907,
+      "step": 7656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.801884677819581e-06,
+      "loss": 2.3691,
+      "step": 7657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.783282876650182e-06,
+      "loss": 2.3009,
+      "step": 7658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.76470268713833e-06,
+      "loss": 2.1289,
+      "step": 7659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.746144112107583e-06,
+      "loss": 2.586,
+      "step": 7660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.727607154378102e-06,
+      "loss": 2.4981,
+      "step": 7661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.709091816766833e-06,
+      "loss": 2.9212,
+      "step": 7662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.690598102087425e-06,
+      "loss": 2.8791,
+      "step": 7663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.672126013150215e-06,
+      "loss": 1.9546,
+      "step": 7664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.653675552762306e-06,
+      "loss": 2.2015,
+      "step": 7665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.635246723727439e-06,
+      "loss": 2.3156,
+      "step": 7666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.616839528846175e-06,
+      "loss": 2.1661,
+      "step": 7667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.598453970915658e-06,
+      "loss": 2.2878,
+      "step": 7668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.580090052729854e-06,
+      "loss": 2.5027,
+      "step": 7669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.5617477770793635e-06,
+      "loss": 2.4789,
+      "step": 7670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.543427146751574e-06,
+      "loss": 1.6894,
+      "step": 7671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.525128164530492e-06,
+      "loss": 2.2013,
+      "step": 7672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.506850833196943e-06,
+      "loss": 2.3409,
+      "step": 7673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.488595155528321e-06,
+      "loss": 1.9773,
+      "step": 7674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.470361134298891e-06,
+      "loss": 2.5511,
+      "step": 7675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.452148772279487e-06,
+      "loss": 2.5165,
+      "step": 7676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.433958072237744e-06,
+      "loss": 2.4672,
+      "step": 7677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.415789036937969e-06,
+      "loss": 2.281,
+      "step": 7678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.397641669141169e-06,
+      "loss": 1.6545,
+      "step": 7679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.379515971605071e-06,
+      "loss": 2.1713,
+      "step": 7680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.361411947084139e-06,
+      "loss": 1.9397,
+      "step": 7681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.343329598329456e-06,
+      "loss": 2.5686,
+      "step": 7682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.325268928088923e-06,
+      "loss": 2.3521,
+      "step": 7683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.307229939107046e-06,
+      "loss": 2.7921,
+      "step": 7684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.289212634125113e-06,
+      "loss": 2.5645,
+      "step": 7685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.2712170158810705e-06,
+      "loss": 2.2611,
+      "step": 7686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.2532430871095785e-06,
+      "loss": 2.6529,
+      "step": 7687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.235290850542019e-06,
+      "loss": 3.3748,
+      "step": 7688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.217360308906478e-06,
+      "loss": 1.2162,
+      "step": 7689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.199451464927674e-06,
+      "loss": 2.7888,
+      "step": 7690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.181564321327165e-06,
+      "loss": 2.579,
+      "step": 7691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.163698880823077e-06,
+      "loss": 2.0683,
+      "step": 7692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.145855146130303e-06,
+      "loss": 1.6202,
+      "step": 7693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.128033119960441e-06,
+      "loss": 1.9638,
+      "step": 7694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.110232805021759e-06,
+      "loss": 2.0808,
+      "step": 7695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.0924542040192765e-06,
+      "loss": 0.9935,
+      "step": 7696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.074697319654632e-06,
+      "loss": 2.0827,
+      "step": 7697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.056962154626267e-06,
+      "loss": 2.2171,
+      "step": 7698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.039248711629209e-06,
+      "loss": 2.003,
+      "step": 7699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.021556993355287e-06,
+      "loss": 2.0018,
+      "step": 7700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.00388700249297e-06,
+      "loss": 1.525,
+      "step": 7701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.986238741727456e-06,
+      "loss": 2.5226,
+      "step": 7702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.968612213740588e-06,
+      "loss": 2.5695,
+      "step": 7703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.9510074212109855e-06,
+      "loss": 2.2968,
+      "step": 7704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.933424366813878e-06,
+      "loss": 2.8616,
+      "step": 7705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.915863053221293e-06,
+      "loss": 2.1628,
+      "step": 7706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.8983234831018306e-06,
+      "loss": 2.6825,
+      "step": 7707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.880805659120925e-06,
+      "loss": 2.6093,
+      "step": 7708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.863309583940579e-06,
+      "loss": 2.5044,
+      "step": 7709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.845835260219584e-06,
+      "loss": 2.1994,
+      "step": 7710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.828382690613349e-06,
+      "loss": 2.163,
+      "step": 7711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.810951877774051e-06,
+      "loss": 2.4902,
+      "step": 7712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.793542824350506e-06,
+      "loss": 2.8278,
+      "step": 7713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.776155532988264e-06,
+      "loss": 2.1533,
+      "step": 7714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.758790006329512e-06,
+      "loss": 2.2805,
+      "step": 7715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.7414462470132214e-06,
+      "loss": 2.2555,
+      "step": 7716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.7241242576749175e-06,
+      "loss": 1.4832,
+      "step": 7717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.706824040946995e-06,
+      "loss": 2.1026,
+      "step": 7718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.689545599458368e-06,
+      "loss": 2.5214,
+      "step": 7719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.672288935834752e-06,
+      "loss": 2.4018,
+      "step": 7720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.6550540526985165e-06,
+      "loss": 2.1483,
+      "step": 7721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.637840952668716e-06,
+      "loss": 2.6783,
+      "step": 7722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.6206496383610915e-06,
+      "loss": 2.4161,
+      "step": 7723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.603480112388104e-06,
+      "loss": 2.5609,
+      "step": 7724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.586332377358866e-06,
+      "loss": 2.6217,
+      "step": 7725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.5692064358792264e-06,
+      "loss": 2.5019,
+      "step": 7726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.55210229055167e-06,
+      "loss": 2.1933,
+      "step": 7727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.5350199439754e-06,
+      "loss": 2.1894,
+      "step": 7728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.5179593987462585e-06,
+      "loss": 2.3572,
+      "step": 7729
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.500920657456888e-06,
+      "loss": 2.3364,
+      "step": 7730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.483903722696482e-06,
+      "loss": 2.4117,
+      "step": 7731
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.466908597051024e-06,
+      "loss": 2.1902,
+      "step": 7732
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.449935283103097e-06,
+      "loss": 2.6655,
+      "step": 7733
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.432983783432072e-06,
+      "loss": 1.9209,
+      "step": 7734
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.416054100613887e-06,
+      "loss": 2.7309,
+      "step": 7735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.39914623722127e-06,
+      "loss": 2.272,
+      "step": 7736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.382260195823563e-06,
+      "loss": 2.2043,
+      "step": 7737
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.365395978986848e-06,
+      "loss": 1.6402,
+      "step": 7738
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.348553589273808e-06,
+      "loss": 2.0966,
+      "step": 7739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.331733029243913e-06,
+      "loss": 2.7657,
+      "step": 7740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.314934301453217e-06,
+      "loss": 2.4231,
+      "step": 7741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.2981574084545295e-06,
+      "loss": 2.501,
+      "step": 7742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.28140235279731e-06,
+      "loss": 2.5651,
+      "step": 7743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.264669137027722e-06,
+      "loss": 2.4751,
+      "step": 7744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.247957763688566e-06,
+      "loss": 2.1554,
+      "step": 7745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.2312682353193586e-06,
+      "loss": 2.8964,
+      "step": 7746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.214600554456306e-06,
+      "loss": 1.4579,
+      "step": 7747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.197954723632231e-06,
+      "loss": 1.8985,
+      "step": 7748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.181330745376728e-06,
+      "loss": 2.4396,
+      "step": 7749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.164728622216009e-06,
+      "loss": 2.3649,
+      "step": 7750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.148148356672955e-06,
+      "loss": 3.1744,
+      "step": 7751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.131589951267202e-06,
+      "loss": 2.6927,
+      "step": 7752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.1150534085149525e-06,
+      "loss": 2.0804,
+      "step": 7753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.0985387309292144e-06,
+      "loss": 2.259,
+      "step": 7754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.08204592101953e-06,
+      "loss": 2.4879,
+      "step": 7755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.065574981292259e-06,
+      "loss": 2.3391,
+      "step": 7756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.049125914250319e-06,
+      "loss": 2.2315,
+      "step": 7757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.032698722393409e-06,
+      "loss": 2.1593,
+      "step": 7758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.0162934082177986e-06,
+      "loss": 2.6184,
+      "step": 7759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.999909974216544e-06,
+      "loss": 2.3127,
+      "step": 7760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.983548422879269e-06,
+      "loss": 2.2711,
+      "step": 7761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.967208756692354e-06,
+      "loss": 2.2916,
+      "step": 7762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.950890978138811e-06,
+      "loss": 2.7229,
+      "step": 7763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.934595089698308e-06,
+      "loss": 1.5834,
+      "step": 7764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.9183210938472616e-06,
+      "loss": 2.6485,
+      "step": 7765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.9020689930586615e-06,
+      "loss": 2.2373,
+      "step": 7766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.8858387898023e-06,
+      "loss": 2.1692,
+      "step": 7767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.869630486544486e-06,
+      "loss": 2.0395,
+      "step": 7768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.853444085748332e-06,
+      "loss": 2.226,
+      "step": 7769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.837279589873523e-06,
+      "loss": 2.5864,
+      "step": 7770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.821137001376508e-06,
+      "loss": 2.5345,
+      "step": 7771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.805016322710343e-06,
+      "loss": 2.5726,
+      "step": 7772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.788917556324768e-06,
+      "loss": 2.2636,
+      "step": 7773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.7728407046661926e-06,
+      "loss": 2.1447,
+      "step": 7774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.75678577017773e-06,
+      "loss": 2.2684,
+      "step": 7775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.740752755299094e-06,
+      "loss": 2.4029,
+      "step": 7776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.7247416624667365e-06,
+      "loss": 2.1534,
+      "step": 7777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.708752494113744e-06,
+      "loss": 2.5734,
+      "step": 7778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.692785252669857e-06,
+      "loss": 2.2377,
+      "step": 7779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.676839940561534e-06,
+      "loss": 2.3839,
+      "step": 7780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.66091656021182e-06,
+      "loss": 2.7389,
+      "step": 7781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.645015114040546e-06,
+      "loss": 2.4662,
+      "step": 7782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.629135604464097e-06,
+      "loss": 2.3681,
+      "step": 7783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.613278033895557e-06,
+      "loss": 2.3916,
+      "step": 7784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.597442404744717e-06,
+      "loss": 2.1445,
+      "step": 7785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.581628719417985e-06,
+      "loss": 1.9304,
+      "step": 7786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.565836980318489e-06,
+      "loss": 2.4829,
+      "step": 7787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.550067189845925e-06,
+      "loss": 2.5662,
+      "step": 7788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.534319350396776e-06,
+      "loss": 2.2603,
+      "step": 7789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.518593464364096e-06,
+      "loss": 2.8232,
+      "step": 7790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.502889534137655e-06,
+      "loss": 2.3369,
+      "step": 7791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.487207562103829e-06,
+      "loss": 2.3517,
+      "step": 7792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.471547550645744e-06,
+      "loss": 2.4698,
+      "step": 7793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.4559095021430966e-06,
+      "loss": 2.386,
+      "step": 7794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.440293418972336e-06,
+      "loss": 2.3589,
+      "step": 7795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.42469930350648e-06,
+      "loss": 2.0887,
+      "step": 7796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.409127158115284e-06,
+      "loss": 2.3875,
+      "step": 7797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.393576985165121e-06,
+      "loss": 2.447,
+      "step": 7798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.3780487870190515e-06,
+      "loss": 2.1275,
+      "step": 7799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.36254256603677e-06,
+      "loss": 2.1515,
+      "step": 7800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.34705832457466e-06,
+      "loss": 2.753,
+      "step": 7801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.331596064985738e-06,
+      "loss": 2.3468,
+      "step": 7802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.3161557896197245e-06,
+      "loss": 2.7457,
+      "step": 7803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.30073750082291e-06,
+      "loss": 1.4741,
+      "step": 7804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.28534120093837e-06,
+      "loss": 2.3704,
+      "step": 7805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.269966892305699e-06,
+      "loss": 2.8785,
+      "step": 7806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.2546145772612795e-06,
+      "loss": 1.9516,
+      "step": 7807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.239284258138077e-06,
+      "loss": 2.5936,
+      "step": 7808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.223975937265712e-06,
+      "loss": 2.1774,
+      "step": 7809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.208689616970507e-06,
+      "loss": 2.3571,
+      "step": 7810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.1934252995753875e-06,
+      "loss": 1.3437,
+      "step": 7811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.178182987399982e-06,
+      "loss": 2.6273,
+      "step": 7812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.162962682760552e-06,
+      "loss": 2.536,
+      "step": 7813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.147764387969999e-06,
+      "loss": 2.6627,
+      "step": 7814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.132588105337926e-06,
+      "loss": 2.72,
+      "step": 7815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.117433837170554e-06,
+      "loss": 2.0354,
+      "step": 7816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.102301585770774e-06,
+      "loss": 2.6019,
+      "step": 7817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.087191353438114e-06,
+      "loss": 2.5137,
+      "step": 7818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.07210314246877e-06,
+      "loss": 2.2456,
+      "step": 7819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.057036955155608e-06,
+      "loss": 2.1624,
+      "step": 7820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.041992793788113e-06,
+      "loss": 2.2179,
+      "step": 7821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.026970660652424e-06,
+      "loss": 2.4799,
+      "step": 7822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.011970558031381e-06,
+      "loss": 2.4912,
+      "step": 7823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.99699248820441e-06,
+      "loss": 2.8958,
+      "step": 7824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.982036453447658e-06,
+      "loss": 2.3011,
+      "step": 7825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.967102456033856e-06,
+      "loss": 2.6033,
+      "step": 7826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.952190498232422e-06,
+      "loss": 1.7524,
+      "step": 7827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.937300582309428e-06,
+      "loss": 2.8306,
+      "step": 7828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9224327105275795e-06,
+      "loss": 2.2521,
+      "step": 7829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.907586885146236e-06,
+      "loss": 1.6199,
+      "step": 7830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.892763108421443e-06,
+      "loss": 2.1088,
+      "step": 7831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.877961382605816e-06,
+      "loss": 2.0822,
+      "step": 7832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.863181709948722e-06,
+      "loss": 2.7609,
+      "step": 7833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.848424092696063e-06,
+      "loss": 2.3856,
+      "step": 7834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.833688533090496e-06,
+      "loss": 2.6334,
+      "step": 7835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.818975033371264e-06,
+      "loss": 2.2658,
+      "step": 7836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.804283595774261e-06,
+      "loss": 2.804,
+      "step": 7837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.789614222532051e-06,
+      "loss": 2.6421,
+      "step": 7838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.774966915873834e-06,
+      "loss": 1.0986,
+      "step": 7839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.760341678025448e-06,
+      "loss": 2.2233,
+      "step": 7840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.745738511209396e-06,
+      "loss": 2.2647,
+      "step": 7841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.7311574176448205e-06,
+      "loss": 2.6997,
+      "step": 7842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.716598399547467e-06,
+      "loss": 2.6295,
+      "step": 7843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.702061459129813e-06,
+      "loss": 2.5304,
+      "step": 7844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.687546598600911e-06,
+      "loss": 2.8141,
+      "step": 7845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6730538201664945e-06,
+      "loss": 2.4587,
+      "step": 7846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.658583126028903e-06,
+      "loss": 2.4848,
+      "step": 7847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.644134518387177e-06,
+      "loss": 2.6417,
+      "step": 7848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6297079994369434e-06,
+      "loss": 2.5752,
+      "step": 7849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.615303571370516e-06,
+      "loss": 1.2727,
+      "step": 7850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.600921236376792e-06,
+      "loss": 2.8994,
+      "step": 7851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.586560996641409e-06,
+      "loss": 2.5067,
+      "step": 7852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.572222854346552e-06,
+      "loss": 2.0486,
+      "step": 7853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.557906811671114e-06,
+      "loss": 2.6179,
+      "step": 7854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.543612870790553e-06,
+      "loss": 2.2225,
+      "step": 7855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.529341033877065e-06,
+      "loss": 2.2464,
+      "step": 7856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.515091303099416e-06,
+      "loss": 2.4022,
+      "step": 7857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.500863680623057e-06,
+      "loss": 2.6413,
+      "step": 7858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.486658168610024e-06,
+      "loss": 2.1179,
+      "step": 7859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.472474769219059e-06,
+      "loss": 2.8819,
+      "step": 7860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.458313484605486e-06,
+      "loss": 2.4257,
+      "step": 7861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.444174316921317e-06,
+      "loss": 2.2584,
+      "step": 7862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.430057268315168e-06,
+      "loss": 1.2478,
+      "step": 7863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.415962340932322e-06,
+      "loss": 2.4475,
+      "step": 7864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.401889536914649e-06,
+      "loss": 2.2278,
+      "step": 7865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.387838858400722e-06,
+      "loss": 2.2963,
+      "step": 7866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.373810307525699e-06,
+      "loss": 2.3637,
+      "step": 7867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.359803886421442e-06,
+      "loss": 2.6666,
+      "step": 7868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.345819597216366e-06,
+      "loss": 2.7149,
+      "step": 7869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.331857442035569e-06,
+      "loss": 1.4291,
+      "step": 7870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.3179174230007886e-06,
+      "loss": 2.3557,
+      "step": 7871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.303999542230396e-06,
+      "loss": 2.3301,
+      "step": 7872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.290103801839368e-06,
+      "loss": 2.7709,
+      "step": 7873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.276230203939379e-06,
+      "loss": 1.4863,
+      "step": 7874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.262378750638662e-06,
+      "loss": 2.8621,
+      "step": 7875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.248549444042149e-06,
+      "loss": 2.0398,
+      "step": 7876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.2347422862513415e-06,
+      "loss": 2.0493,
+      "step": 7877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.220957279364479e-06,
+      "loss": 2.1416,
+      "step": 7878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.207194425476318e-06,
+      "loss": 2.2677,
+      "step": 7879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.193453726678336e-06,
+      "loss": 2.7169,
+      "step": 7880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.179735185058564e-06,
+      "loss": 2.6183,
+      "step": 7881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.1660388027017655e-06,
+      "loss": 2.1403,
+      "step": 7882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.152364581689227e-06,
+      "loss": 2.1263,
+      "step": 7883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.138712524098969e-06,
+      "loss": 2.5048,
+      "step": 7884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.125082632005566e-06,
+      "loss": 1.9053,
+      "step": 7885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.111474907480278e-06,
+      "loss": 2.2885,
+      "step": 7886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.097889352590949e-06,
+      "loss": 2.3837,
+      "step": 7887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.084325969402113e-06,
+      "loss": 2.4382,
+      "step": 7888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.070784759974855e-06,
+      "loss": 2.4664,
+      "step": 7889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.057265726366976e-06,
+      "loss": 2.208,
+      "step": 7890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.043768870632835e-06,
+      "loss": 2.5151,
+      "step": 7891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.0302941948234915e-06,
+      "loss": 2.3572,
+      "step": 7892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.0168417009865565e-06,
+      "loss": 1.9367,
+      "step": 7893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.0034113911663446e-06,
+      "loss": 1.8439,
+      "step": 7894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.990003267403724e-06,
+      "loss": 2.683,
+      "step": 7895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.976617331736265e-06,
+      "loss": 2.1338,
+      "step": 7896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.963253586198123e-06,
+      "loss": 1.6507,
+      "step": 7897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.949912032820057e-06,
+      "loss": 2.1595,
+      "step": 7898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.936592673629546e-06,
+      "loss": 2.9766,
+      "step": 7899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.923295510650621e-06,
+      "loss": 2.382,
+      "step": 7900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.910020545903914e-06,
+      "loss": 2.539,
+      "step": 7901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.896767781406779e-06,
+      "loss": 2.4846,
+      "step": 7902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.883537219173105e-06,
+      "loss": 1.6192,
+      "step": 7903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.870328861213484e-06,
+      "loss": 2.6724,
+      "step": 7904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.85714270953506e-06,
+      "loss": 2.3415,
+      "step": 7905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.843978766141664e-06,
+      "loss": 2.9445,
+      "step": 7906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8308370330337125e-06,
+      "loss": 2.4464,
+      "step": 7907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8177175122082734e-06,
+      "loss": 2.1637,
+      "step": 7908
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.804620205659004e-06,
+      "loss": 2.1809,
+      "step": 7909
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7915451153762265e-06,
+      "loss": 2.3586,
+      "step": 7910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.77849224334687e-06,
+      "loss": 2.0182,
+      "step": 7911
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.765461591554464e-06,
+      "loss": 2.2684,
+      "step": 7912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.752453161979224e-06,
+      "loss": 2.071,
+      "step": 7913
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.739466956597903e-06,
+      "loss": 2.2746,
+      "step": 7914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.726502977383955e-06,
+      "loss": 2.4107,
+      "step": 7915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.713561226307388e-06,
+      "loss": 2.2423,
+      "step": 7916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7006417053349115e-06,
+      "loss": 2.317,
+      "step": 7917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.687744416429789e-06,
+      "loss": 2.9295,
+      "step": 7918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6748693615519354e-06,
+      "loss": 2.754,
+      "step": 7919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.662016542657853e-06,
+      "loss": 2.5748,
+      "step": 7920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6491859617007456e-06,
+      "loss": 2.2292,
+      "step": 7921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6363776206303375e-06,
+      "loss": 2.5334,
+      "step": 7922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6235915213930546e-06,
+      "loss": 1.9856,
+      "step": 7923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6108276659318757e-06,
+      "loss": 2.3797,
+      "step": 7924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.598086056186467e-06,
+      "loss": 2.7388,
+      "step": 7925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5853666940930615e-06,
+      "loss": 2.2039,
+      "step": 7926
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.572669581584531e-06,
+      "loss": 1.7798,
+      "step": 7927
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5599947205903478e-06,
+      "loss": 2.2958,
+      "step": 7928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5473421130366554e-06,
+      "loss": 2.3762,
+      "step": 7929
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.534711760846165e-06,
+      "loss": 2.9188,
+      "step": 7930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5221036659382086e-06,
+      "loss": 2.2815,
+      "step": 7931
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5095178302287697e-06,
+      "loss": 2.245,
+      "step": 7932
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.496954255630402e-06,
+      "loss": 1.9009,
+      "step": 7933
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4844129440523115e-06,
+      "loss": 2.4821,
+      "step": 7934
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4718938974003407e-06,
+      "loss": 2.6672,
+      "step": 7935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4593971175768674e-06,
+      "loss": 2.2025,
+      "step": 7936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.44692260648099e-06,
+      "loss": 2.1964,
+      "step": 7937
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.434470366008324e-06,
+      "loss": 2.698,
+      "step": 7938
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4220403980511733e-06,
+      "loss": 2.4624,
+      "step": 7939
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.409632704498427e-06,
+      "loss": 1.875,
+      "step": 7940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3972472872356106e-06,
+      "loss": 2.5683,
+      "step": 7941
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3848841481448173e-06,
+      "loss": 2.5002,
+      "step": 7942
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3725432891047955e-06,
+      "loss": 2.7443,
+      "step": 7943
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.360224711990911e-06,
+      "loss": 1.3167,
+      "step": 7944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3479284186751167e-06,
+      "loss": 1.9912,
+      "step": 7945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.335654411025984e-06,
+      "loss": 2.8865,
+      "step": 7946
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.323402690908722e-06,
+      "loss": 1.7962,
+      "step": 7947
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3111732601851404e-06,
+      "loss": 2.377,
+      "step": 7948
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2989661207136363e-06,
+      "loss": 2.8237,
+      "step": 7949
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2867812743492593e-06,
+      "loss": 2.4002,
+      "step": 7950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2746187229436615e-06,
+      "loss": 2.2059,
+      "step": 7951
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2624784683450646e-06,
+      "loss": 2.3004,
+      "step": 7952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2503605123983754e-06,
+      "loss": 2.4287,
+      "step": 7953
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2382648569450554e-06,
+      "loss": 2.2841,
+      "step": 7954
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2261915038232e-06,
+      "loss": 2.7726,
+      "step": 7955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2141404548674755e-06,
+      "loss": 2.5637,
+      "step": 7956
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2021117119092497e-06,
+      "loss": 1.3977,
+      "step": 7957
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.190105276776428e-06,
+      "loss": 2.011,
+      "step": 7958
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1781211512935167e-06,
+      "loss": 2.9092,
+      "step": 7959
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.166159337281676e-06,
+      "loss": 2.613,
+      "step": 7960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1542198365586523e-06,
+      "loss": 2.4088,
+      "step": 7961
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1423026509388107e-06,
+      "loss": 1.9009,
+      "step": 7962
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1304077822331354e-06,
+      "loss": 2.2398,
+      "step": 7963
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1185352322491643e-06,
+      "loss": 2.635,
+      "step": 7964
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1066850027911206e-06,
+      "loss": 2.9858,
+      "step": 7965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.094857095659764e-06,
+      "loss": 1.9486,
+      "step": 7966
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0830515126525568e-06,
+      "loss": 2.6488,
+      "step": 7967
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0712682555634305e-06,
+      "loss": 2.2589,
+      "step": 7968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.059507326183086e-06,
+      "loss": 2.1276,
+      "step": 7969
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0477687262986606e-06,
+      "loss": 2.343,
+      "step": 7970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0360524576940603e-06,
+      "loss": 2.5001,
+      "step": 7971
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.024358522149678e-06,
+      "loss": 2.4241,
+      "step": 7972
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.012686921442592e-06,
+      "loss": 2.0814,
+      "step": 7973
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0010376573464003e-06,
+      "loss": 2.3449,
+      "step": 7974
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9894107316314196e-06,
+      "loss": 2.0319,
+      "step": 7975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9778061460644543e-06,
+      "loss": 2.3077,
+      "step": 7976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966223902409026e-06,
+      "loss": 2.0999,
+      "step": 7977
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9546640024251444e-06,
+      "loss": 2.3303,
+      "step": 7978
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9431264478695536e-06,
+      "loss": 1.9768,
+      "step": 7979
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9316112404954682e-06,
+      "loss": 2.0436,
+      "step": 7980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.920118382052805e-06,
+      "loss": 2.6693,
+      "step": 7981
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9086478742880503e-06,
+      "loss": 2.1636,
+      "step": 7982
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.897199718944293e-06,
+      "loss": 2.7796,
+      "step": 7983
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.885773917761225e-06,
+      "loss": 2.2796,
+      "step": 7984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8743704724751236e-06,
+      "loss": 1.9219,
+      "step": 7985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.862989384818937e-06,
+      "loss": 2.3196,
+      "step": 7986
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8516306565221303e-06,
+      "loss": 2.6504,
+      "step": 7987
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.840294289310824e-06,
+      "loss": 2.8449,
+      "step": 7988
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.828980284907706e-06,
+      "loss": 2.2062,
+      "step": 7989
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.817688645032101e-06,
+      "loss": 2.1016,
+      "step": 7990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8064193713999207e-06,
+      "loss": 2.2012,
+      "step": 7991
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.795172465723661e-06,
+      "loss": 2.2684,
+      "step": 7992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7839479297124545e-06,
+      "loss": 1.836,
+      "step": 7993
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.772745765072004e-06,
+      "loss": 2.4899,
+      "step": 7994
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.761565973504615e-06,
+      "loss": 2.5614,
+      "step": 7995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7504085567092106e-06,
+      "loss": 2.6574,
+      "step": 7996
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.739273516381285e-06,
+      "loss": 2.696,
+      "step": 7997
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7281608542129852e-06,
+      "loss": 2.2188,
+      "step": 7998
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7170705718929932e-06,
+      "loss": 2.0899,
+      "step": 7999
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7060026711066275e-06,
+      "loss": 2.5588,
+      "step": 8000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6949571535358093e-06,
+      "loss": 2.1669,
+      "step": 8001
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.68393402085903e-06,
+      "loss": 2.8809,
+      "step": 8002
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6729332747513988e-06,
+      "loss": 2.2508,
+      "step": 8003
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.661954916884629e-06,
+      "loss": 2.3927,
+      "step": 8004
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.650998948927019e-06,
+      "loss": 2.4493,
+      "step": 8005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6400653725434874e-06,
+      "loss": 2.2264,
+      "step": 8006
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.629154189395488e-06,
+      "loss": 2.2596,
+      "step": 8007
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.618265401141145e-06,
+      "loss": 2.4124,
+      "step": 8008
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6073990094351517e-06,
+      "loss": 1.6661,
+      "step": 8009
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5965550159288028e-06,
+      "loss": 2.6462,
+      "step": 8010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5857334222699477e-06,
+      "loss": 2.3381,
+      "step": 8011
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.574934230103104e-06,
+      "loss": 2.4544,
+      "step": 8012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5641574410693255e-06,
+      "loss": 2.5574,
+      "step": 8013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5534030568062856e-06,
+      "loss": 2.4306,
+      "step": 8014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5426710789482607e-06,
+      "loss": 1.1375,
+      "step": 8015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5319615091261127e-06,
+      "loss": 2.2307,
+      "step": 8016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.521274348967289e-06,
+      "loss": 2.2671,
+      "step": 8017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.510609600095842e-06,
+      "loss": 1.6689,
+      "step": 8018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.499967264132441e-06,
+      "loss": 2.6151,
+      "step": 8019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4893473426942757e-06,
+      "loss": 3.0112,
+      "step": 8020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4787498373952386e-06,
+      "loss": 2.2468,
+      "step": 8021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4681747498457083e-06,
+      "loss": 2.4801,
+      "step": 8022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4576220816527324e-06,
+      "loss": 2.2292,
+      "step": 8023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.447091834419912e-06,
+      "loss": 2.3513,
+      "step": 8024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.43658400974745e-06,
+      "loss": 2.3146,
+      "step": 8025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4260986092321686e-06,
+      "loss": 2.7416,
+      "step": 8026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4156356344674445e-06,
+      "loss": 2.8143,
+      "step": 8027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4051950870432545e-06,
+      "loss": 2.4965,
+      "step": 8028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.39477696854618e-06,
+      "loss": 2.4033,
+      "step": 8029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.384381280559389e-06,
+      "loss": 2.4382,
+      "step": 8030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3740080246626325e-06,
+      "loss": 2.8606,
+      "step": 8031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.363657202432284e-06,
+      "loss": 2.402,
+      "step": 8032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.353328815441252e-06,
+      "loss": 1.7743,
+      "step": 8033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3430228652590812e-06,
+      "loss": 2.2252,
+      "step": 8034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.332739353451901e-06,
+      "loss": 2.4095,
+      "step": 8035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3224782815824127e-06,
+      "loss": 2.4452,
+      "step": 8036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3122396512099185e-06,
+      "loss": 2.1805,
+      "step": 8037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3020234638903234e-06,
+      "loss": 2.568,
+      "step": 8038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2918297211760862e-06,
+      "loss": 2.5647,
+      "step": 8039
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2816584246163004e-06,
+      "loss": 1.6638,
+      "step": 8040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2715095757566138e-06,
+      "loss": 2.2642,
+      "step": 8041
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2613831761392755e-06,
+      "loss": 1.9723,
+      "step": 8042
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2512792273031045e-06,
+      "loss": 2.4825,
+      "step": 8043
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.241197730783556e-06,
+      "loss": 2.5517,
+      "step": 8044
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2311386881126214e-06,
+      "loss": 2.2741,
+      "step": 8045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2211021008189112e-06,
+      "loss": 2.1177,
+      "step": 8046
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.211087970427605e-06,
+      "loss": 2.1432,
+      "step": 8047
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.201096298460503e-06,
+      "loss": 2.4657,
+      "step": 8048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1911270864359398e-06,
+      "loss": 2.3294,
+      "step": 8049
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1811803358688706e-06,
+      "loss": 2.4073,
+      "step": 8050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1712560482708363e-06,
+      "loss": 2.7906,
+      "step": 8051
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1613542251499804e-06,
+      "loss": 2.4552,
+      "step": 8052
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.151474868010966e-06,
+      "loss": 2.7543,
+      "step": 8053
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1416179783551424e-06,
+      "loss": 1.2911,
+      "step": 8054
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1317835576803445e-06,
+      "loss": 2.5383,
+      "step": 8055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.12197160748106e-06,
+      "loss": 2.4264,
+      "step": 8056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.11218212924833e-06,
+      "loss": 2.7427,
+      "step": 8057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.102415124469814e-06,
+      "loss": 2.0537,
+      "step": 8058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0926705946297086e-06,
+      "loss": 2.0491,
+      "step": 8059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0829485412088286e-06,
+      "loss": 2.6585,
+      "step": 8060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.073248965684593e-06,
+      "loss": 1.9938,
+      "step": 8061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0635718695309223e-06,
+      "loss": 2.5991,
+      "step": 8062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0539172542184236e-06,
+      "loss": 2.4572,
+      "step": 8063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0442851212142064e-06,
+      "loss": 2.4557,
+      "step": 8064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.034675471982017e-06,
+      "loss": 1.709,
+      "step": 8065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.02508830798217e-06,
+      "loss": 2.8114,
+      "step": 8066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.015523630671534e-06,
+      "loss": 2.6847,
+      "step": 8067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0059814415036126e-06,
+      "loss": 2.7778,
+      "step": 8068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.996461741928446e-06,
+      "loss": 2.4541,
+      "step": 8069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9869645333926597e-06,
+      "loss": 2.5137,
+      "step": 8070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.977489817339517e-06,
+      "loss": 2.3696,
+      "step": 8071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.968037595208799e-06,
+      "loss": 2.293,
+      "step": 8072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586078684368898e-06,
+      "loss": 2.5867,
+      "step": 8073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9492006384567603e-06,
+      "loss": 2.3965,
+      "step": 8074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.939815906697967e-06,
+      "loss": 2.3632,
+      "step": 8075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.930453674586635e-06,
+      "loss": 2.5306,
+      "step": 8076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921113943545477e-06,
+      "loss": 2.1595,
+      "step": 8077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.911796714993774e-06,
+      "loss": 2.2254,
+      "step": 8078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.902501990347427e-06,
+      "loss": 2.1622,
+      "step": 8079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8932297710188726e-06,
+      "loss": 2.6083,
+      "step": 8080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8839800584171338e-06,
+      "loss": 2.758,
+      "step": 8081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8747528539478362e-06,
+      "loss": 2.6533,
+      "step": 8082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8655481590131583e-06,
+      "loss": 2.8444,
+      "step": 8083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.856365975011881e-06,
+      "loss": 2.1006,
+      "step": 8084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8472063033393713e-06,
+      "loss": 2.8926,
+      "step": 8085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8380691453875163e-06,
+      "loss": 2.1222,
+      "step": 8086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8289545025448716e-06,
+      "loss": 2.839,
+      "step": 8087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8198623761964958e-06,
+      "loss": 2.7118,
+      "step": 8088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8107927677240675e-06,
+      "loss": 2.4472,
+      "step": 8089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.801745678505817e-06,
+      "loss": 2.177,
+      "step": 8090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7927211099165784e-06,
+      "loss": 2.3602,
+      "step": 8091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7837190633277376e-06,
+      "loss": 2.74,
+      "step": 8092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7747395401073006e-06,
+      "loss": 2.6436,
+      "step": 8093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7657825416197924e-06,
+      "loss": 2.6598,
+      "step": 8094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.756848069226341e-06,
+      "loss": 2.5568,
+      "step": 8095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7479361242846768e-06,
+      "loss": 2.3309,
+      "step": 8096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7390467081490666e-06,
+      "loss": 2.5566,
+      "step": 8097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.73017982217038e-06,
+      "loss": 2.7896,
+      "step": 8098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7213354676960556e-06,
+      "loss": 2.1682,
+      "step": 8099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7125136460701016e-06,
+      "loss": 2.3194,
+      "step": 8100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7037143586331125e-06,
+      "loss": 1.5527,
+      "step": 8101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6949376067222352e-06,
+      "loss": 2.0556,
+      "step": 8102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6861833916712364e-06,
+      "loss": 2.2602,
+      "step": 8103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6774517148104183e-06,
+      "loss": 2.3623,
+      "step": 8104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6687425774666696e-06,
+      "loss": 2.6745,
+      "step": 8105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.660055980963465e-06,
+      "loss": 2.5711,
+      "step": 8106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.651391926620832e-06,
+      "loss": 1.7804,
+      "step": 8107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.642750415755384e-06,
+      "loss": 1.7412,
+      "step": 8108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.634131449680337e-06,
+      "loss": 2.3992,
+      "step": 8109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.625535029705427e-06,
+      "loss": 2.2224,
+      "step": 8110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6169611571369756e-06,
+      "loss": 2.4534,
+      "step": 8111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6084098332779404e-06,
+      "loss": 2.6852,
+      "step": 8112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5998810594277655e-06,
+      "loss": 1.6026,
+      "step": 8113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.591374836882514e-06,
+      "loss": 2.983,
+      "step": 8114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5828911669348354e-06,
+      "loss": 2.462,
+      "step": 8115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5744300508739149e-06,
+      "loss": 2.5457,
+      "step": 8116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.565991489985541e-06,
+      "loss": 2.8082,
+      "step": 8117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5575754855520372e-06,
+      "loss": 2.6488,
+      "step": 8118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5491820388523479e-06,
+      "loss": 2.2372,
+      "step": 8119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5408111511619525e-06,
+      "loss": 2.484,
+      "step": 8120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5324628237529169e-06,
+      "loss": 2.6081,
+      "step": 8121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.524137057893876e-06,
+      "loss": 2.6959,
+      "step": 8122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5158338548500515e-06,
+      "loss": 2.5283,
+      "step": 8123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5075532158832004e-06,
+      "loss": 1.996,
+      "step": 8124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.499295142251683e-06,
+      "loss": 2.4328,
+      "step": 8125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4910596352104286e-06,
+      "loss": 2.3764,
+      "step": 8126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.482846696010903e-06,
+      "loss": 1.4315,
+      "step": 8127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4746563259011913e-06,
+      "loss": 2.2517,
+      "step": 8128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4664885261259141e-06,
+      "loss": 2.1681,
+      "step": 8129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4583432979262787e-06,
+      "loss": 2.1907,
+      "step": 8130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4502206425400454e-06,
+      "loss": 2.7786,
+      "step": 8131
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4421205612015763e-06,
+      "loss": 2.2492,
+      "step": 8132
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4340430551417704e-06,
+      "loss": 2.2839,
+      "step": 8133
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4259881255881123e-06,
+      "loss": 2.6555,
+      "step": 8134
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4179557737646397e-06,
+      "loss": 2.3354,
+      "step": 8135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4099460008919927e-06,
+      "loss": 1.8347,
+      "step": 8136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4019588081873478e-06,
+      "loss": 2.3685,
+      "step": 8137
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3939941968644507e-06,
+      "loss": 2.2785,
+      "step": 8138
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.38605216813365e-06,
+      "loss": 2.5887,
+      "step": 8139
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3781327232018302e-06,
+      "loss": 2.6575,
+      "step": 8140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3702358632724452e-06,
+      "loss": 2.5582,
+      "step": 8141
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3623615895455186e-06,
+      "loss": 2.7759,
+      "step": 8142
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3545099032176599e-06,
+      "loss": 2.209,
+      "step": 8143
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3466808054820476e-06,
+      "loss": 2.4971,
+      "step": 8144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3388742975283807e-06,
+      "loss": 2.4605,
+      "step": 8145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.331090380542993e-06,
+      "loss": 2.263,
+      "step": 8146
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.323329055708705e-06,
+      "loss": 2.2049,
+      "step": 8147
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.315590324205007e-06,
+      "loss": 2.5738,
+      "step": 8148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3078741872078414e-06,
+      "loss": 1.2958,
+      "step": 8149
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.30018064588982e-06,
+      "loss": 2.333,
+      "step": 8150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.292509701420058e-06,
+      "loss": 2.2492,
+      "step": 8151
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2848613549642394e-06,
+      "loss": 2.2947,
+      "step": 8152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2772356076846514e-06,
+      "loss": 2.3288,
+      "step": 8153
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2696324607401165e-06,
+      "loss": 2.473,
+      "step": 8154
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2620519152860275e-06,
+      "loss": 2.7029,
+      "step": 8155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.254493972474363e-06,
+      "loss": 2.7261,
+      "step": 8156
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2469586334536042e-06,
+      "loss": 2.6206,
+      "step": 8157
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2394458993689016e-06,
+      "loss": 1.2545,
+      "step": 8158
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2319557713618755e-06,
+      "loss": 2.1288,
+      "step": 8159
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2244882505707488e-06,
+      "loss": 2.5742,
+      "step": 8160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2170433381303135e-06,
+      "loss": 1.7756,
+      "step": 8161
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2096210351719149e-06,
+      "loss": 2.1699,
+      "step": 8162
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.202221342823484e-06,
+      "loss": 2.4568,
+      "step": 8163
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.194844262209471e-06,
+      "loss": 2.8131,
+      "step": 8164
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1874897944509465e-06,
+      "loss": 2.4976,
+      "step": 8165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1801579406654992e-06,
+      "loss": 2.489,
+      "step": 8166
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.172848701967305e-06,
+      "loss": 2.2447,
+      "step": 8167
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1655620794670918e-06,
+      "loss": 2.4713,
+      "step": 8168
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.158298074272157e-06,
+      "loss": 1.8528,
+      "step": 8169
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.151056687486368e-06,
+      "loss": 2.3188,
+      "step": 8170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1438379202101445e-06,
+      "loss": 1.6656,
+      "step": 8171
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1366417735404588e-06,
+      "loss": 2.3151,
+      "step": 8172
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1294682485708695e-06,
+      "loss": 2.5824,
+      "step": 8173
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1223173463914714e-06,
+      "loss": 1.5359,
+      "step": 8174
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1151890680889453e-06,
+      "loss": 1.1584,
+      "step": 8175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1080834147465412e-06,
+      "loss": 2.6441,
+      "step": 8176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1010003874440288e-06,
+      "loss": 2.5177,
+      "step": 8177
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0939399872577637e-06,
+      "loss": 2.5913,
+      "step": 8178
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0869022152606877e-06,
+      "loss": 2.517,
+      "step": 8179
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0798870725222785e-06,
+      "loss": 2.4662,
+      "step": 8180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0728945601085503e-06,
+      "loss": 2.8676,
+      "step": 8181
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0659246790821197e-06,
+      "loss": 2.5792,
+      "step": 8182
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0589774305021559e-06,
+      "loss": 2.6431,
+      "step": 8183
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.052052815424398e-06,
+      "loss": 2.5483,
+      "step": 8184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0451508349010873e-06,
+      "loss": 2.1505,
+      "step": 8185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0382714899811018e-06,
+      "loss": 2.195,
+      "step": 8186
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0314147817098383e-06,
+      "loss": 1.2228,
+      "step": 8187
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0245807111292637e-06,
+      "loss": 2.4399,
+      "step": 8188
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0177692792778802e-06,
+      "loss": 2.1373,
+      "step": 8189
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.01098048719081e-06,
+      "loss": 2.2496,
+      "step": 8190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.004214335899678e-06,
+      "loss": 2.2258,
+      "step": 8191
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.97470826432678e-07,
+      "loss": 2.3501,
+      "step": 8192
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.907499598145907e-07,
+      "loss": 2.1275,
+      "step": 8193
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.840517370667322e-07,
+      "loss": 2.5695,
+      "step": 8194
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773761592069718e-07,
+      "loss": 2.5163,
+      "step": 8195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.70723227249781e-07,
+      "loss": 2.6859,
+      "step": 8196
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640929422061183e-07,
+      "loss": 2.2958,
+      "step": 8197
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.574853050835607e-07,
+      "loss": 2.2788,
+      "step": 8198
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.509003168862383e-07,
+      "loss": 2.7199,
+      "step": 8199
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.443379786148009e-07,
+      "loss": 1.9805,
+      "step": 8200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.377982912664839e-07,
+      "loss": 2.314,
+      "step": 8201
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.312812558350924e-07,
+      "loss": 2.3137,
+      "step": 8202
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.247868733109343e-07,
+      "loss": 1.8762,
+      "step": 8203
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.1831514468097e-07,
+      "loss": 2.0741,
+      "step": 8204
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.118660709286296e-07,
+      "loss": 2.2547,
+      "step": 8205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.054396530339292e-07,
+      "loss": 2.3372,
+      "step": 8206
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.990358919734542e-07,
+      "loss": 2.9761,
+      "step": 8207
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.926547887203428e-07,
+      "loss": 2.5516,
+      "step": 8208
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.862963442442861e-07,
+      "loss": 2.0758,
+      "step": 8209
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.799605595115111e-07,
+      "loss": 2.2987,
+      "step": 8210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.736474354848477e-07,
+      "loss": 2.932,
+      "step": 8211
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.673569731236451e-07,
+      "loss": 1.7484,
+      "step": 8212
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.610891733838055e-07,
+      "loss": 2.0116,
+      "step": 8213
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.548440372178333e-07,
+      "loss": 2.0279,
+      "step": 8214
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.486215655747364e-07,
+      "loss": 2.3103,
+      "step": 8215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.424217594000915e-07,
+      "loss": 2.6079,
+      "step": 8216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.362446196360617e-07,
+      "loss": 2.4605,
+      "step": 8217
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.300901472213295e-07,
+      "loss": 2.1729,
+      "step": 8218
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.239583430911467e-07,
+      "loss": 2.3689,
+      "step": 8219
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.17849208177318e-07,
+      "loss": 2.4309,
+      "step": 8220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.117627434082174e-07,
+      "loss": 2.3122,
+      "step": 8221
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.056989497087385e-07,
+      "loss": 2.6668,
+      "step": 8222
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.996578280003774e-07,
+      "loss": 2.2297,
+      "step": 8223
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.936393792011497e-07,
+      "loss": 2.0248,
+      "step": 8224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.87643604225624e-07,
+      "loss": 2.186,
+      "step": 8225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.816705039849547e-07,
+      "loss": 2.2388,
+      "step": 8226
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.757200793868157e-07,
+      "loss": 2.3482,
+      "step": 8227
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.69792331335467e-07,
+      "loss": 2.0771,
+      "step": 8228
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.638872607316882e-07,
+      "loss": 2.4299,
+      "step": 8229
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.580048684728612e-07,
+      "loss": 2.3699,
+      "step": 8230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.52145155452838e-07,
+      "loss": 2.3738,
+      "step": 8231
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.463081225621226e-07,
+      "loss": 2.469,
+      "step": 8232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.404937706877056e-07,
+      "loss": 2.1706,
+      "step": 8233
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.347021007131804e-07,
+      "loss": 3.0061,
+      "step": 8234
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.289331135186261e-07,
+      "loss": 2.7626,
+      "step": 8235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.231868099807415e-07,
+      "loss": 2.4842,
+      "step": 8236
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.174631909727446e-07,
+      "loss": 2.1758,
+      "step": 8237
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.117622573644066e-07,
+      "loss": 2.4159,
+      "step": 8238
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.06084010022051e-07,
+      "loss": 2.5129,
+      "step": 8239
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.004284498085877e-07,
+      "loss": 2.6311,
+      "step": 8240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.94795577583429e-07,
+      "loss": 2.2793,
+      "step": 8241
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.891853942025738e-07,
+      "loss": 2.6411,
+      "step": 8242
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.835979005185565e-07,
+      "loss": 2.4008,
+      "step": 8243
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.780330973804649e-07,
+      "loss": 2.3241,
+      "step": 8244
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.724909856339555e-07,
+      "loss": 2.7367,
+      "step": 8245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.669715661212049e-07,
+      "loss": 2.7325,
+      "step": 8246
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.614748396809755e-07,
+      "loss": 2.0421,
+      "step": 8247
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.560008071485489e-07,
+      "loss": 2.2279,
+      "step": 8248
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.505494693557934e-07,
+      "loss": 1.7504,
+      "step": 8249
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.45120827131096e-07,
+      "loss": 2.6349,
+      "step": 8250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.397148812994301e-07,
+      "loss": 2.1694,
+      "step": 8251
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.343316326822556e-07,
+      "loss": 2.147,
+      "step": 8252
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.289710820976846e-07,
+      "loss": 2.8021,
+      "step": 8253
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.236332303602653e-07,
+      "loss": 2.6269,
+      "step": 8254
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.183180782811825e-07,
+      "loss": 2.6111,
+      "step": 8255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.130256266681399e-07,
+      "loss": 2.2284,
+      "step": 8256
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.077558763253942e-07,
+      "loss": 2.4162,
+      "step": 8257
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.025088280537549e-07,
+      "loss": 2.117,
+      "step": 8258
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.972844826505674e-07,
+      "loss": 2.2862,
+      "step": 8259
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.9208284090973e-07,
+      "loss": 2.5295,
+      "step": 8260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.869039036217271e-07,
+      "loss": 2.3114,
+      "step": 8261
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.817476715735458e-07,
+      "loss": 1.883,
+      "step": 8262
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.766141455487428e-07,
+      "loss": 2.0114,
+      "step": 8263
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.715033263274271e-07,
+      "loss": 2.5007,
+      "step": 8264
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.664152146862444e-07,
+      "loss": 2.5081,
+      "step": 8265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.613498113984094e-07,
+      "loss": 2.7483,
+      "step": 8266
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.56307117233673e-07,
+      "loss": 2.3022,
+      "step": 8267
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.51287132958339e-07,
+      "loss": 2.6202,
+      "step": 8268
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.462898593352638e-07,
+      "loss": 2.396,
+      "step": 8269
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.413152971238233e-07,
+      "loss": 2.4869,
+      "step": 8270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.36363447079996e-07,
+      "loss": 2.0411,
+      "step": 8271
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.314343099562635e-07,
+      "loss": 1.9319,
+      "step": 8272
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.265278865016598e-07,
+      "loss": 2.0574,
+      "step": 8273
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.216441774618052e-07,
+      "loss": 2.7566,
+      "step": 8274
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.167831835788394e-07,
+      "loss": 2.4723,
+      "step": 8275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.119449055914216e-07,
+      "loss": 2.8594,
+      "step": 8276
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.071293442348301e-07,
+      "loss": 2.9135,
+      "step": 8277
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.02336500240813e-07,
+      "loss": 2.119,
+      "step": 8278
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.975663743377379e-07,
+      "loss": 2.5033,
+      "step": 8279
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.928189672504746e-07,
+      "loss": 2.1629,
+      "step": 8280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.880942797004461e-07,
+      "loss": 2.086,
+      "step": 8281
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.833923124056449e-07,
+      "loss": 2.0744,
+      "step": 8282
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.78713066080566e-07,
+      "loss": 2.5212,
+      "step": 8283
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.7405654143630736e-07,
+      "loss": 2.5581,
+      "step": 8284
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.694227391804861e-07,
+      "loss": 2.6364,
+      "step": 8285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6481166001727246e-07,
+      "loss": 2.2177,
+      "step": 8286
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6022330464733915e-07,
+      "loss": 2.3153,
+      "step": 8287
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.556576737679951e-07,
+      "loss": 2.5676,
+      "step": 8288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.5111476807303537e-07,
+      "loss": 2.0714,
+      "step": 8289
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.4659458825279105e-07,
+      "loss": 2.4623,
+      "step": 8290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.420971349941793e-07,
+      "loss": 2.8652,
+      "step": 8291
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.3762240898063685e-07,
+      "loss": 2.4541,
+      "step": 8292
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.3317041089216965e-07,
+      "loss": 2.7429,
+      "step": 8293
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.287411414053032e-07,
+      "loss": 2.5175,
+      "step": 8294
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.2433460119313257e-07,
+      "loss": 2.5474,
+      "step": 8295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.1995079092527196e-07,
+      "loss": 3.2978,
+      "step": 8296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.155897112679052e-07,
+      "loss": 1.979,
+      "step": 8297
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.112513628837688e-07,
+      "loss": 2.322,
+      "step": 8298
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.0693574643210215e-07,
+      "loss": 2.4222,
+      "step": 8299
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.026428625687472e-07,
+      "loss": 1.2942,
+      "step": 8300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.983727119460489e-07,
+      "loss": 1.7528,
+      "step": 8301
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.9412529521290483e-07,
+      "loss": 2.5189,
+      "step": 8302
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.899006130147653e-07,
+      "loss": 2.3195,
+      "step": 8303
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8569866599365006e-07,
+      "loss": 1.1498,
+      "step": 8304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8151945478808175e-07,
+      "loss": 2.6546,
+      "step": 8305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.773629800331357e-07,
+      "loss": 2.3023,
+      "step": 8306
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.732292423604566e-07,
+      "loss": 1.8535,
+      "step": 8307
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6911824239820865e-07,
+      "loss": 2.5157,
+      "step": 8308
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.650299807711088e-07,
+      "loss": 2.6695,
+      "step": 8309
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.609644581004267e-07,
+      "loss": 2.9049,
+      "step": 8310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5692167500398496e-07,
+      "loss": 2.64,
+      "step": 8311
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.529016320961253e-07,
+      "loss": 2.1156,
+      "step": 8312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.489043299877259e-07,
+      "loss": 2.762,
+      "step": 8313
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.449297692862507e-07,
+      "loss": 2.3768,
+      "step": 8314
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4097795059568333e-07,
+      "loss": 2.212,
+      "step": 8315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3704887451656003e-07,
+      "loss": 1.499,
+      "step": 8316
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3314254164591994e-07,
+      "loss": 2.7631,
+      "step": 8317
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.292589525774214e-07,
+      "loss": 2.3738,
+      "step": 8318
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.253981079012091e-07,
+      "loss": 2.4187,
+      "step": 8319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.215600082039804e-07,
+      "loss": 2.4062,
+      "step": 8320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.177446540690021e-07,
+      "loss": 2.4907,
+      "step": 8321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1395204607604364e-07,
+      "loss": 2.4069,
+      "step": 8322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.1018218480146075e-07,
+      "loss": 2.5308,
+      "step": 8323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.064350708181118e-07,
+      "loss": 2.1024,
+      "step": 8324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0271070469544137e-07,
+      "loss": 2.1333,
+      "step": 8325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9900908699939665e-07,
+      "loss": 2.7736,
+      "step": 8326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.953302182924944e-07,
+      "loss": 2.3786,
+      "step": 8327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.916740991337874e-07,
+      "loss": 2.2796,
+      "step": 8328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.880407300788812e-07,
+      "loss": 2.9273,
+      "step": 8329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8443011167988417e-07,
+      "loss": 2.3365,
+      "step": 8330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8084224448549076e-07,
+      "loss": 2.5717,
+      "step": 8331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7727712904093146e-07,
+      "loss": 2.3513,
+      "step": 8332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7373476588795626e-07,
+      "loss": 2.6403,
+      "step": 8333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.702151555648846e-07,
+      "loss": 2.4847,
+      "step": 8334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.667182986065719e-07,
+      "loss": 2.657,
+      "step": 8335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6324419554439315e-07,
+      "loss": 1.2437,
+      "step": 8336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5979284690629267e-07,
+      "loss": 2.365,
+      "step": 8337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5636425321673426e-07,
+      "loss": 2.3053,
+      "step": 8338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5295841499676786e-07,
+      "loss": 2.2881,
+      "step": 8339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4957533276391275e-07,
+      "loss": 2.6329,
+      "step": 8340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4621500703230767e-07,
+      "loss": 2.4795,
+      "step": 8341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4287743831257754e-07,
+      "loss": 2.7053,
+      "step": 8342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3956262711191667e-07,
+      "loss": 2.8309,
+      "step": 8343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3627057393405536e-07,
+      "loss": 2.5949,
+      "step": 8344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3300127927926016e-07,
+      "loss": 2.4852,
+      "step": 8345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2975474364433365e-07,
+      "loss": 2.0823,
+      "step": 8346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2653096752263123e-07,
+      "loss": 3.1799,
+      "step": 8347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2332995140406096e-07,
+      "loss": 2.1076,
+      "step": 8348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2015169577505042e-07,
+      "loss": 2.0843,
+      "step": 8349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1699620111857996e-07,
+      "loss": 2.356,
+      "step": 8350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.138634679141493e-07,
+      "loss": 2.8258,
+      "step": 8351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1075349663784435e-07,
+      "loss": 2.412,
+      "step": 8352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0766628776223705e-07,
+      "loss": 2.6518,
+      "step": 8353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0460184175650207e-07,
+      "loss": 2.7868,
+      "step": 8354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.015601590863003e-07,
+      "loss": 1.9976,
+      "step": 8355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9854124021384533e-07,
+      "loss": 2.5808,
+      "step": 8356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554508559792015e-07,
+      "loss": 2.8821,
+      "step": 8357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9257169569381058e-07,
+      "loss": 2.5384,
+      "step": 8358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.896210709533885e-07,
+      "loss": 2.2091,
+      "step": 8359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8669321182501196e-07,
+      "loss": 2.2639,
+      "step": 8360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8378811875360832e-07,
+      "loss": 2.6449,
+      "step": 8361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8090579218067446e-07,
+      "loss": 2.8168,
+      "step": 8362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7804623254417671e-07,
+      "loss": 2.5374,
+      "step": 8363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.752094402786841e-07,
+      "loss": 2.5168,
+      "step": 8364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7239541581528515e-07,
+      "loss": 2.7627,
+      "step": 8365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6960415958160444e-07,
+      "loss": 2.0583,
+      "step": 8366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.668356720018027e-07,
+      "loss": 2.4472,
+      "step": 8367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6408995349659336e-07,
+      "loss": 2.5738,
+      "step": 8368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6136700448322603e-07,
+      "loss": 2.3193,
+      "step": 8369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5866682537548637e-07,
+      "loss": 2.0785,
+      "step": 8370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5598941658369613e-07,
+      "loss": 2.4005,
+      "step": 8371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.533347785147132e-07,
+      "loss": 2.8972,
+      "step": 8372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.507029115719649e-07,
+      "loss": 2.5271,
+      "step": 8373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4809381615538128e-07,
+      "loss": 2.6176,
+      "step": 8374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.455074926614619e-07,
+      "loss": 2.1707,
+      "step": 8375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.455074926614619e-07,
+      "loss": 1.5324,
+      "step": 8376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.42943941483209e-07,
+      "loss": 2.3908,
+      "step": 8377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4040316301021093e-07,
+      "loss": 2.8616,
+      "step": 8378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3788515762855888e-07,
+      "loss": 2.2608,
+      "step": 8379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3538992572091346e-07,
+      "loss": 2.7353,
+      "step": 8380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3291746766643795e-07,
+      "loss": 1.6758,
+      "step": 8381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3046778384084856e-07,
+      "loss": 2.4255,
+      "step": 8382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2804087461643076e-07,
+      "loss": 2.511,
+      "step": 8383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.256367403619729e-07,
+      "loss": 2.4258,
+      "step": 8384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2325538144279945e-07,
+      "loss": 2.5796,
+      "step": 8385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2089679822080421e-07,
+      "loss": 2.1885,
+      "step": 8386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1856099105441719e-07,
+      "loss": 2.7544,
+      "step": 8387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1624796029857109e-07,
+      "loss": 2.3913,
+      "step": 8388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1395770630475143e-07,
+      "loss": 2.2032,
+      "step": 8389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1169022942102978e-07,
+      "loss": 3.0857,
+      "step": 8390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0944552999194722e-07,
+      "loss": 2.4537,
+      "step": 8391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0722360835863087e-07,
+      "loss": 2.2891,
+      "step": 8392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0502446485871064e-07,
+      "loss": 2.0255,
+      "step": 8393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0284809982640252e-07,
+      "loss": 1.8977,
+      "step": 8394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0069451359240866e-07,
+      "loss": 2.5334,
+      "step": 8395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.856370648401723e-08,
+      "loss": 2.2814,
+      "step": 8396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64556788250026e-08,
+      "loss": 2.7738,
+      "step": 8397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.437043093574181e-08,
+      "loss": 2.1535,
+      "step": 8398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.230796313308142e-08,
+      "loss": 2.2586,
+      "step": 8399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.026827573045403e-08,
+      "loss": 1.9008,
+      "step": 8400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.825136903781171e-08,
+      "loss": 2.4066,
+      "step": 8401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.625724336167595e-08,
+      "loss": 1.9289,
+      "step": 8402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.428589900503768e-08,
+      "loss": 2.141,
+      "step": 8403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.233733626750726e-08,
+      "loss": 2.5928,
+      "step": 8404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.041155544516454e-08,
+      "loss": 2.5052,
+      "step": 8405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.850855683065871e-08,
+      "loss": 2.3668,
+      "step": 8406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.66283407131918e-08,
+      "loss": 2.9801,
+      "step": 8407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.477090737848523e-08,
+      "loss": 2.8238,
+      "step": 8408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.293625710877993e-08,
+      "loss": 2.2693,
+      "step": 8409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.112439018290284e-08,
+      "loss": 2.6409,
+      "step": 8410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.933530687616706e-08,
+      "loss": 2.7203,
+      "step": 8411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.756900746045513e-08,
+      "loss": 2.3599,
+      "step": 8412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.582549220418565e-08,
+      "loss": 2.1739,
+      "step": 8413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.410476137228004e-08,
+      "loss": 2.2927,
+      "step": 8414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.240681522626245e-08,
+      "loss": 2.1473,
+      "step": 8415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.073165402414315e-08,
+      "loss": 1.8608,
+      "step": 8416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.907927802048518e-08,
+      "loss": 2.3724,
+      "step": 8417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.7449687466387674e-08,
+      "loss": 2.6738,
+      "step": 8418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.584288260948589e-08,
+      "loss": 2.558,
+      "step": 8419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.4258863693934504e-08,
+      "loss": 2.2522,
+      "step": 8420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.2697630960490953e-08,
+      "loss": 2.3888,
+      "step": 8421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.115918464636548e-08,
+      "loss": 2.6212,
+      "step": 8422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9643524985371055e-08,
+      "loss": 2.0799,
+      "step": 8423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.815065220780678e-08,
+      "loss": 2.428,
+      "step": 8424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.668056654055785e-08,
+      "loss": 2.3237,
+      "step": 8425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.523326820701223e-08,
+      "loss": 2.5399,
+      "step": 8426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.3808757427094e-08,
+      "loss": 2.5038,
+      "step": 8427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.2407034417296646e-08,
+      "loss": 2.2636,
+      "step": 8428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.102809939061647e-08,
+      "loss": 2.4227,
+      "step": 8429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.967195255660249e-08,
+      "loss": 2.0099,
+      "step": 8430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8338594121339885e-08,
+      "loss": 2.1343,
+      "step": 8431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.702802428744989e-08,
+      "loss": 2.6973,
+      "step": 8432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.574024325408986e-08,
+      "loss": 2.2763,
+      "step": 8433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4475251216969925e-08,
+      "loss": 2.4854,
+      "step": 8434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.323304836828633e-08,
+      "loss": 2.5481,
+      "step": 8435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.201363489683806e-08,
+      "loss": 2.3353,
+      "step": 8436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.081701098792688e-08,
+      "loss": 3.2778,
+      "step": 8437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9643176823374026e-08,
+      "loss": 2.1632,
+      "step": 8438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8492132581586778e-08,
+      "loss": 2.4492,
+      "step": 8439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7363878437458577e-08,
+      "loss": 1.6577,
+      "step": 8440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.625841456246891e-08,
+      "loss": 2.4897,
+      "step": 8441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5175741124566773e-08,
+      "loss": 2.4965,
+      "step": 8442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4115858288320523e-08,
+      "loss": 2.5395,
+      "step": 8443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.307876621476801e-08,
+      "loss": 1.8886,
+      "step": 8444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.206446506151649e-08,
+      "loss": 1.8232,
+      "step": 8445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.107295498269268e-08,
+      "loss": 2.3811,
+      "step": 8446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.01042361289927e-08,
+      "loss": 2.3832,
+      "step": 8447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.915830864759882e-08,
+      "loss": 2.5729,
+      "step": 8448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8235172682279364e-08,
+      "loss": 1.7884,
+      "step": 8449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.733482837328881e-08,
+      "loss": 2.6148,
+      "step": 8450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.64572758574677e-08,
+      "loss": 1.4496,
+      "step": 8451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.560251526817602e-08,
+      "loss": 2.636,
+      "step": 8452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4770546735293209e-08,
+      "loss": 2.483,
+      "step": 8453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.396137038525147e-08,
+      "loss": 2.2336,
+      "step": 8454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3174986341002447e-08,
+      "loss": 2.6057,
+      "step": 8455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2411394722067203e-08,
+      "loss": 2.5925,
+      "step": 8456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1670595644486247e-08,
+      "loss": 2.1398,
+      "step": 8457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0952589220802888e-08,
+      "loss": 2.2949,
+      "step": 8458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.025737556016315e-08,
+      "loss": 2.5769,
+      "step": 8459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.584954768199205e-09,
+      "loss": 3.0434,
+      "step": 8460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.935326947079324e-09,
+      "loss": 2.6608,
+      "step": 8461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.308492195541194e-09,
+      "loss": 1.2395,
+      "step": 8462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.70445060884195e-09,
+      "loss": 2.1728,
+      "step": 8463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.1232022787581775e-09,
+      "loss": 2.4736,
+      "step": 8464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.564747293635875e-09,
+      "loss": 2.345,
+      "step": 8465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.029085738323836e-09,
+      "loss": 2.063,
+      "step": 8466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.516217694240266e-09,
+      "loss": 2.5709,
+      "step": 8467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.0261432393061685e-09,
+      "loss": 2.315,
+      "step": 8468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.558862447995304e-09,
+      "loss": 2.7094,
+      "step": 8469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.114375391317537e-09,
+      "loss": 2.2237,
+      "step": 8470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6926821368354896e-09,
+      "loss": 2.369,
+      "step": 8471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2937827485979283e-09,
+      "loss": 2.3934,
+      "step": 8472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.917677287256337e-09,
+      "loss": 2.5286,
+      "step": 8473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5643658099483435e-09,
+      "loss": 2.0695,
+      "step": 8474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.233848370364333e-09,
+      "loss": 2.2428,
+      "step": 8475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.926125018730795e-09,
+      "loss": 2.7602,
+      "step": 8476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6411958018269777e-09,
+      "loss": 2.3769,
+      "step": 8477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3790607629182716e-09,
+      "loss": 2.108,
+      "step": 8478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1397199418727853e-09,
+      "loss": 2.3603,
+      "step": 8479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.231733750447723e-10,
+      "loss": 2.5105,
+      "step": 8480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.294210953245894e-10,
+      "loss": 2.2865,
+      "step": 8481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.584631321886579e-10,
+      "loss": 2.2824,
+      "step": 8482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.1029951161619667e-10,
+      "loss": 2.6411,
+      "step": 8483
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8493025608922193e-10,
+      "loss": 1.9773,
+      "step": 8484
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.823553846758141e-10,
+      "loss": 2.11,
+      "step": 8485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0257491298015785e-10,
+      "loss": 1.9986,
+      "step": 8486
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5588853092581865e-11,
+      "loss": 1.8562,
+      "step": 8487
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1397213706132447e-11,
+      "loss": 2.6486,
+      "step": 8488
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0,
+      "loss": 2.7549,
+      "step": 8489
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1397213706132447e-11,
+      "loss": 2.4984,
+      "step": 8490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5588853092581865e-11,
+      "loss": 2.8088,
+      "step": 8491
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0257491298015785e-10,
+      "loss": 2.2933,
+      "step": 8492
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.823553846758141e-10,
+      "loss": 2.1761,
+      "step": 8493
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8493025608922193e-10,
+      "loss": 2.132,
+      "step": 8494
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.1029951159954335e-10,
+      "loss": 2.312,
+      "step": 8495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.584631321886579e-10,
+      "loss": 2.1976,
+      "step": 8496
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.294210953245894e-10,
+      "loss": 2.3844,
+      "step": 8497
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.231733750447723e-10,
+      "loss": 2.2832,
+      "step": 8498
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1397199418727853e-09,
+      "loss": 2.8098,
+      "step": 8499
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3790607629182716e-09,
+      "loss": 2.1798,
+      "step": 8500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6411958018269777e-09,
+      "loss": 2.4076,
+      "step": 8501
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.926125018730795e-09,
+      "loss": 1.9991,
+      "step": 8502
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.233848370364333e-09,
+      "loss": 2.4586,
+      "step": 8503
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5643658099483435e-09,
+      "loss": 2.5182,
+      "step": 8504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.917677287256337e-09,
+      "loss": 2.4474,
+      "step": 8505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2937827485979283e-09,
+      "loss": 1.9297,
+      "step": 8506
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.6926821368188364e-09,
+      "loss": 2.3623,
+      "step": 8507
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.114375391317537e-09,
+      "loss": 2.6773,
+      "step": 8508
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.558862447995304e-09,
+      "loss": 2.8854,
+      "step": 8509
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.0261432393061685e-09,
+      "loss": 2.4487,
+      "step": 8510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.516217694240266e-09,
+      "loss": 1.3269,
+      "step": 8511
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.029085738323836e-09,
+      "loss": 2.4383,
+      "step": 8512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.564747293635875e-09,
+      "loss": 2.0243,
+      "step": 8513
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.1232022787581775e-09,
+      "loss": 2.3073,
+      "step": 8514
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.70445060884195e-09,
+      "loss": 2.2245,
+      "step": 8515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.308492195541194e-09,
+      "loss": 2.4557,
+      "step": 8516
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.935326947079324e-09,
+      "loss": 2.4766,
+      "step": 8517
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584954768199205e-09,
+      "loss": 2.5378,
+      "step": 8518
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.025737556016315e-08,
+      "loss": 2.3413,
+      "step": 8519
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0952589220802888e-08,
+      "loss": 2.6593,
+      "step": 8520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1670595644486247e-08,
+      "loss": 2.3059,
+      "step": 8521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2411394722067203e-08,
+      "loss": 1.941,
+      "step": 8522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3174986341002447e-08,
+      "loss": 2.9225,
+      "step": 8523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.396137038525147e-08,
+      "loss": 2.8797,
+      "step": 8524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4770546735293209e-08,
+      "loss": 2.2192,
+      "step": 8525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.560251526817602e-08,
+      "loss": 2.8391,
+      "step": 8526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.64572758574677e-08,
+      "loss": 2.3191,
+      "step": 8527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.733482837328881e-08,
+      "loss": 2.6958,
+      "step": 8528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8235172682279364e-08,
+      "loss": 2.4547,
+      "step": 8529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915830864759882e-08,
+      "loss": 1.9919,
+      "step": 8530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.01042361289927e-08,
+      "loss": 2.1527,
+      "step": 8531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.107295498269268e-08,
+      "loss": 2.4297,
+      "step": 8532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.206446506151649e-08,
+      "loss": 2.7044,
+      "step": 8533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.307876621476801e-08,
+      "loss": 2.441,
+      "step": 8534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4115858288320523e-08,
+      "loss": 2.4377,
+      "step": 8535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5175741124566773e-08,
+      "loss": 2.2096,
+      "step": 8536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.625841456246891e-08,
+      "loss": 2.0434,
+      "step": 8537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7363878437458577e-08,
+      "loss": 2.6862,
+      "step": 8538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8492132581586778e-08,
+      "loss": 1.7199,
+      "step": 8539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9643176823374026e-08,
+      "loss": 2.59,
+      "step": 8540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.081701098792688e-08,
+      "loss": 2.8155,
+      "step": 8541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.201363489683806e-08,
+      "loss": 2.3202,
+      "step": 8542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.323304836828633e-08,
+      "loss": 2.6627,
+      "step": 8543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4475251216969925e-08,
+      "loss": 2.1096,
+      "step": 8544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.574024325408986e-08,
+      "loss": 2.8063,
+      "step": 8545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.702802428744989e-08,
+      "loss": 2.2841,
+      "step": 8546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.8338594121339885e-08,
+      "loss": 2.6541,
+      "step": 8547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.967195255660249e-08,
+      "loss": 2.3097,
+      "step": 8548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.102809939061647e-08,
+      "loss": 1.2769,
+      "step": 8549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.2407034417296646e-08,
+      "loss": 2.5321,
+      "step": 8550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3808757427094e-08,
+      "loss": 2.4064,
+      "step": 8551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.523326820701223e-08,
+      "loss": 2.603,
+      "step": 8552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.668056654055785e-08,
+      "loss": 2.4415,
+      "step": 8553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.815065220780678e-08,
+      "loss": 2.4353,
+      "step": 8554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.9643524985371055e-08,
+      "loss": 1.9736,
+      "step": 8555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.115918464636548e-08,
+      "loss": 2.5365,
+      "step": 8556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.2697630960490953e-08,
+      "loss": 2.5916,
+      "step": 8557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.4258863693934504e-08,
+      "loss": 2.5213,
+      "step": 8558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.584288260946923e-08,
+      "loss": 2.1145,
+      "step": 8559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.7449687466387674e-08,
+      "loss": 2.3483,
+      "step": 8560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.907927802048518e-08,
+      "loss": 2.3725,
+      "step": 8561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.073165402414315e-08,
+      "loss": 2.3026,
+      "step": 8562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.240681522626245e-08,
+      "loss": 2.7205,
+      "step": 8563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.410476137228004e-08,
+      "loss": 1.7156,
+      "step": 8564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.5825492204169e-08,
+      "loss": 2.1554,
+      "step": 8565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.756900746045513e-08,
+      "loss": 2.2804,
+      "step": 8566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.933530687616706e-08,
+      "loss": 2.8021,
+      "step": 8567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.112439018290284e-08,
+      "loss": 2.2476,
+      "step": 8568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.293625710877993e-08,
+      "loss": 2.4989,
+      "step": 8569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.477090737848523e-08,
+      "loss": 2.5101,
+      "step": 8570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.66283407131918e-08,
+      "loss": 2.0088,
+      "step": 8571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.850855683065871e-08,
+      "loss": 2.4276,
+      "step": 8572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.041155544516454e-08,
+      "loss": 2.5103,
+      "step": 8573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.233733626750726e-08,
+      "loss": 2.3041,
+      "step": 8574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.428589900503768e-08,
+      "loss": 1.6561,
+      "step": 8575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.625724336167595e-08,
+      "loss": 2.2888,
+      "step": 8576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.825136903782836e-08,
+      "loss": 2.3602,
+      "step": 8577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.026827573045403e-08,
+      "loss": 2.6215,
+      "step": 8578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.230796313308142e-08,
+      "loss": 2.711,
+      "step": 8579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.437043093572516e-08,
+      "loss": 2.6827,
+      "step": 8580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.64556788250026e-08,
+      "loss": 2.2636,
+      "step": 8581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.856370648401723e-08,
+      "loss": 2.4579,
+      "step": 8582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0069451359240866e-07,
+      "loss": 2.2606,
+      "step": 8583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0284809982640252e-07,
+      "loss": 2.2895,
+      "step": 8584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0502446485871064e-07,
+      "loss": 2.0024,
+      "step": 8585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0722360835863087e-07,
+      "loss": 2.0096,
+      "step": 8586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0944552999194722e-07,
+      "loss": 2.0783,
+      "step": 8587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1169022942102978e-07,
+      "loss": 2.3943,
+      "step": 8588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1395770630475143e-07,
+      "loss": 2.6999,
+      "step": 8589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1624796029855443e-07,
+      "loss": 2.4999,
+      "step": 8590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1856099105441719e-07,
+      "loss": 2.5705,
+      "step": 8591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2089679822080421e-07,
+      "loss": 1.5295,
+      "step": 8592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2325538144279945e-07,
+      "loss": 2.9135,
+      "step": 8593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.256367403619729e-07,
+      "loss": 2.2559,
+      "step": 8594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2804087461643076e-07,
+      "loss": 2.0523,
+      "step": 8595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3046778384084856e-07,
+      "loss": 2.2268,
+      "step": 8596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3291746766643795e-07,
+      "loss": 2.5569,
+      "step": 8597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3538992572091346e-07,
+      "loss": 2.5801,
+      "step": 8598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3788515762855888e-07,
+      "loss": 2.3415,
+      "step": 8599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4040316301021093e-07,
+      "loss": 2.418,
+      "step": 8600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.42943941483209e-07,
+      "loss": 3.1294,
+      "step": 8601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4550749266144524e-07,
+      "loss": 2.4668,
+      "step": 8602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4809381615538128e-07,
+      "loss": 2.4719,
+      "step": 8603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5070291157194824e-07,
+      "loss": 2.2623,
+      "step": 8604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.533347785147132e-07,
+      "loss": 2.5031,
+      "step": 8605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5598941658367948e-07,
+      "loss": 2.3677,
+      "step": 8606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5866682537548637e-07,
+      "loss": 2.7185,
+      "step": 8607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6136700448322603e-07,
+      "loss": 2.6421,
+      "step": 8608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6408995349659336e-07,
+      "loss": 2.5271,
+      "step": 8609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.668356720018027e-07,
+      "loss": 2.4635,
+      "step": 8610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6960415958160444e-07,
+      "loss": 2.5222,
+      "step": 8611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7239541581528515e-07,
+      "loss": 2.4103,
+      "step": 8612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.752094402786841e-07,
+      "loss": 2.1008,
+      "step": 8613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7804623254417671e-07,
+      "loss": 2.5104,
+      "step": 8614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.809057921806578e-07,
+      "loss": 2.6549,
+      "step": 8615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8378811875360832e-07,
+      "loss": 2.133,
+      "step": 8616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8669321182501196e-07,
+      "loss": 2.3311,
+      "step": 8617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8962107095337186e-07,
+      "loss": 2.0709,
+      "step": 8618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9257169569381058e-07,
+      "loss": 2.5553,
+      "step": 8619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.955450855979035e-07,
+      "loss": 2.5903,
+      "step": 8620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9854124021384533e-07,
+      "loss": 2.4013,
+      "step": 8621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0156015908628365e-07,
+      "loss": 2.1972,
+      "step": 8622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0460184175650207e-07,
+      "loss": 2.0936,
+      "step": 8623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0766628776223705e-07,
+      "loss": 2.513,
+      "step": 8624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1075349663784435e-07,
+      "loss": 2.8617,
+      "step": 8625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.138634679141493e-07,
+      "loss": 2.4289,
+      "step": 8626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1699620111857996e-07,
+      "loss": 2.3954,
+      "step": 8627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2015169577505042e-07,
+      "loss": 1.4175,
+      "step": 8628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2332995140406096e-07,
+      "loss": 2.2321,
+      "step": 8629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2653096752263123e-07,
+      "loss": 2.2136,
+      "step": 8630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2975474364433365e-07,
+      "loss": 2.3005,
+      "step": 8631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3300127927926016e-07,
+      "loss": 2.2316,
+      "step": 8632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3627057393405536e-07,
+      "loss": 2.3902,
+      "step": 8633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3956262711191667e-07,
+      "loss": 2.0648,
+      "step": 8634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4287743831257754e-07,
+      "loss": 2.6115,
+      "step": 8635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4621500703230767e-07,
+      "loss": 2.8221,
+      "step": 8636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4957533276391275e-07,
+      "loss": 2.4741,
+      "step": 8637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5295841499676786e-07,
+      "loss": 2.0192,
+      "step": 8638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5636425321673426e-07,
+      "loss": 2.1241,
+      "step": 8639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.59792846906276e-07,
+      "loss": 2.4011,
+      "step": 8640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6324419554439315e-07,
+      "loss": 2.0447,
+      "step": 8641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.667182986065719e-07,
+      "loss": 2.157,
+      "step": 8642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.702151555648846e-07,
+      "loss": 2.2254,
+      "step": 8643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7373476588795626e-07,
+      "loss": 2.6331,
+      "step": 8644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7727712904093146e-07,
+      "loss": 2.7387,
+      "step": 8645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8084224448549076e-07,
+      "loss": 2.2021,
+      "step": 8646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8443011167988417e-07,
+      "loss": 2.5854,
+      "step": 8647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.880407300788812e-07,
+      "loss": 2.2266,
+      "step": 8648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.916740991337874e-07,
+      "loss": 2.4807,
+      "step": 8649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.953302182924944e-07,
+      "loss": 2.2386,
+      "step": 8650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9900908699939665e-07,
+      "loss": 1.9463,
+      "step": 8651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0271070469544137e-07,
+      "loss": 2.6196,
+      "step": 8652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.064350708181118e-07,
+      "loss": 1.5733,
+      "step": 8653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1018218480144407e-07,
+      "loss": 2.3896,
+      "step": 8654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1395204607604364e-07,
+      "loss": 2.5791,
+      "step": 8655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.177446540689854e-07,
+      "loss": 2.7416,
+      "step": 8656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.215600082039804e-07,
+      "loss": 2.6806,
+      "step": 8657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.253981079012091e-07,
+      "loss": 2.9112,
+      "step": 8658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.292589525774214e-07,
+      "loss": 2.4175,
+      "step": 8659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3314254164591994e-07,
+      "loss": 2.4833,
+      "step": 8660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.370488745165434e-07,
+      "loss": 2.9432,
+      "step": 8661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4097795059568333e-07,
+      "loss": 2.4899,
+      "step": 8662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.449297692862507e-07,
+      "loss": 1.9198,
+      "step": 8663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.489043299877259e-07,
+      "loss": 2.7717,
+      "step": 8664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.529016320961087e-07,
+      "loss": 2.2433,
+      "step": 8665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.5692167500398496e-07,
+      "loss": 2.3406,
+      "step": 8666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.609644581004267e-07,
+      "loss": 2.2053,
+      "step": 8667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.650299807711088e-07,
+      "loss": 2.4428,
+      "step": 8668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.69118242398192e-07,
+      "loss": 2.4963,
+      "step": 8669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.732292423604566e-07,
+      "loss": 2.3688,
+      "step": 8670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.773629800331357e-07,
+      "loss": 1.8989,
+      "step": 8671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.8151945478808175e-07,
+      "loss": 2.0838,
+      "step": 8672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.8569866599365006e-07,
+      "loss": 2.2455,
+      "step": 8673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.899006130147819e-07,
+      "loss": 2.3286,
+      "step": 8674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.9412529521290483e-07,
+      "loss": 2.8623,
+      "step": 8675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.983727119460323e-07,
+      "loss": 2.4096,
+      "step": 8676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.0264286256873055e-07,
+      "loss": 1.3863,
+      "step": 8677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.0693574643210215e-07,
+      "loss": 2.5116,
+      "step": 8678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.112513628837688e-07,
+      "loss": 2.3615,
+      "step": 8679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.155897112679052e-07,
+      "loss": 1.966,
+      "step": 8680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.1995079092527196e-07,
+      "loss": 2.5968,
+      "step": 8681
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.2433460119313257e-07,
+      "loss": 2.4199,
+      "step": 8682
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.287411414053032e-07,
+      "loss": 2.4265,
+      "step": 8683
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3317041089216965e-07,
+      "loss": 2.2277,
+      "step": 8684
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3762240898063685e-07,
+      "loss": 2.4495,
+      "step": 8685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.420971349941793e-07,
+      "loss": 2.107,
+      "step": 8686
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.4659458825279105e-07,
+      "loss": 2.4161,
+      "step": 8687
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5111476807303537e-07,
+      "loss": 2.772,
+      "step": 8688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.556576737679951e-07,
+      "loss": 1.8543,
+      "step": 8689
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.602233046473558e-07,
+      "loss": 2.6128,
+      "step": 8690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.648116600172558e-07,
+      "loss": 1.9374,
+      "step": 8691
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.694227391804861e-07,
+      "loss": 2.657,
+      "step": 8692
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.7405654143630736e-07,
+      "loss": 2.2944,
+      "step": 8693
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.78713066080566e-07,
+      "loss": 2.3211,
+      "step": 8694
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.833923124056283e-07,
+      "loss": 2.5533,
+      "step": 8695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.880942797004461e-07,
+      "loss": 2.6142,
+      "step": 8696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.928189672504578e-07,
+      "loss": 2.3796,
+      "step": 8697
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.975663743377379e-07,
+      "loss": 2.2117,
+      "step": 8698
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.02336500240813e-07,
+      "loss": 2.4623,
+      "step": 8699
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.071293442348134e-07,
+      "loss": 2.9812,
+      "step": 8700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.119449055914216e-07,
+      "loss": 2.0273,
+      "step": 8701
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.167831835788394e-07,
+      "loss": 2.2181,
+      "step": 8702
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.216441774618052e-07,
+      "loss": 2.8848,
+      "step": 8703
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.265278865016598e-07,
+      "loss": 2.5471,
+      "step": 8704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.314343099562468e-07,
+      "loss": 2.1305,
+      "step": 8705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.36363447079996e-07,
+      "loss": 2.2403,
+      "step": 8706
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.413152971238233e-07,
+      "loss": 2.2465,
+      "step": 8707
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.462898593352638e-07,
+      "loss": 1.5101,
+      "step": 8708
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.51287132958339e-07,
+      "loss": 2.7304,
+      "step": 8709
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.56307117233673e-07,
+      "loss": 2.2814,
+      "step": 8710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.613498113984094e-07,
+      "loss": 2.573,
+      "step": 8711
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.664152146862444e-07,
+      "loss": 1.7757,
+      "step": 8712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.715033263274105e-07,
+      "loss": 2.2053,
+      "step": 8713
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.766141455487261e-07,
+      "loss": 2.3763,
+      "step": 8714
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.817476715735458e-07,
+      "loss": 1.7246,
+      "step": 8715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.869039036217271e-07,
+      "loss": 2.2391,
+      "step": 8716
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.9208284090973e-07,
+      "loss": 2.6261,
+      "step": 8717
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.972844826505674e-07,
+      "loss": 2.4751,
+      "step": 8718
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.025088280537549e-07,
+      "loss": 2.443,
+      "step": 8719
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.077558763253942e-07,
+      "loss": 2.2944,
+      "step": 8720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.130256266681399e-07,
+      "loss": 2.599,
+      "step": 8721
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.183180782811825e-07,
+      "loss": 2.2991,
+      "step": 8722
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.236332303602653e-07,
+      "loss": 2.532,
+      "step": 8723
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.289710820976679e-07,
+      "loss": 2.2877,
+      "step": 8724
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.343316326822556e-07,
+      "loss": 1.7084,
+      "step": 8725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.397148812994135e-07,
+      "loss": 2.0276,
+      "step": 8726
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.45120827131096e-07,
+      "loss": 2.794,
+      "step": 8727
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.505494693557934e-07,
+      "loss": 3.0669,
+      "step": 8728
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.560008071485489e-07,
+      "loss": 2.1523,
+      "step": 8729
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.614748396809588e-07,
+      "loss": 2.4367,
+      "step": 8730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.669715661212049e-07,
+      "loss": 0.9575,
+      "step": 8731
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.724909856339555e-07,
+      "loss": 2.8748,
+      "step": 8732
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.780330973804649e-07,
+      "loss": 2.4027,
+      "step": 8733
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.835979005185565e-07,
+      "loss": 2.4618,
+      "step": 8734
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.891853942025738e-07,
+      "loss": 2.6585,
+      "step": 8735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.94795577583429e-07,
+      "loss": 2.2896,
+      "step": 8736
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.004284498085877e-07,
+      "loss": 2.8231,
+      "step": 8737
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.06084010022051e-07,
+      "loss": 2.7142,
+      "step": 8738
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.117622573644066e-07,
+      "loss": 2.2327,
+      "step": 8739
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.174631909727446e-07,
+      "loss": 2.7421,
+      "step": 8740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.231868099807415e-07,
+      "loss": 2.2446,
+      "step": 8741
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.289331135186261e-07,
+      "loss": 2.2638,
+      "step": 8742
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.347021007131804e-07,
+      "loss": 2.1821,
+      "step": 8743
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.404937706877056e-07,
+      "loss": 2.4512,
+      "step": 8744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.463081225621226e-07,
+      "loss": 2.2407,
+      "step": 8745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.52145155452838e-07,
+      "loss": 2.6682,
+      "step": 8746
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.580048684728446e-07,
+      "loss": 2.2617,
+      "step": 8747
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.638872607316882e-07,
+      "loss": 2.5677,
+      "step": 8748
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.69792331335467e-07,
+      "loss": 2.5445,
+      "step": 8749
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.757200793868157e-07,
+      "loss": 2.3567,
+      "step": 8750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.816705039849547e-07,
+      "loss": 2.1502,
+      "step": 8751
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.87643604225624e-07,
+      "loss": 1.8032,
+      "step": 8752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.936393792011497e-07,
+      "loss": 2.1923,
+      "step": 8753
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.996578280003774e-07,
+      "loss": 1.4596,
+      "step": 8754
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.056989497087551e-07,
+      "loss": 2.4526,
+      "step": 8755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.117627434082174e-07,
+      "loss": 2.4266,
+      "step": 8756
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.17849208177318e-07,
+      "loss": 2.2725,
+      "step": 8757
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.239583430911467e-07,
+      "loss": 1.8362,
+      "step": 8758
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.300901472213295e-07,
+      "loss": 2.6671,
+      "step": 8759
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.362446196360617e-07,
+      "loss": 2.7886,
+      "step": 8760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.424217594000915e-07,
+      "loss": 2.3467,
+      "step": 8761
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.486215655747197e-07,
+      "loss": 2.8759,
+      "step": 8762
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.548440372178333e-07,
+      "loss": 2.4347,
+      "step": 8763
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.610891733838055e-07,
+      "loss": 2.754,
+      "step": 8764
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.673569731236285e-07,
+      "loss": 2.2853,
+      "step": 8765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.736474354848311e-07,
+      "loss": 2.525,
+      "step": 8766
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.799605595115111e-07,
+      "loss": 2.2088,
+      "step": 8767
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.862963442442861e-07,
+      "loss": 1.5123,
+      "step": 8768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.926547887203428e-07,
+      "loss": 2.3674,
+      "step": 8769
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.990358919734542e-07,
+      "loss": 2.8502,
+      "step": 8770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.054396530339292e-07,
+      "loss": 2.5644,
+      "step": 8771
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.118660709286296e-07,
+      "loss": 1.9539,
+      "step": 8772
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.1831514468097e-07,
+      "loss": 2.8236,
+      "step": 8773
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.247868733109343e-07,
+      "loss": 2.4943,
+      "step": 8774
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.312812558350758e-07,
+      "loss": 2.6812,
+      "step": 8775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.377982912664839e-07,
+      "loss": 2.2135,
+      "step": 8776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.443379786147843e-07,
+      "loss": 2.1555,
+      "step": 8777
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.509003168862383e-07,
+      "loss": 2.8037,
+      "step": 8778
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574853050835773e-07,
+      "loss": 2.0419,
+      "step": 8779
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.640929422061183e-07,
+      "loss": 2.3048,
+      "step": 8780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.707232272497646e-07,
+      "loss": 2.3484,
+      "step": 8781
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.773761592069718e-07,
+      "loss": 2.1497,
+      "step": 8782
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.840517370667157e-07,
+      "loss": 2.4582,
+      "step": 8783
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.907499598145907e-07,
+      "loss": 2.8193,
+      "step": 8784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.97470826432678e-07,
+      "loss": 2.3936,
+      "step": 8785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0042143358996613e-06,
+      "loss": 1.7489,
+      "step": 8786
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0109804871907933e-06,
+      "loss": 2.2841,
+      "step": 8787
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0177692792778802e-06,
+      "loss": 3.0059,
+      "step": 8788
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.024580711129247e-06,
+      "loss": 2.8728,
+      "step": 8789
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0314147817098218e-06,
+      "loss": 2.3611,
+      "step": 8790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0382714899810851e-06,
+      "loss": 1.6393,
+      "step": 8791
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0451508349010873e-06,
+      "loss": 2.2642,
+      "step": 8792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0520528154243812e-06,
+      "loss": 2.2657,
+      "step": 8793
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0589774305021559e-06,
+      "loss": 2.2694,
+      "step": 8794
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0659246790821197e-06,
+      "loss": 2.571,
+      "step": 8795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0728945601085503e-06,
+      "loss": 2.7869,
+      "step": 8796
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.079887072522262e-06,
+      "loss": 2.3223,
+      "step": 8797
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0869022152606877e-06,
+      "loss": 2.4139,
+      "step": 8798
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0939399872577637e-06,
+      "loss": 2.2355,
+      "step": 8799
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1010003874440288e-06,
+      "loss": 2.2321,
+      "step": 8800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1080834147465412e-06,
+      "loss": 2.1336,
+      "step": 8801
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1151890680889453e-06,
+      "loss": 2.6447,
+      "step": 8802
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1223173463914549e-06,
+      "loss": 2.3006,
+      "step": 8803
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1294682485708527e-06,
+      "loss": 2.5533,
+      "step": 8804
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1366417735404588e-06,
+      "loss": 2.4893,
+      "step": 8805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1438379202101445e-06,
+      "loss": 2.5573,
+      "step": 8806
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.151056687486368e-06,
+      "loss": 2.3926,
+      "step": 8807
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1582980742721737e-06,
+      "loss": 2.4441,
+      "step": 8808
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1655620794670918e-06,
+      "loss": 2.8244,
+      "step": 8809
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.172848701967305e-06,
+      "loss": 2.4595,
+      "step": 8810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1801579406654992e-06,
+      "loss": 2.185,
+      "step": 8811
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1874897944509465e-06,
+      "loss": 2.744,
+      "step": 8812
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.194844262209471e-06,
+      "loss": 2.5807,
+      "step": 8813
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2022213428234672e-06,
+      "loss": 2.391,
+      "step": 8814
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2096210351719149e-06,
+      "loss": 2.3043,
+      "step": 8815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2170433381303135e-06,
+      "loss": 1.1276,
+      "step": 8816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2244882505707488e-06,
+      "loss": 1.4955,
+      "step": 8817
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2319557713618588e-06,
+      "loss": 2.6208,
+      "step": 8818
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2394458993688849e-06,
+      "loss": 2.3732,
+      "step": 8819
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2469586334536207e-06,
+      "loss": 2.1209,
+      "step": 8820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2544939724743465e-06,
+      "loss": 2.3471,
+      "step": 8821
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2620519152860275e-06,
+      "loss": 2.3134,
+      "step": 8822
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2696324607401165e-06,
+      "loss": 3.0338,
+      "step": 8823
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2772356076846514e-06,
+      "loss": 2.3756,
+      "step": 8824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2848613549642394e-06,
+      "loss": 2.3314,
+      "step": 8825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2925097014200414e-06,
+      "loss": 2.7509,
+      "step": 8826
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3001806458898035e-06,
+      "loss": 2.4608,
+      "step": 8827
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3078741872078414e-06,
+      "loss": 2.7806,
+      "step": 8828
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3155903242049902e-06,
+      "loss": 2.6127,
+      "step": 8829
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.323329055708705e-06,
+      "loss": 2.0025,
+      "step": 8830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3310903805429762e-06,
+      "loss": 2.4162,
+      "step": 8831
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3388742975283807e-06,
+      "loss": 2.7953,
+      "step": 8832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3466808054820476e-06,
+      "loss": 2.4933,
+      "step": 8833
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3545099032176599e-06,
+      "loss": 2.3431,
+      "step": 8834
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3623615895455186e-06,
+      "loss": 2.645,
+      "step": 8835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3702358632724287e-06,
+      "loss": 1.8094,
+      "step": 8836
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3781327232018136e-06,
+      "loss": 2.0196,
+      "step": 8837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3860521681336335e-06,
+      "loss": 3.0416,
+      "step": 8838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3939941968644507e-06,
+      "loss": 2.3648,
+      "step": 8839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4019588081873478e-06,
+      "loss": 2.2798,
+      "step": 8840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4099460008919927e-06,
+      "loss": 2.7298,
+      "step": 8841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4179557737646397e-06,
+      "loss": 1.941,
+      "step": 8842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4259881255880956e-06,
+      "loss": 2.276,
+      "step": 8843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4340430551417704e-06,
+      "loss": 2.1673,
+      "step": 8844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4421205612015763e-06,
+      "loss": 2.6926,
+      "step": 8845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4502206425400454e-06,
+      "loss": 1.9985,
+      "step": 8846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4583432979262622e-06,
+      "loss": 2.1191,
+      "step": 8847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4664885261259141e-06,
+      "loss": 1.5741,
+      "step": 8848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4746563259011913e-06,
+      "loss": 2.0848,
+      "step": 8849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.482846696010903e-06,
+      "loss": 2.4986,
+      "step": 8850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.491059635210412e-06,
+      "loss": 2.1884,
+      "step": 8851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.499295142251683e-06,
+      "loss": 2.8289,
+      "step": 8852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5075532158832004e-06,
+      "loss": 2.3561,
+      "step": 8853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5158338548500515e-06,
+      "loss": 2.5878,
+      "step": 8854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.524137057893876e-06,
+      "loss": 2.5332,
+      "step": 8855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5324628237529001e-06,
+      "loss": 2.6574,
+      "step": 8856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5408111511619525e-06,
+      "loss": 2.2014,
+      "step": 8857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5491820388523479e-06,
+      "loss": 2.7532,
+      "step": 8858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5575754855520205e-06,
+      "loss": 2.1471,
+      "step": 8859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5659914899855242e-06,
+      "loss": 2.3246,
+      "step": 8860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5744300508739149e-06,
+      "loss": 2.3689,
+      "step": 8861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5828911669348354e-06,
+      "loss": 2.5482,
+      "step": 8862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.591374836882514e-06,
+      "loss": 2.0809,
+      "step": 8863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5998810594277488e-06,
+      "loss": 2.47,
+      "step": 8864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6084098332779404e-06,
+      "loss": 2.1783,
+      "step": 8865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6169611571369756e-06,
+      "loss": 2.5142,
+      "step": 8866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6255350297054104e-06,
+      "loss": 2.8151,
+      "step": 8867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6341314496803203e-06,
+      "loss": 2.7341,
+      "step": 8868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6427504157554005e-06,
+      "loss": 2.3876,
+      "step": 8869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.651391926620832e-06,
+      "loss": 2.82,
+      "step": 8870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6600559809634483e-06,
+      "loss": 2.1645,
+      "step": 8871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6687425774666529e-06,
+      "loss": 2.3797,
+      "step": 8872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6774517148104183e-06,
+      "loss": 2.3684,
+      "step": 8873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6861833916712364e-06,
+      "loss": 2.6044,
+      "step": 8874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6949376067222352e-06,
+      "loss": 2.749,
+      "step": 8875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.703714358633096e-06,
+      "loss": 2.7981,
+      "step": 8876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7125136460701016e-06,
+      "loss": 1.7795,
+      "step": 8877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7213354676960556e-06,
+      "loss": 2.4516,
+      "step": 8878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.73017982217038e-06,
+      "loss": 2.5382,
+      "step": 8879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.73904670814905e-06,
+      "loss": 2.246,
+      "step": 8880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7479361242846768e-06,
+      "loss": 2.5545,
+      "step": 8881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.756848069226341e-06,
+      "loss": 2.4731,
+      "step": 8882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7657825416197757e-06,
+      "loss": 1.5868,
+      "step": 8883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.774739540107284e-06,
+      "loss": 2.02,
+      "step": 8884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7837190633277543e-06,
+      "loss": 2.3773,
+      "step": 8885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7927211099165784e-06,
+      "loss": 2.4327,
+      "step": 8886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.801745678505817e-06,
+      "loss": 2.1827,
+      "step": 8887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8107927677240508e-06,
+      "loss": 2.1221,
+      "step": 8888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8198623761964958e-06,
+      "loss": 2.6261,
+      "step": 8889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8289545025448716e-06,
+      "loss": 2.2287,
+      "step": 8890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8380691453875163e-06,
+      "loss": 2.632,
+      "step": 8891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8472063033393546e-06,
+      "loss": 2.0824,
+      "step": 8892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.856365975011881e-06,
+      "loss": 1.5727,
+      "step": 8893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8655481590131583e-06,
+      "loss": 2.5583,
+      "step": 8894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8747528539478197e-06,
+      "loss": 2.4293,
+      "step": 8895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.883980058417117e-06,
+      "loss": 1.6607,
+      "step": 8896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8932297710188726e-06,
+      "loss": 2.5268,
+      "step": 8897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.902501990347427e-06,
+      "loss": 2.3086,
+      "step": 8898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911796714993774e-06,
+      "loss": 2.9294,
+      "step": 8899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.921113943545477e-06,
+      "loss": 2.8305,
+      "step": 8900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930453674586635e-06,
+      "loss": 2.0558,
+      "step": 8901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.939815906697967e-06,
+      "loss": 2.08,
+      "step": 8902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492006384567603e-06,
+      "loss": 2.5275,
+      "step": 8903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9586078684368733e-06,
+      "loss": 2.4394,
+      "step": 8904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.968037595208799e-06,
+      "loss": 2.8439,
+      "step": 8905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.977489817339517e-06,
+      "loss": 2.678,
+      "step": 8906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9869645333926597e-06,
+      "loss": 1.5591,
+      "step": 8907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.996461741928429e-06,
+      "loss": 2.2981,
+      "step": 8908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0059814415036126e-06,
+      "loss": 2.6669,
+      "step": 8909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.015523630671534e-06,
+      "loss": 2.4195,
+      "step": 8910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0250883079821533e-06,
+      "loss": 2.1824,
+      "step": 8911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0346754719820003e-06,
+      "loss": 2.3498,
+      "step": 8912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0442851212142064e-06,
+      "loss": 2.4116,
+      "step": 8913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0539172542184236e-06,
+      "loss": 2.1397,
+      "step": 8914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0635718695309223e-06,
+      "loss": 2.7506,
+      "step": 8915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0732489656845762e-06,
+      "loss": 2.7032,
+      "step": 8916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0829485412088455e-06,
+      "loss": 2.7986,
+      "step": 8917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0926705946297086e-06,
+      "loss": 2.1291,
+      "step": 8918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1024151244697973e-06,
+      "loss": 2.5381,
+      "step": 8919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1121821292483134e-06,
+      "loss": 2.0916,
+      "step": 8920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1219716074810433e-06,
+      "loss": 2.2547,
+      "step": 8921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1317835576803445e-06,
+      "loss": 2.7762,
+      "step": 8922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.141617978355126e-06,
+      "loss": 1.9854,
+      "step": 8923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.151474868010966e-06,
+      "loss": 2.2879,
+      "step": 8924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.161354225149964e-06,
+      "loss": 2.6616,
+      "step": 8925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1712560482708363e-06,
+      "loss": 2.2895,
+      "step": 8926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.181180335868854e-06,
+      "loss": 2.0853,
+      "step": 8927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1911270864359398e-06,
+      "loss": 2.0083,
+      "step": 8928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2010962984604864e-06,
+      "loss": 2.279,
+      "step": 8929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.211087970427605e-06,
+      "loss": 2.4281,
+      "step": 8930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2211021008189112e-06,
+      "loss": 2.0582,
+      "step": 8931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.231138688112605e-06,
+      "loss": 2.6212,
+      "step": 8932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.241197730783539e-06,
+      "loss": 1.7275,
+      "step": 8933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2512792273031045e-06,
+      "loss": 2.3726,
+      "step": 8934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2613831761392586e-06,
+      "loss": 2.1122,
+      "step": 8935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2715095757565972e-06,
+      "loss": 2.0886,
+      "step": 8936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.281658424616284e-06,
+      "loss": 2.8078,
+      "step": 8937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2918297211761028e-06,
+      "loss": 2.2733,
+      "step": 8938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3020234638903234e-06,
+      "loss": 2.3544,
+      "step": 8939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3122396512099185e-06,
+      "loss": 2.5225,
+      "step": 8940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3224782815823962e-06,
+      "loss": 2.4466,
+      "step": 8941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.332739353451901e-06,
+      "loss": 2.9961,
+      "step": 8942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3430228652590812e-06,
+      "loss": 2.0958,
+      "step": 8943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3533288154412356e-06,
+      "loss": 2.2409,
+      "step": 8944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3636572024322675e-06,
+      "loss": 2.3223,
+      "step": 8945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.374008024662649e-06,
+      "loss": 2.508,
+      "step": 8946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.384381280559389e-06,
+      "loss": 2.6794,
+      "step": 8947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3947769685461637e-06,
+      "loss": 2.7607,
+      "step": 8948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.405195087043238e-06,
+      "loss": 2.563,
+      "step": 8949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4156356344674445e-06,
+      "loss": 2.1162,
+      "step": 8950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4260986092321686e-06,
+      "loss": 2.5021,
+      "step": 8951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.43658400974745e-06,
+      "loss": 2.5867,
+      "step": 8952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4470918344198955e-06,
+      "loss": 2.4743,
+      "step": 8953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.457622081652716e-06,
+      "loss": 2.4259,
+      "step": 8954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4681747498457083e-06,
+      "loss": 1.8809,
+      "step": 8955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.478749837395222e-06,
+      "loss": 2.2548,
+      "step": 8956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4893473426942757e-06,
+      "loss": 2.6819,
+      "step": 8957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4999672641324243e-06,
+      "loss": 2.4007,
+      "step": 8958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.510609600095842e-06,
+      "loss": 2.3114,
+      "step": 8959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5212743489672727e-06,
+      "loss": 2.7212,
+      "step": 8960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5319615091261127e-06,
+      "loss": 2.4752,
+      "step": 8961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5426710789482607e-06,
+      "loss": 2.152,
+      "step": 8962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5534030568062856e-06,
+      "loss": 1.7018,
+      "step": 8963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.564157441069309e-06,
+      "loss": 2.3275,
+      "step": 8964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5749342301030873e-06,
+      "loss": 2.4499,
+      "step": 8965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5857334222699477e-06,
+      "loss": 2.371,
+      "step": 8966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5965550159287863e-06,
+      "loss": 2.2567,
+      "step": 8967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6073990094351517e-06,
+      "loss": 2.0204,
+      "step": 8968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6182654011411286e-06,
+      "loss": 2.5493,
+      "step": 8969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.629154189395488e-06,
+      "loss": 2.2966,
+      "step": 8970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6400653725434704e-06,
+      "loss": 2.409,
+      "step": 8971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.650998948927019e-06,
+      "loss": 2.2072,
+      "step": 8972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.661954916884612e-06,
+      "loss": 2.4571,
+      "step": 8973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6729332747513988e-06,
+      "loss": 2.421,
+      "step": 8974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.68393402085903e-06,
+      "loss": 2.3926,
+      "step": 8975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6949571535357927e-06,
+      "loss": 1.9043,
+      "step": 8976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7060026711066106e-06,
+      "loss": 2.3168,
+      "step": 8977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7170705718929932e-06,
+      "loss": 2.3675,
+      "step": 8978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7281608542129852e-06,
+      "loss": 2.7334,
+      "step": 8979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.739273516381285e-06,
+      "loss": 2.2302,
+      "step": 8980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7504085567091937e-06,
+      "loss": 2.5319,
+      "step": 8981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.761565973504615e-06,
+      "loss": 2.3748,
+      "step": 8982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.772745765072004e-06,
+      "loss": 2.5303,
+      "step": 8983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7839479297124545e-06,
+      "loss": 2.5328,
+      "step": 8984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.795172465723661e-06,
+      "loss": 2.7146,
+      "step": 8985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8064193713999207e-06,
+      "loss": 2.6925,
+      "step": 8986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.817688645032101e-06,
+      "loss": 2.6056,
+      "step": 8987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8289802849076894e-06,
+      "loss": 2.0637,
+      "step": 8988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.840294289310807e-06,
+      "loss": 2.3388,
+      "step": 8989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8516306565221303e-06,
+      "loss": 2.5233,
+      "step": 8990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.862989384818937e-06,
+      "loss": 1.991,
+      "step": 8991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8743704724751236e-06,
+      "loss": 1.7652,
+      "step": 8992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8857739177611918e-06,
+      "loss": 2.3855,
+      "step": 8993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.897199718944293e-06,
+      "loss": 2.5234,
+      "step": 8994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9086478742880503e-06,
+      "loss": 3.0535,
+      "step": 8995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920118382052805e-06,
+      "loss": 2.6193,
+      "step": 8996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9316112404954517e-06,
+      "loss": 2.3919,
+      "step": 8997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9431264478695206e-06,
+      "loss": 2.4292,
+      "step": 8998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.954664002425161e-06,
+      "loss": 2.163,
+      "step": 8999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9662239024090095e-06,
+      "loss": 2.3182,
+      "step": 9000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9778061460644543e-06,
+      "loss": 2.2061,
+      "step": 9001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.989410731631403e-06,
+      "loss": 2.325,
+      "step": 9002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.001037657346417e-06,
+      "loss": 2.446,
+      "step": 9003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0126869214425754e-06,
+      "loss": 1.0723,
+      "step": 9004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.024358522149678e-06,
+      "loss": 2.3,
+      "step": 9005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.036052457694044e-06,
+      "loss": 2.8163,
+      "step": 9006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.047768726298677e-06,
+      "loss": 2.368,
+      "step": 9007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0595073261830694e-06,
+      "loss": 2.6842,
+      "step": 9008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0712682555634305e-06,
+      "loss": 2.0458,
+      "step": 9009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0830515126525233e-06,
+      "loss": 2.0008,
+      "step": 9010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0948570956597808e-06,
+      "loss": 2.4758,
+      "step": 9011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.106685002791104e-06,
+      "loss": 2.5705,
+      "step": 9012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1185352322491473e-06,
+      "loss": 2.0944,
+      "step": 9013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.130407782233102e-06,
+      "loss": 2.2261,
+      "step": 9014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1423026509388107e-06,
+      "loss": 2.4546,
+      "step": 9015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1542198365586523e-06,
+      "loss": 2.501,
+      "step": 9016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1661593372816597e-06,
+      "loss": 2.9445,
+      "step": 9017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1781211512935e-06,
+      "loss": 2.488,
+      "step": 9018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1901052767764114e-06,
+      "loss": 2.4378,
+      "step": 9019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2021117119092497e-06,
+      "loss": 2.3071,
+      "step": 9020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2141404548674755e-06,
+      "loss": 2.6244,
+      "step": 9021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2261915038231834e-06,
+      "loss": 2.5477,
+      "step": 9022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2382648569450554e-06,
+      "loss": 2.6277,
+      "step": 9023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2503605123983754e-06,
+      "loss": 2.1208,
+      "step": 9024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2624784683450646e-06,
+      "loss": 2.3668,
+      "step": 9025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.274618722943645e-06,
+      "loss": 2.5657,
+      "step": 9026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2867812743492762e-06,
+      "loss": 2.6438,
+      "step": 9027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2989661207136363e-06,
+      "loss": 1.2377,
+      "step": 9028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.311173260185124e-06,
+      "loss": 2.3601,
+      "step": 9029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3234026909087055e-06,
+      "loss": 2.5307,
+      "step": 9030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.335654411025984e-06,
+      "loss": 2.7757,
+      "step": 9031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3479284186750998e-06,
+      "loss": 2.2044,
+      "step": 9032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3602247119908943e-06,
+      "loss": 2.1966,
+      "step": 9033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.372543289104779e-06,
+      "loss": 2.0995,
+      "step": 9034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3848841481448173e-06,
+      "loss": 2.2387,
+      "step": 9035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3972472872355937e-06,
+      "loss": 1.8064,
+      "step": 9036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.409632704498427e-06,
+      "loss": 2.3699,
+      "step": 9037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.422040398051157e-06,
+      "loss": 2.258,
+      "step": 9038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.434470366008324e-06,
+      "loss": 2.2874,
+      "step": 9039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.446922606480973e-06,
+      "loss": 2.1525,
+      "step": 9040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.459397117576851e-06,
+      "loss": 2.7466,
+      "step": 9041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4718938974003073e-06,
+      "loss": 3.0207,
+      "step": 9042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4844129440523115e-06,
+      "loss": 2.6951,
+      "step": 9043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.496954255630402e-06,
+      "loss": 1.833,
+      "step": 9044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.509517830228753e-06,
+      "loss": 1.6742,
+      "step": 9045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.5221036659381917e-06,
+      "loss": 2.361,
+      "step": 9046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.534711760846165e-06,
+      "loss": 2.9756,
+      "step": 9047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.5473421130366554e-06,
+      "loss": 2.7883,
+      "step": 9048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.5599947205903478e-06,
+      "loss": 2.6925,
+      "step": 9049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.572669581584514e-06,
+      "loss": 2.4257,
+      "step": 9050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.5853666940930615e-06,
+      "loss": 2.2408,
+      "step": 9051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.598086056186467e-06,
+      "loss": 2.4012,
+      "step": 9052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.6108276659318757e-06,
+      "loss": 2.564,
+      "step": 9053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.6235915213930377e-06,
+      "loss": 2.5373,
+      "step": 9054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.636377620630354e-06,
+      "loss": 2.0653,
+      "step": 9055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.6491859617007456e-06,
+      "loss": 2.3374,
+      "step": 9056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.662016542657853e-06,
+      "loss": 2.2674,
+      "step": 9057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.674869361551902e-06,
+      "loss": 2.4135,
+      "step": 9058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.687744416429789e-06,
+      "loss": 2.0081,
+      "step": 9059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.7006417053349115e-06,
+      "loss": 2.4943,
+      "step": 9060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.713561226307388e-06,
+      "loss": 2.5351,
+      "step": 9061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.726502977383922e-06,
+      "loss": 2.4479,
+      "step": 9062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.739466956597903e-06,
+      "loss": 2.0175,
+      "step": 9063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.7524531619792077e-06,
+      "loss": 2.5164,
+      "step": 9064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.765461591554464e-06,
+      "loss": 2.1596,
+      "step": 9065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.7784922433468535e-06,
+      "loss": 2.1651,
+      "step": 9066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.7915451153762265e-06,
+      "loss": 2.3953,
+      "step": 9067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.804620205659004e-06,
+      "loss": 2.0991,
+      "step": 9068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.817717512208257e-06,
+      "loss": 1.8809,
+      "step": 9069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.830837033033696e-06,
+      "loss": 2.1831,
+      "step": 9070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.843978766141648e-06,
+      "loss": 2.1529,
+      "step": 9071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.85714270953506e-06,
+      "loss": 2.2255,
+      "step": 9072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.870328861213484e-06,
+      "loss": 2.3017,
+      "step": 9073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.883537219173105e-06,
+      "loss": 2.6021,
+      "step": 9074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.896767781406762e-06,
+      "loss": 2.9018,
+      "step": 9075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.910020545903914e-06,
+      "loss": 2.4512,
+      "step": 9076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.923295510650604e-06,
+      "loss": 2.3547,
+      "step": 9077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.936592673629546e-06,
+      "loss": 1.5361,
+      "step": 9078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.949912032820057e-06,
+      "loss": 2.6944,
+      "step": 9079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.963253586198106e-06,
+      "loss": 2.3285,
+      "step": 9080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.976617331736248e-06,
+      "loss": 2.5556,
+      "step": 9081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.990003267403708e-06,
+      "loss": 2.3557,
+      "step": 9082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.003411391166328e-06,
+      "loss": 2.0563,
+      "step": 9083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.016841700986573e-06,
+      "loss": 2.1324,
+      "step": 9084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.0302941948234915e-06,
+      "loss": 2.3631,
+      "step": 9085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.043768870632835e-06,
+      "loss": 1.9255,
+      "step": 9086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.05726572636696e-06,
+      "loss": 2.5329,
+      "step": 9087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.070784759974871e-06,
+      "loss": 2.7796,
+      "step": 9088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.084325969402097e-06,
+      "loss": 2.0725,
+      "step": 9089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.097889352590949e-06,
+      "loss": 2.5256,
+      "step": 9090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.111474907480261e-06,
+      "loss": 2.374,
+      "step": 9091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.125082632005566e-06,
+      "loss": 2.3473,
+      "step": 9092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.138712524098969e-06,
+      "loss": 2.678,
+      "step": 9093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.152364581689211e-06,
+      "loss": 2.4051,
+      "step": 9094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.1660388027017324e-06,
+      "loss": 2.2478,
+      "step": 9095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.179735185058564e-06,
+      "loss": 2.9597,
+      "step": 9096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.19345372667832e-06,
+      "loss": 2.2439,
+      "step": 9097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.207194425476302e-06,
+      "loss": 2.4355,
+      "step": 9098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.220957279364462e-06,
+      "loss": 2.5675,
+      "step": 9099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.234742286251358e-06,
+      "loss": 2.0875,
+      "step": 9100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.248549444042132e-06,
+      "loss": 2.6311,
+      "step": 9101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.262378750638645e-06,
+      "loss": 2.7383,
+      "step": 9102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.276230203939346e-06,
+      "loss": 2.694,
+      "step": 9103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.290103801839368e-06,
+      "loss": 2.4004,
+      "step": 9104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.303999542230396e-06,
+      "loss": 2.4139,
+      "step": 9105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.3179174230007886e-06,
+      "loss": 2.1457,
+      "step": 9106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.331857442035552e-06,
+      "loss": 1.7998,
+      "step": 9107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.345819597216366e-06,
+      "loss": 2.2177,
+      "step": 9108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.359803886421426e-06,
+      "loss": 2.4921,
+      "step": 9109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.373810307525699e-06,
+      "loss": 2.282,
+      "step": 9110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.387838858400705e-06,
+      "loss": 2.6825,
+      "step": 9111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.401889536914649e-06,
+      "loss": 2.4167,
+      "step": 9112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.415962340932305e-06,
+      "loss": 2.874,
+      "step": 9113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.430057268315151e-06,
+      "loss": 2.2973,
+      "step": 9114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.444174316921301e-06,
+      "loss": 1.9855,
+      "step": 9115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.458313484605503e-06,
+      "loss": 2.1875,
+      "step": 9116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.472474769219059e-06,
+      "loss": 2.6091,
+      "step": 9117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.486658168610008e-06,
+      "loss": 2.4345,
+      "step": 9118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.50086368062304e-06,
+      "loss": 1.9539,
+      "step": 9119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.515091303099433e-06,
+      "loss": 2.6852,
+      "step": 9120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.529341033877065e-06,
+      "loss": 2.8208,
+      "step": 9121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.543612870790536e-06,
+      "loss": 2.4395,
+      "step": 9122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.557906811671081e-06,
+      "loss": 2.7777,
+      "step": 9123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.572222854346552e-06,
+      "loss": 2.2453,
+      "step": 9124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.586560996641409e-06,
+      "loss": 2.7974,
+      "step": 9125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.600921236376792e-06,
+      "loss": 2.0309,
+      "step": 9126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.615303571370482e-06,
+      "loss": 1.9863,
+      "step": 9127
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6297079994369434e-06,
+      "loss": 2.0606,
+      "step": 9128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.644134518387177e-06,
+      "loss": 2.4403,
+      "step": 9129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.658583126028903e-06,
+      "loss": 2.6064,
+      "step": 9130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.673053820166478e-06,
+      "loss": 2.529,
+      "step": 9131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.687546598600911e-06,
+      "loss": 2.5434,
+      "step": 9132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.702061459129813e-06,
+      "loss": 2.1462,
+      "step": 9133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.71659839954745e-06,
+      "loss": 2.317,
+      "step": 9134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.7311574176447874e-06,
+      "loss": 1.9197,
+      "step": 9135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.7457385112093795e-06,
+      "loss": 2.4062,
+      "step": 9136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.760341678025431e-06,
+      "loss": 2.4271,
+      "step": 9137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.774966915873834e-06,
+      "loss": 2.1105,
+      "step": 9138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.789614222532051e-06,
+      "loss": 2.1749,
+      "step": 9139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.804283595774261e-06,
+      "loss": 2.1429,
+      "step": 9140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.818975033371264e-06,
+      "loss": 1.905,
+      "step": 9141
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.833688533090496e-06,
+      "loss": 3.1132,
+      "step": 9142
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.848424092696063e-06,
+      "loss": 2.5917,
+      "step": 9143
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.863181709948705e-06,
+      "loss": 2.1208,
+      "step": 9144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.877961382605833e-06,
+      "loss": 2.2884,
+      "step": 9145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.892763108421427e-06,
+      "loss": 2.3456,
+      "step": 9146
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.907586885146236e-06,
+      "loss": 2.3148,
+      "step": 9147
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.9224327105275626e-06,
+      "loss": 2.266,
+      "step": 9148
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.937300582309428e-06,
+      "loss": 2.3949,
+      "step": 9149
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.952190498232422e-06,
+      "loss": 2.3227,
+      "step": 9150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.967102456033839e-06,
+      "loss": 2.3805,
+      "step": 9151
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.982036453447641e-06,
+      "loss": 2.6251,
+      "step": 9152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.996992488204427e-06,
+      "loss": 2.3808,
+      "step": 9153
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.011970558031381e-06,
+      "loss": 2.5196,
+      "step": 9154
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.026970660652407e-06,
+      "loss": 2.3289,
+      "step": 9155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.04199279378808e-06,
+      "loss": 2.1843,
+      "step": 9156
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.057036955155608e-06,
+      "loss": 2.087,
+      "step": 9157
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.07210314246877e-06,
+      "loss": 2.0698,
+      "step": 9158
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.087191353438097e-06,
+      "loss": 2.3777,
+      "step": 9159
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.10230158577074e-06,
+      "loss": 2.2882,
+      "step": 9160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.117433837170554e-06,
+      "loss": 1.6224,
+      "step": 9161
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.132588105337926e-06,
+      "loss": 2.2358,
+      "step": 9162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.147764387969983e-06,
+      "loss": 2.5114,
+      "step": 9163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.162962682760519e-06,
+      "loss": 2.4772,
+      "step": 9164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.178182987399965e-06,
+      "loss": 2.2944,
+      "step": 9165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.1934252995753875e-06,
+      "loss": 2.4366,
+      "step": 9166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.208689616970491e-06,
+      "loss": 2.4928,
+      "step": 9167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.223975937265695e-06,
+      "loss": 2.3959,
+      "step": 9168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.239284258138077e-06,
+      "loss": 2.6606,
+      "step": 9169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.2546145772612795e-06,
+      "loss": 2.0477,
+      "step": 9170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.269966892305699e-06,
+      "loss": 2.3098,
+      "step": 9171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.285341200938354e-06,
+      "loss": 2.5783,
+      "step": 9172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.300737500822927e-06,
+      "loss": 2.4808,
+      "step": 9173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.316155789619708e-06,
+      "loss": 3.1734,
+      "step": 9174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.331596064985721e-06,
+      "loss": 1.9547,
+      "step": 9175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.347058324574643e-06,
+      "loss": 2.0576,
+      "step": 9176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.36254256603677e-06,
+      "loss": 2.5729,
+      "step": 9177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.378048787019035e-06,
+      "loss": 2.4934,
+      "step": 9178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.393576985165105e-06,
+      "loss": 2.4178,
+      "step": 9179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.409127158115251e-06,
+      "loss": 2.3699,
+      "step": 9180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.42469930350648e-06,
+      "loss": 2.5076,
+      "step": 9181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.440293418972319e-06,
+      "loss": 1.5846,
+      "step": 9182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.4559095021430966e-06,
+      "loss": 2.5767,
+      "step": 9183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.471547550645711e-06,
+      "loss": 2.0026,
+      "step": 9184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.487207562103829e-06,
+      "loss": 2.3016,
+      "step": 9185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.502889534137639e-06,
+      "loss": 2.2683,
+      "step": 9186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.518593464364079e-06,
+      "loss": 2.0931,
+      "step": 9187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.53431935039676e-06,
+      "loss": 2.2568,
+      "step": 9188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.550067189845942e-06,
+      "loss": 2.5972,
+      "step": 9189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.5658369803184726e-06,
+      "loss": 2.5627,
+      "step": 9190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.581628719417985e-06,
+      "loss": 2.2443,
+      "step": 9191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.597442404744684e-06,
+      "loss": 2.4851,
+      "step": 9192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.61327803389554e-06,
+      "loss": 2.5359,
+      "step": 9193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.62913560446408e-06,
+      "loss": 2.2371,
+      "step": 9194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.645015114040529e-06,
+      "loss": 2.2525,
+      "step": 9195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.66091656021182e-06,
+      "loss": 2.6282,
+      "step": 9196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.676839940561518e-06,
+      "loss": 2.7048,
+      "step": 9197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.692785252669841e-06,
+      "loss": 2.4211,
+      "step": 9198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.708752494113744e-06,
+      "loss": 2.2032,
+      "step": 9199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.7247416624667365e-06,
+      "loss": 2.2597,
+      "step": 9200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.740752755299094e-06,
+      "loss": 1.4041,
+      "step": 9201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.75678577017773e-06,
+      "loss": 2.4372,
+      "step": 9202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.7728407046661926e-06,
+      "loss": 2.2928,
+      "step": 9203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.788917556324751e-06,
+      "loss": 1.0785,
+      "step": 9204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.805016322710343e-06,
+      "loss": 2.6629,
+      "step": 9205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.821137001376508e-06,
+      "loss": 2.4816,
+      "step": 9206
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.837279589873506e-06,
+      "loss": 2.8335,
+      "step": 9207
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.853444085748299e-06,
+      "loss": 2.2437,
+      "step": 9208
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.869630486544486e-06,
+      "loss": 2.2986,
+      "step": 9209
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.885838789802283e-06,
+      "loss": 1.845,
+      "step": 9210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.9020689930586615e-06,
+      "loss": 2.4459,
+      "step": 9211
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.918321093847228e-06,
+      "loss": 2.384,
+      "step": 9212
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.934595089698308e-06,
+      "loss": 2.2265,
+      "step": 9213
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.950890978138795e-06,
+      "loss": 2.1596,
+      "step": 9214
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.967208756692338e-06,
+      "loss": 2.0744,
+      "step": 9215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.983548422879253e-06,
+      "loss": 0.9839,
+      "step": 9216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.99990997421651e-06,
+      "loss": 2.6739,
+      "step": 9217
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.016293408217815e-06,
+      "loss": 2.367,
+      "step": 9218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.032698722393392e-06,
+      "loss": 2.1476,
+      "step": 9219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.0491259142503025e-06,
+      "loss": 2.0519,
+      "step": 9220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.065574981292226e-06,
+      "loss": 2.8227,
+      "step": 9221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.08204592101953e-06,
+      "loss": 1.5298,
+      "step": 9222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.0985387309291975e-06,
+      "loss": 2.497,
+      "step": 9223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.1150534085149525e-06,
+      "loss": 2.3536,
+      "step": 9224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.131589951267168e-06,
+      "loss": 1.2057,
+      "step": 9225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.148148356672955e-06,
+      "loss": 2.6854,
+      "step": 9226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.164728622216009e-06,
+      "loss": 2.3399,
+      "step": 9227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.181330745376712e-06,
+      "loss": 2.4266,
+      "step": 9228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.197954723632214e-06,
+      "loss": 2.2009,
+      "step": 9229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.21460055445629e-06,
+      "loss": 2.8015,
+      "step": 9230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.2312682353193425e-06,
+      "loss": 2.0812,
+      "step": 9231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.247957763688549e-06,
+      "loss": 2.1597,
+      "step": 9232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.264669137027706e-06,
+      "loss": 2.7354,
+      "step": 9233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.281402352797327e-06,
+      "loss": 2.2767,
+      "step": 9234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.2981574084545295e-06,
+      "loss": 2.5235,
+      "step": 9235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.314934301453201e-06,
+      "loss": 2.7888,
+      "step": 9236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.3317330292438796e-06,
+      "loss": 2.438,
+      "step": 9237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.348553589273808e-06,
+      "loss": 1.9535,
+      "step": 9238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.365395978986831e-06,
+      "loss": 2.1856,
+      "step": 9239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.382260195823563e-06,
+      "loss": 2.421,
+      "step": 9240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.399146237221253e-06,
+      "loss": 2.8483,
+      "step": 9241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.416054100613904e-06,
+      "loss": 1.8178,
+      "step": 9242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.432983783432056e-06,
+      "loss": 2.0862,
+      "step": 9243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.449935283103097e-06,
+      "loss": 2.5112,
+      "step": 9244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.46690859705099e-06,
+      "loss": 2.319,
+      "step": 9245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.483903722696499e-06,
+      "loss": 2.5723,
+      "step": 9246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.500920657456888e-06,
+      "loss": 2.7724,
+      "step": 9247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.5179593987462585e-06,
+      "loss": 2.5374,
+      "step": 9248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.535019943975367e-06,
+      "loss": 2.2239,
+      "step": 9249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.55210229055167e-06,
+      "loss": 2.4876,
+      "step": 9250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.5692064358792264e-06,
+      "loss": 2.2338,
+      "step": 9251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.586332377358866e-06,
+      "loss": 2.6454,
+      "step": 9252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.603480112388071e-06,
+      "loss": 1.1048,
+      "step": 9253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.620649638361075e-06,
+      "loss": 1.8191,
+      "step": 9254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.637840952668699e-06,
+      "loss": 2.4283,
+      "step": 9255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.6550540526985e-06,
+      "loss": 2.3921,
+      "step": 9256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.6722889358347356e-06,
+      "loss": 2.255,
+      "step": 9257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.689545599458368e-06,
+      "loss": 2.6094,
+      "step": 9258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.7068240409469785e-06,
+      "loss": 2.3387,
+      "step": 9259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.7241242576749175e-06,
+      "loss": 2.2436,
+      "step": 9260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.7414462470132045e-06,
+      "loss": 2.4542,
+      "step": 9261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.758790006329529e-06,
+      "loss": 1.9054,
+      "step": 9262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.776155532988247e-06,
+      "loss": 2.279,
+      "step": 9263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.793542824350489e-06,
+      "loss": 2.4292,
+      "step": 9264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.810951877774034e-06,
+      "loss": 2.5227,
+      "step": 9265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.828382690613349e-06,
+      "loss": 2.5,
+      "step": 9266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.845835260219568e-06,
+      "loss": 2.209,
+      "step": 9267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.863309583940563e-06,
+      "loss": 2.3604,
+      "step": 9268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.880805659120892e-06,
+      "loss": 2.3048,
+      "step": 9269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.8983234831018475e-06,
+      "loss": 2.6768,
+      "step": 9270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.915863053221277e-06,
+      "loss": 2.4237,
+      "step": 9271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.9334243668138615e-06,
+      "loss": 1.8745,
+      "step": 9272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.9510074212109524e-06,
+      "loss": 1.6165,
+      "step": 9273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.968612213740588e-06,
+      "loss": 2.6977,
+      "step": 9274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.98623874172744e-06,
+      "loss": 2.4416,
+      "step": 9275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.003887002492953e-06,
+      "loss": 1.7943,
+      "step": 9276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.021556993355271e-06,
+      "loss": 2.5747,
+      "step": 9277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.039248711629209e-06,
+      "loss": 2.077,
+      "step": 9278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.05696215462625e-06,
+      "loss": 2.8547,
+      "step": 9279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.074697319654615e-06,
+      "loss": 2.2456,
+      "step": 9280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.0924542040192435e-06,
+      "loss": 2.0934,
+      "step": 9281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.110232805021776e-06,
+      "loss": 2.7448,
+      "step": 9282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.128033119960441e-06,
+      "loss": 2.6572,
+      "step": 9283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.145855146130286e-06,
+      "loss": 2.3871,
+      "step": 9284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.16369888082306e-06,
+      "loss": 2.3378,
+      "step": 9285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.181564321327149e-06,
+      "loss": 2.1226,
+      "step": 9286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.199451464927674e-06,
+      "loss": 2.4326,
+      "step": 9287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.217360308906461e-06,
+      "loss": 2.3472,
+      "step": 9288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.235290850542019e-06,
+      "loss": 2.0914,
+      "step": 9289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.253243087109562e-06,
+      "loss": 2.1865,
+      "step": 9290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.2712170158810705e-06,
+      "loss": 2.0581,
+      "step": 9291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.289212634125097e-06,
+      "loss": 2.533,
+      "step": 9292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.307229939107029e-06,
+      "loss": 2.2874,
+      "step": 9293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.325268928088906e-06,
+      "loss": 2.8927,
+      "step": 9294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.343329598329456e-06,
+      "loss": 2.8028,
+      "step": 9295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.361411947084123e-06,
+      "loss": 2.933,
+      "step": 9296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.379515971605054e-06,
+      "loss": 3.0838,
+      "step": 9297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.3976416691411356e-06,
+      "loss": 1.9225,
+      "step": 9298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.415789036937969e-06,
+      "loss": 2.1255,
+      "step": 9299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.433958072237744e-06,
+      "loss": 2.8505,
+      "step": 9300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.4521487722794706e-06,
+      "loss": 2.703,
+      "step": 9301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.470361134298858e-06,
+      "loss": 2.1906,
+      "step": 9302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.488595155528338e-06,
+      "loss": 2.7967,
+      "step": 9303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.506850833196926e-06,
+      "loss": 2.9889,
+      "step": 9304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.525128164530492e-06,
+      "loss": 1.92,
+      "step": 9305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.543427146751541e-06,
+      "loss": 2.2371,
+      "step": 9306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.56174777707938e-06,
+      "loss": 1.7724,
+      "step": 9307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.580090052729854e-06,
+      "loss": 2.5417,
+      "step": 9308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.598453970915641e-06,
+      "loss": 2.2331,
+      "step": 9309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.616839528846142e-06,
+      "loss": 2.1604,
+      "step": 9310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.635246723727456e-06,
+      "loss": 2.4277,
+      "step": 9311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.65367555276229e-06,
+      "loss": 1.7894,
+      "step": 9312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.672126013150198e-06,
+      "loss": 2.2998,
+      "step": 9313
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.690598102087392e-06,
+      "loss": 2.1631,
+      "step": 9314
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.709091816766816e-06,
+      "loss": 2.1936,
+      "step": 9315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.727607154378102e-06,
+      "loss": 2.3252,
+      "step": 9316
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.746144112107566e-06,
+      "loss": 2.3141,
+      "step": 9317
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.76470268713833e-06,
+      "loss": 2.3734,
+      "step": 9318
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.783282876650165e-06,
+      "loss": 2.5997,
+      "step": 9319
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.801884677819565e-06,
+      "loss": 2.0846,
+      "step": 9320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.820508087819755e-06,
+      "loss": 2.6574,
+      "step": 9321
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.839153103820667e-06,
+      "loss": 2.5495,
+      "step": 9322
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.857819722988967e-06,
+      "loss": 2.4414,
+      "step": 9323
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.876507942487991e-06,
+      "loss": 2.3203,
+      "step": 9324
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.895217759477828e-06,
+      "loss": 2.3147,
+      "step": 9325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.913949171115302e-06,
+      "loss": 1.9988,
+      "step": 9326
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.932702174553956e-06,
+      "loss": 2.2152,
+      "step": 9327
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.951476766943954e-06,
+      "loss": 2.8859,
+      "step": 9328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.97027294543231e-06,
+      "loss": 2.0312,
+      "step": 9329
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.989090707162694e-06,
+      "loss": 2.1601,
+      "step": 9330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.00793004927554e-06,
+      "loss": 2.5555,
+      "step": 9331
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.026790968907892e-06,
+      "loss": 2.3272,
+      "step": 9332
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.045673463193624e-06,
+      "loss": 2.4582,
+      "step": 9333
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.064577529263317e-06,
+      "loss": 2.6281,
+      "step": 9334
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.08350316424428e-06,
+      "loss": 2.3809,
+      "step": 9335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.102450365260438e-06,
+      "loss": 2.9938,
+      "step": 9336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.121419129432572e-06,
+      "loss": 2.1282,
+      "step": 9337
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.140409453878121e-06,
+      "loss": 2.7996,
+      "step": 9338
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.15942133571133e-06,
+      "loss": 2.6087,
+      "step": 9339
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.178454772042975e-06,
+      "loss": 2.5058,
+      "step": 9340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.197509759980753e-06,
+      "loss": 2.1835,
+      "step": 9341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.216586296628997e-06,
+      "loss": 2.3884,
+      "step": 9342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.235684379088841e-06,
+      "loss": 2.254,
+      "step": 9343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.254804004457977e-06,
+      "loss": 2.4895,
+      "step": 9344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.273945169831009e-06,
+      "loss": 2.2531,
+      "step": 9345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.293107872299165e-06,
+      "loss": 2.9139,
+      "step": 9346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.312292108950441e-06,
+      "loss": 2.4095,
+      "step": 9347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.331497876869537e-06,
+      "loss": 1.9912,
+      "step": 9348
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.350725173137901e-06,
+      "loss": 1.7516,
+      "step": 9349
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.369973994833656e-06,
+      "loss": 2.7661,
+      "step": 9350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.389244339031759e-06,
+      "loss": 2.4174,
+      "step": 9351
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.408536202803783e-06,
+      "loss": 2.1816,
+      "step": 9352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.427849583218093e-06,
+      "loss": 2.1199,
+      "step": 9353
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.447184477339785e-06,
+      "loss": 2.3344,
+      "step": 9354
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.466540882230728e-06,
+      "loss": 2.3881,
+      "step": 9355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.485918794949337e-06,
+      "loss": 2.4095,
+      "step": 9356
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.505318212550987e-06,
+      "loss": 2.4221,
+      "step": 9357
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.524739132087637e-06,
+      "loss": 2.8922,
+      "step": 9358
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.544181550608114e-06,
+      "loss": 2.3104,
+      "step": 9359
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.56364546515778e-06,
+      "loss": 1.8569,
+      "step": 9360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.583130872778888e-06,
+      "loss": 2.3537,
+      "step": 9361
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.602637770510384e-06,
+      "loss": 1.5231,
+      "step": 9362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.622166155387961e-06,
+      "loss": 2.6109,
+      "step": 9363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.641716024444056e-06,
+      "loss": 2.2993,
+      "step": 9364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.661287374707732e-06,
+      "loss": 2.1719,
+      "step": 9365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.680880203204899e-06,
+      "loss": 2.3029,
+      "step": 9366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.700494506958205e-06,
+      "loss": 2.7214,
+      "step": 9367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.72013028298702e-06,
+      "loss": 2.3456,
+      "step": 9368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.73978752830738e-06,
+      "loss": 2.7052,
+      "step": 9369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.759466239932123e-06,
+      "loss": 2.2426,
+      "step": 9370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.779166414870824e-06,
+      "loss": 2.4171,
+      "step": 9371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.798888050129848e-06,
+      "loss": 2.7502,
+      "step": 9372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.818631142712157e-06,
+      "loss": 2.2947,
+      "step": 9373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.83839568961755e-06,
+      "loss": 2.1853,
+      "step": 9374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.858181687842563e-06,
+      "loss": 2.594,
+      "step": 9375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.877989134380486e-06,
+      "loss": 1.6801,
+      "step": 9376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.897818026221293e-06,
+      "loss": 2.372,
+      "step": 9377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.917668360351709e-06,
+      "loss": 2.4205,
+      "step": 9378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.93754013375525e-06,
+      "loss": 2.6617,
+      "step": 9379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.957433343412163e-06,
+      "loss": 2.6549,
+      "step": 9380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.977347986299382e-06,
+      "loss": 2.4714,
+      "step": 9381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.99728405939063e-06,
+      "loss": 2.4827,
+      "step": 9382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.017241559656379e-06,
+      "loss": 1.8557,
+      "step": 9383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.037220484063839e-06,
+      "loss": 2.4679,
+      "step": 9384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.057220829576906e-06,
+      "loss": 2.9241,
+      "step": 9385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.077242593156309e-06,
+      "loss": 2.2932,
+      "step": 9386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.097285771759466e-06,
+      "loss": 2.4279,
+      "step": 9387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.11735036234063e-06,
+      "loss": 2.5153,
+      "step": 9388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.137436361850608e-06,
+      "loss": 2.1596,
+      "step": 9389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.15754376723714e-06,
+      "loss": 2.4496,
+      "step": 9390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.177672575444655e-06,
+      "loss": 2.0737,
+      "step": 9391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.197822783414333e-06,
+      "loss": 1.8453,
+      "step": 9392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.217994388084038e-06,
+      "loss": 2.496,
+      "step": 9393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.238187386388457e-06,
+      "loss": 2.1479,
+      "step": 9394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.258401775259011e-06,
+      "loss": 2.6056,
+      "step": 9395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.278637551623907e-06,
+      "loss": 2.52,
+      "step": 9396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.298894712407972e-06,
+      "loss": 2.5419,
+      "step": 9397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.31917325453292e-06,
+      "loss": 2.4957,
+      "step": 9398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.339473174917161e-06,
+      "loss": 3.1203,
+      "step": 9399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.3597944704759e-06,
+      "loss": 2.1636,
+      "step": 9400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.380137138120975e-06,
+      "loss": 2.8162,
+      "step": 9401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.400501174761105e-06,
+      "loss": 2.3177,
+      "step": 9402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.420886577301683e-06,
+      "loss": 2.3859,
+      "step": 9403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.441293342644973e-06,
+      "loss": 2.5176,
+      "step": 9404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.461721467689786e-06,
+      "loss": 2.6293,
+      "step": 9405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.482170949331857e-06,
+      "loss": 2.3281,
+      "step": 9406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502641784463638e-06,
+      "loss": 1.8238,
+      "step": 9407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523133969974317e-06,
+      "loss": 2.0376,
+      "step": 9408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543647502749817e-06,
+      "loss": 2.3873,
+      "step": 9409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.564182379672903e-06,
+      "loss": 2.2637,
+      "step": 9410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.584738597622948e-06,
+      "loss": 2.7451,
+      "step": 9411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.605316153476256e-06,
+      "loss": 2.4895,
+      "step": 9412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.625915044105742e-06,
+      "loss": 2.1918,
+      "step": 9413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.646535266381161e-06,
+      "loss": 2.5101,
+      "step": 9414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66717681716902e-06,
+      "loss": 2.4402,
+      "step": 9415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687839693332594e-06,
+      "loss": 2.6117,
+      "step": 9416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708523891731813e-06,
+      "loss": 2.4989,
+      "step": 9417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.729229409223505e-06,
+      "loss": 2.5602,
+      "step": 9418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.749956242661173e-06,
+      "loss": 2.5009,
+      "step": 9419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.770704388895168e-06,
+      "loss": 2.5652,
+      "step": 9420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.791473844772462e-06,
+      "loss": 2.1283,
+      "step": 9421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.8122646071369e-06,
+      "loss": 2.2633,
+      "step": 9422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.833076672829055e-06,
+      "loss": 2.644,
+      "step": 9423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.853910038686325e-06,
+      "loss": 2.455,
+      "step": 9424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.874764701542725e-06,
+      "loss": 2.3206,
+      "step": 9425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.89564065822916e-06,
+      "loss": 2.1586,
+      "step": 9426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.916537905573246e-06,
+      "loss": 2.3806,
+      "step": 9427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.937456440399443e-06,
+      "loss": 2.2688,
+      "step": 9428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.958396259528811e-06,
+      "loss": 2.8696,
+      "step": 9429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.979357359779312e-06,
+      "loss": 2.5665,
+      "step": 9430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0000339737965658e-05,
+      "loss": 2.7349,
+      "step": 9431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.002134339089935e-05,
+      "loss": 2.7545,
+      "step": 9432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.004236831538851e-05,
+      "loss": 2.6438,
+      "step": 9433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.006341450823816e-05,
+      "loss": 2.2356,
+      "step": 9434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0084481966250108e-05,
+      "loss": 1.8636,
+      "step": 9435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0105570686222848e-05,
+      "loss": 2.4002,
+      "step": 9436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0126680664951692e-05,
+      "loss": 1.1835,
+      "step": 9437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0147811899228725e-05,
+      "loss": 2.6399,
+      "step": 9438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0168964385842732e-05,
+      "loss": 2.9037,
+      "step": 9439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0190138121579333e-05,
+      "loss": 2.2785,
+      "step": 9440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.021133310322097e-05,
+      "loss": 3.1583,
+      "step": 9441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0232549327546702e-05,
+      "loss": 2.4001,
+      "step": 9442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0253786791332508e-05,
+      "loss": 2.2695,
+      "step": 9443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0275045491351103e-05,
+      "loss": 2.5115,
+      "step": 9444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0296325424371904e-05,
+      "loss": 2.338,
+      "step": 9445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0317626587161149e-05,
+      "loss": 1.9325,
+      "step": 9446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0338948976481875e-05,
+      "loss": 2.6789,
+      "step": 9447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0360292589093877e-05,
+      "loss": 2.4356,
+      "step": 9448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0381657421753747e-05,
+      "loss": 1.3065,
+      "step": 9449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.040304347121475e-05,
+      "loss": 1.8825,
+      "step": 9450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0424450734227035e-05,
+      "loss": 2.0214,
+      "step": 9451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0445879207537506e-05,
+      "loss": 2.3004,
+      "step": 9452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0467328887889865e-05,
+      "loss": 2.8523,
+      "step": 9453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0488799772024454e-05,
+      "loss": 2.4807,
+      "step": 9454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0510291856678566e-05,
+      "loss": 2.6293,
+      "step": 9455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0531805138586196e-05,
+      "loss": 2.3883,
+      "step": 9456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0553339614478162e-05,
+      "loss": 2.8725,
+      "step": 9457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0574895281081947e-05,
+      "loss": 2.7029,
+      "step": 9458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0596472135121919e-05,
+      "loss": 1.9619,
+      "step": 9459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0618070173319221e-05,
+      "loss": 2.7736,
+      "step": 9460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0639689392391776e-05,
+      "loss": 2.4707,
+      "step": 9461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0661329789054179e-05,
+      "loss": 2.2957,
+      "step": 9462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0682991360017945e-05,
+      "loss": 2.5863,
+      "step": 9463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0704674101991307e-05,
+      "loss": 2.0965,
+      "step": 9464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0726378011679366e-05,
+      "loss": 2.1321,
+      "step": 9465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0748103085783827e-05,
+      "loss": 1.2246,
+      "step": 9466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0769849321003316e-05,
+      "loss": 2.8901,
+      "step": 9467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.079161671403324e-05,
+      "loss": 2.4611,
+      "step": 9468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0813405261565782e-05,
+      "loss": 2.2338,
+      "step": 9469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0835214960289823e-05,
+      "loss": 2.472,
+      "step": 9470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.085704580689118e-05,
+      "loss": 2.3303,
+      "step": 9471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0878897798052305e-05,
+      "loss": 2.4807,
+      "step": 9472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0900770930452557e-05,
+      "loss": 2.4327,
+      "step": 9473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0922665200767993e-05,
+      "loss": 2.4868,
+      "step": 9474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.094458060567151e-05,
+      "loss": 2.5861,
+      "step": 9475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0966517141832786e-05,
+      "loss": 2.7557,
+      "step": 9476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.098847480591834e-05,
+      "loss": 2.3167,
+      "step": 9477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1010453594591307e-05,
+      "loss": 1.9943,
+      "step": 9478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.103245350451179e-05,
+      "loss": 2.3944,
+      "step": 9479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1054474532336632e-05,
+      "loss": 2.0691,
+      "step": 9480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1076516674719476e-05,
+      "loss": 2.0836,
+      "step": 9481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1098579928310653e-05,
+      "loss": 2.4961,
+      "step": 9482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1120664289757408e-05,
+      "loss": 2.1421,
+      "step": 9483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1142769755703745e-05,
+      "loss": 2.0755,
+      "step": 9484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1164896322790502e-05,
+      "loss": 2.2328,
+      "step": 9485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1187043987655181e-05,
+      "loss": 2.186,
+      "step": 9486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1209212746932178e-05,
+      "loss": 2.2531,
+      "step": 9487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1231402597252687e-05,
+      "loss": 2.213,
+      "step": 9488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1253613535244721e-05,
+      "loss": 1.7626,
+      "step": 9489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1275845557532932e-05,
+      "loss": 2.6047,
+      "step": 9490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1298098660738936e-05,
+      "loss": 2.8119,
+      "step": 9491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.132037284148109e-05,
+      "loss": 2.5933,
+      "step": 9492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1342668096374602e-05,
+      "loss": 2.0676,
+      "step": 9493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1364984422031315e-05,
+      "loss": 2.8929,
+      "step": 9494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1387321815060024e-05,
+      "loss": 2.9773,
+      "step": 9495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.140968027206628e-05,
+      "loss": 2.7679,
+      "step": 9496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1432059789652448e-05,
+      "loss": 2.528,
+      "step": 9497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1454460364417617e-05,
+      "loss": 2.1424,
+      "step": 9498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1476881992957793e-05,
+      "loss": 2.3731,
+      "step": 9499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1499324671865684e-05,
+      "loss": 2.8598,
+      "step": 9500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1521788397730852e-05,
+      "loss": 2.3644,
+      "step": 9501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1544273167139628e-05,
+      "loss": 2.19,
+      "step": 9502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1566778976675162e-05,
+      "loss": 2.3228,
+      "step": 9503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1589305822917423e-05,
+      "loss": 2.1701,
+      "step": 9504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1611853702443213e-05,
+      "loss": 1.9953,
+      "step": 9505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1634422611825994e-05,
+      "loss": 2.9206,
+      "step": 9506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1657012547636174e-05,
+      "loss": 2.5918,
+      "step": 9507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1679623506440933e-05,
+      "loss": 2.36,
+      "step": 9508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1702255484804218e-05,
+      "loss": 2.6079,
+      "step": 9509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1724908479286899e-05,
+      "loss": 1.9969,
+      "step": 9510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1747582486446444e-05,
+      "loss": 3.5932,
+      "step": 9511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1770277502837294e-05,
+      "loss": 1.9601,
+      "step": 9512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1792993525010641e-05,
+      "loss": 2.3448,
+      "step": 9513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1815730549514562e-05,
+      "loss": 2.1521,
+      "step": 9514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1838488572893757e-05,
+      "loss": 2.2037,
+      "step": 9515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1861267591689904e-05,
+      "loss": 2.7886,
+      "step": 9516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.188406760244146e-05,
+      "loss": 2.327,
+      "step": 9517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1906888601683711e-05,
+      "loss": 1.9608,
+      "step": 9518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1929730585948583e-05,
+      "loss": 2.1727,
+      "step": 9519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1952593551765033e-05,
+      "loss": 2.5709,
+      "step": 9520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1975477495658725e-05,
+      "loss": 2.4149,
+      "step": 9521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1998382414152192e-05,
+      "loss": 2.6206,
+      "step": 9522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2021308303764648e-05,
+      "loss": 2.3733,
+      "step": 9523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.204425516101225e-05,
+      "loss": 2.5275,
+      "step": 9524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.206722298240792e-05,
+      "loss": 2.7334,
+      "step": 9525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2090211764461482e-05,
+      "loss": 2.5832,
+      "step": 9526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2113221503679382e-05,
+      "loss": 2.5652,
+      "step": 9527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2136252196565032e-05,
+      "loss": 3.2398,
+      "step": 9528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2159303839618635e-05,
+      "loss": 2.7967,
+      "step": 9529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2182376429337221e-05,
+      "loss": 1.5104,
+      "step": 9530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2205469962214565e-05,
+      "loss": 1.7674,
+      "step": 9531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.222858443474134e-05,
+      "loss": 2.1411,
+      "step": 9532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2251719843404989e-05,
+      "loss": 1.9926,
+      "step": 9533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2274876184689808e-05,
+      "loss": 2.3256,
+      "step": 9534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2298053455076845e-05,
+      "loss": 2.4924,
+      "step": 9535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2321251651044068e-05,
+      "loss": 1.9901,
+      "step": 9536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2344470769066195e-05,
+      "loss": 2.3712,
+      "step": 9537
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2367710805614833e-05,
+      "loss": 2.5147,
+      "step": 9538
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2390971757158258e-05,
+      "loss": 2.4905,
+      "step": 9539
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2414253620161712e-05,
+      "loss": 2.3021,
+      "step": 9540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2437556391087245e-05,
+      "loss": 2.5704,
+      "step": 9541
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2460880066393719e-05,
+      "loss": 1.8791,
+      "step": 9542
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2484224642536723e-05,
+      "loss": 2.3974,
+      "step": 9543
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2507590115968792e-05,
+      "loss": 2.4789,
+      "step": 9544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2530976483139233e-05,
+      "loss": 2.4467,
+      "step": 9545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2554383740494256e-05,
+      "loss": 2.2715,
+      "step": 9546
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2577811884476723e-05,
+      "loss": 2.2924,
+      "step": 9547
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2601260911526468e-05,
+      "loss": 2.47,
+      "step": 9548
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2624730818080102e-05,
+      "loss": 2.2094,
+      "step": 9549
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2648221600571135e-05,
+      "loss": 2.5133,
+      "step": 9550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.267173325542975e-05,
+      "loss": 2.098,
+      "step": 9551
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2695265779083074e-05,
+      "loss": 2.5875,
+      "step": 9552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2718819167955052e-05,
+      "loss": 1.9996,
+      "step": 9553
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2742393418466495e-05,
+      "loss": 2.1639,
+      "step": 9554
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2765988527034904e-05,
+      "loss": 1.602,
+      "step": 9555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2789604490074711e-05,
+      "loss": 2.1618,
+      "step": 9556
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2813241303997208e-05,
+      "loss": 2.2185,
+      "step": 9557
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.283689896521048e-05,
+      "loss": 2.2584,
+      "step": 9558
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2860577470119387e-05,
+      "loss": 2.1006,
+      "step": 9559
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.288427681512576e-05,
+      "loss": 2.1988,
+      "step": 9560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2907996996628078e-05,
+      "loss": 3.0038,
+      "step": 9561
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2931738011021858e-05,
+      "loss": 2.7186,
+      "step": 9562
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2955499854699253e-05,
+      "loss": 1.3901,
+      "step": 9563
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2979282524049383e-05,
+      "loss": 2.5526,
+      "step": 9564
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3003086015458191e-05,
+      "loss": 2.5288,
+      "step": 9565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3026910325308454e-05,
+      "loss": 2.2141,
+      "step": 9566
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3050755449979666e-05,
+      "loss": 2.3674,
+      "step": 9567
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3074621385848311e-05,
+      "loss": 2.3728,
+      "step": 9568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3098508129287638e-05,
+      "loss": 1.7927,
+      "step": 9569
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3122415676667836e-05,
+      "loss": 1.9749,
+      "step": 9570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3146344024355692e-05,
+      "loss": 2.6889,
+      "step": 9571
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3170293168715084e-05,
+      "loss": 2.0553,
+      "step": 9572
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3194263106106607e-05,
+      "loss": 2.4092,
+      "step": 9573
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3218253832887771e-05,
+      "loss": 2.5006,
+      "step": 9574
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3242265345412795e-05,
+      "loss": 2.5432,
+      "step": 9575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3266297640032847e-05,
+      "loss": 3.1217,
+      "step": 9576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3290350713095933e-05,
+      "loss": 2.5172,
+      "step": 9577
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3314424560946924e-05,
+      "loss": 1.6435,
+      "step": 9578
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3338519179927365e-05,
+      "loss": 1.2151,
+      "step": 9579
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3362634566375852e-05,
+      "loss": 2.49,
+      "step": 9580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3386770716627732e-05,
+      "loss": 2.5193,
+      "step": 9581
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3410927627015188e-05,
+      "loss": 1.2473,
+      "step": 9582
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3435105293867344e-05,
+      "loss": 2.1355,
+      "step": 9583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3459303713509984e-05,
+      "loss": 2.6995,
+      "step": 9584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3483522882265902e-05,
+      "loss": 2.3553,
+      "step": 9585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3507762796454657e-05,
+      "loss": 0.7652,
+      "step": 9586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3532023452392765e-05,
+      "loss": 2.799,
+      "step": 9587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.355630484639339e-05,
+      "loss": 2.4557,
+      "step": 9588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3580606974766717e-05,
+      "loss": 2.6492,
+      "step": 9589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3604929833819733e-05,
+      "loss": 2.2563,
+      "step": 9590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.362927341985628e-05,
+      "loss": 1.7613,
+      "step": 9591
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3653637729176997e-05,
+      "loss": 2.2952,
+      "step": 9592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3678022758079448e-05,
+      "loss": 2.6327,
+      "step": 9593
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3702428502857981e-05,
+      "loss": 2.873,
+      "step": 9594
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.372685495980388e-05,
+      "loss": 2.0651,
+      "step": 9595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3751302125205177e-05,
+      "loss": 2.2303,
+      "step": 9596
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.377576999534683e-05,
+      "loss": 2.6716,
+      "step": 9597
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.380025856651063e-05,
+      "loss": 2.2763,
+      "step": 9598
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3824767834975286e-05,
+      "loss": 2.7778,
+      "step": 9599
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3849297797016195e-05,
+      "loss": 2.2432,
+      "step": 9600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3873848448905755e-05,
+      "loss": 1.9893,
+      "step": 9601
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3898419786913184e-05,
+      "loss": 2.5541,
+      "step": 9602
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3923011807304619e-05,
+      "loss": 2.0628,
+      "step": 9603
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3947624506342864e-05,
+      "loss": 2.4413,
+      "step": 9604
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3972257880287763e-05,
+      "loss": 2.8503,
+      "step": 9605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3996911925395943e-05,
+      "loss": 2.4961,
+      "step": 9606
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4021586637920983e-05,
+      "loss": 2.5325,
+      "step": 9607
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4046282014113136e-05,
+      "loss": 2.4037,
+      "step": 9608
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.407099805021965e-05,
+      "loss": 2.2219,
+      "step": 9609
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4095734742484615e-05,
+      "loss": 2.5092,
+      "step": 9610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4120492087149054e-05,
+      "loss": 2.4627,
+      "step": 9611
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4145270080450626e-05,
+      "loss": 2.6263,
+      "step": 9612
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4170068718624061e-05,
+      "loss": 2.5195,
+      "step": 9613
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4194887997900906e-05,
+      "loss": 2.6546,
+      "step": 9614
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4219727914509577e-05,
+      "loss": 2.8993,
+      "step": 9615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4244588464675244e-05,
+      "loss": 1.4059,
+      "step": 9616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4269469644620064e-05,
+      "loss": 2.3323,
+      "step": 9617
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4294371450563041e-05,
+      "loss": 2.5299,
+      "step": 9618
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.431929387872004e-05,
+      "loss": 2.278,
+      "step": 9619
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4344236925303737e-05,
+      "loss": 0.8335,
+      "step": 9620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4369200586523749e-05,
+      "loss": 2.5733,
+      "step": 9621
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4394184858586493e-05,
+      "loss": 2.064,
+      "step": 9622
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4419189737695325e-05,
+      "loss": 2.3605,
+      "step": 9623
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4444215220050382e-05,
+      "loss": 2.1272,
+      "step": 9624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4469261301848756e-05,
+      "loss": 1.6777,
+      "step": 9625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4494327979284359e-05,
+      "loss": 2.2686,
+      "step": 9626
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4519415248548072e-05,
+      "loss": 2.1278,
+      "step": 9627
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4544523105827411e-05,
+      "loss": 2.7803,
+      "step": 9628
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4569651547306993e-05,
+      "loss": 2.3819,
+      "step": 9629
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4594800569168225e-05,
+      "loss": 2.3166,
+      "step": 9630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4619970167589462e-05,
+      "loss": 2.4387,
+      "step": 9631
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4645160338745715e-05,
+      "loss": 2.2806,
+      "step": 9632
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4670371078809096e-05,
+      "loss": 2.3147,
+      "step": 9633
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4695602383948501e-05,
+      "loss": 2.7972,
+      "step": 9634
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.472085425032975e-05,
+      "loss": 2.5147,
+      "step": 9635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.474612667411541e-05,
+      "loss": 2.2658,
+      "step": 9636
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4771419651465055e-05,
+      "loss": 2.0508,
+      "step": 9637
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4796733178535108e-05,
+      "loss": 2.4929,
+      "step": 9638
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4822067251478881e-05,
+      "loss": 1.9951,
+      "step": 9639
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4847421866446434e-05,
+      "loss": 2.4669,
+      "step": 9640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4872797019584887e-05,
+      "loss": 2.4881,
+      "step": 9641
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.489819270703812e-05,
+      "loss": 1.9754,
+      "step": 9642
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4923608924947023e-05,
+      "loss": 2.3762,
+      "step": 9643
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.494904566944915e-05,
+      "loss": 2.1504,
+      "step": 9644
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4974502936679095e-05,
+      "loss": 2.4881,
+      "step": 9645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.4999980722768334e-05,
+      "loss": 2.5481,
+      "step": 9646
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.50254790238452e-05,
+      "loss": 1.9118,
+      "step": 9647
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5050997836034855e-05,
+      "loss": 2.1742,
+      "step": 9648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5076537155459423e-05,
+      "loss": 2.113,
+      "step": 9649
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5102096978237837e-05,
+      "loss": 2.8548,
+      "step": 9650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5127677300485991e-05,
+      "loss": 1.9402,
+      "step": 9651
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5153278118316626e-05,
+      "loss": 1.9434,
+      "step": 9652
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.517889942783932e-05,
+      "loss": 2.1158,
+      "step": 9653
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5204541225160655e-05,
+      "loss": 2.4149,
+      "step": 9654
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5230203506383965e-05,
+      "loss": 2.4614,
+      "step": 9655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5255886267609601e-05,
+      "loss": 2.7795,
+      "step": 9656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5281589504934672e-05,
+      "loss": 2.4864,
+      "step": 9657
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5307313214453266e-05,
+      "loss": 1.7197,
+      "step": 9658
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5333057392256343e-05,
+      "loss": 1.9503,
+      "step": 9659
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5358822034431818e-05,
+      "loss": 2.4841,
+      "step": 9660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.538460713706427e-05,
+      "loss": 2.3932,
+      "step": 9661
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.54104126962354e-05,
+      "loss": 2.57,
+      "step": 9662
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5436238708023735e-05,
+      "loss": 2.1301,
+      "step": 9663
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5462085168504707e-05,
+      "loss": 2.2898,
+      "step": 9664
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5487952073750525e-05,
+      "loss": 2.307,
+      "step": 9665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5513839419830437e-05,
+      "loss": 2.4358,
+      "step": 9666
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5539747202810504e-05,
+      "loss": 2.1371,
+      "step": 9667
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.556567541875379e-05,
+      "loss": 1.2425,
+      "step": 9668
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5591624063720036e-05,
+      "loss": 2.3111,
+      "step": 9669
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5617593133766078e-05,
+      "loss": 2.5402,
+      "step": 9670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5643582624945582e-05,
+      "loss": 2.3966,
+      "step": 9671
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5669592533309167e-05,
+      "loss": 2.1462,
+      "step": 9672
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5695622854904173e-05,
+      "loss": 2.698,
+      "step": 9673
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.572167358577503e-05,
+      "loss": 2.6332,
+      "step": 9674
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5747744721962975e-05,
+      "loss": 2.4049,
+      "step": 9675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.577383625950624e-05,
+      "loss": 1.9449,
+      "step": 9676
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.579994819443975e-05,
+      "loss": 1.9993,
+      "step": 9677
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.582608052279554e-05,
+      "loss": 2.7163,
+      "step": 9678
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5852233240602447e-05,
+      "loss": 2.3104,
+      "step": 9679
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5878406343886256e-05,
+      "loss": 1.5407,
+      "step": 9680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.590459982866959e-05,
+      "loss": 2.5015,
+      "step": 9681
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5930813690972045e-05,
+      "loss": 2.2845,
+      "step": 9682
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5957047926810056e-05,
+      "loss": 2.4994,
+      "step": 9683
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.5983302532197034e-05,
+      "loss": 2.7641,
+      "step": 9684
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6009577503143202e-05,
+      "loss": 2.3668,
+      "step": 9685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6035872835655767e-05,
+      "loss": 2.6319,
+      "step": 9686
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.606218852573883e-05,
+      "loss": 2.4637,
+      "step": 9687
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.608852456939342e-05,
+      "loss": 2.3779,
+      "step": 9688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.611488096261732e-05,
+      "loss": 2.393,
+      "step": 9689
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.614125770140541e-05,
+      "loss": 2.7763,
+      "step": 9690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.61676547817494e-05,
+      "loss": 3.0011,
+      "step": 9691
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6194072199637993e-05,
+      "loss": 2.3839,
+      "step": 9692
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6220509951056558e-05,
+      "loss": 2.7256,
+      "step": 9693
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.624696803198765e-05,
+      "loss": 2.3636,
+      "step": 9694
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.627344643841058e-05,
+      "loss": 2.5657,
+      "step": 9695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.629994516630172e-05,
+      "loss": 2.4471,
+      "step": 9696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.632646421163407e-05,
+      "loss": 2.4321,
+      "step": 9697
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6353003570377844e-05,
+      "loss": 2.5868,
+      "step": 9698
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6379563238499992e-05,
+      "loss": 2.2491,
+      "step": 9699
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6406143211964516e-05,
+      "loss": 2.2831,
+      "step": 9700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6432743486732132e-05,
+      "loss": 2.3556,
+      "step": 9701
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6459364058760615e-05,
+      "loss": 1.102,
+      "step": 9702
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6486004924004676e-05,
+      "loss": 2.3491,
+      "step": 9703
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6512666078415914e-05,
+      "loss": 2.7589,
+      "step": 9704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6539347517942725e-05,
+      "loss": 2.422,
+      "step": 9705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6566049238530564e-05,
+      "loss": 2.0459,
+      "step": 9706
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6592771236121783e-05,
+      "loss": 2.2717,
+      "step": 9707
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6619513506655642e-05,
+      "loss": 2.6009,
+      "step": 9708
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6646276046068256e-05,
+      "loss": 2.5268,
+      "step": 9709
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6673058850292787e-05,
+      "loss": 3.0617,
+      "step": 9710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6699861915259167e-05,
+      "loss": 1.0959,
+      "step": 9711
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.672668523689439e-05,
+      "loss": 2.7747,
+      "step": 9712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.675352881112223e-05,
+      "loss": 2.5331,
+      "step": 9713
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.678039263386352e-05,
+      "loss": 1.6903,
+      "step": 9714
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6807276701035975e-05,
+      "loss": 2.4116,
+      "step": 9715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.68341810085542e-05,
+      "loss": 2.6517,
+      "step": 9716
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6861105552329674e-05,
+      "loss": 2.5061,
+      "step": 9717
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6888050328270924e-05,
+      "loss": 2.5725,
+      "step": 9718
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6915015332283345e-05,
+      "loss": 2.3436,
+      "step": 9719
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.6942000560269307e-05,
+      "loss": 1.7416,
+      "step": 9720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.696900600812795e-05,
+      "loss": 2.4408,
+      "step": 9721
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.69960316717555e-05,
+      "loss": 2.5984,
+      "step": 9722
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7023077547045062e-05,
+      "loss": 2.4369,
+      "step": 9723
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.705014362988667e-05,
+      "loss": 2.76,
+      "step": 9724
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7077229916167352e-05,
+      "loss": 2.3878,
+      "step": 9725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7104336401770848e-05,
+      "loss": 1.2033,
+      "step": 9726
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7131463082578085e-05,
+      "loss": 2.3808,
+      "step": 9727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7158609954466778e-05,
+      "loss": 2.5625,
+      "step": 9728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.718577701331168e-05,
+      "loss": 2.6343,
+      "step": 9729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7212964254984308e-05,
+      "loss": 2.4081,
+      "step": 9730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.724017167535325e-05,
+      "loss": 2.2904,
+      "step": 9731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7267399270283976e-05,
+      "loss": 2.8348,
+      "step": 9732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7294647035639003e-05,
+      "loss": 2.1331,
+      "step": 9733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.732191496727754e-05,
+      "loss": 1.5921,
+      "step": 9734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7349203061055915e-05,
+      "loss": 2.6128,
+      "step": 9735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7376511312827386e-05,
+      "loss": 2.102,
+      "step": 9736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7403839718442163e-05,
+      "loss": 2.2474,
+      "step": 9737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7431188273747226e-05,
+      "loss": 2.5367,
+      "step": 9738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7458556974586645e-05,
+      "loss": 1.7921,
+      "step": 9739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7485945816801435e-05,
+      "loss": 2.5309,
+      "step": 9740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7513354796229502e-05,
+      "loss": 2.1039,
+      "step": 9741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7540783908705663e-05,
+      "loss": 2.3645,
+      "step": 9742
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7568233150061755e-05,
+      "loss": 2.5193,
+      "step": 9743
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7595702516126483e-05,
+      "loss": 2.2716,
+      "step": 9744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7623192002725544e-05,
+      "loss": 2.4871,
+      "step": 9745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7650701605681524e-05,
+      "loss": 2.7156,
+      "step": 9746
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7678231320814008e-05,
+      "loss": 2.8142,
+      "step": 9747
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7705781143939497e-05,
+      "loss": 1.9107,
+      "step": 9748
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7733351070871522e-05,
+      "loss": 2.4202,
+      "step": 9749
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7760941097420316e-05,
+      "loss": 2.1665,
+      "step": 9750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7788551219393338e-05,
+      "loss": 2.2727,
+      "step": 9751
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7816181432594822e-05,
+      "loss": 2.2149,
+      "step": 9752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7843831732826097e-05,
+      "loss": 2.9203,
+      "step": 9753
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7871502115885205e-05,
+      "loss": 2.343,
+      "step": 9754
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7899192577567346e-05,
+      "loss": 2.462,
+      "step": 9755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7926903113664582e-05,
+      "loss": 2.5837,
+      "step": 9756
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7954633719966034e-05,
+      "loss": 2.1868,
+      "step": 9757
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.798238439225752e-05,
+      "loss": 2.8225,
+      "step": 9758
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8010155126322068e-05,
+      "loss": 2.5057,
+      "step": 9759
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8037945917939534e-05,
+      "loss": 1.7604,
+      "step": 9760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8065756762886813e-05,
+      "loss": 2.4385,
+      "step": 9761
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8093587656937568e-05,
+      "loss": 2.6578,
+      "step": 9762
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8121438595862602e-05,
+      "loss": 2.3088,
+      "step": 9763
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.81493095754296e-05,
+      "loss": 2.2528,
+      "step": 9764
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.817720059140328e-05,
+      "loss": 2.7481,
+      "step": 9765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8205111639545118e-05,
+      "loss": 2.3281,
+      "step": 9766
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8233042715613737e-05,
+      "loss": 2.3109,
+      "step": 9767
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.826099381536464e-05,
+      "loss": 2.6851,
+      "step": 9768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.828896493455033e-05,
+      "loss": 2.598,
+      "step": 9769
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8316956068920185e-05,
+      "loss": 2.3398,
+      "step": 9770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.834496721422065e-05,
+      "loss": 2.4069,
+      "step": 9771
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.837299836619502e-05,
+      "loss": 2.3762,
+      "step": 9772
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8401049520583627e-05,
+      "loss": 2.5473,
+      "step": 9773
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.842912067312372e-05,
+      "loss": 2.4733,
+      "step": 9774
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8457211819549527e-05,
+      "loss": 2.5821,
+      "step": 9775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8485322955592248e-05,
+      "loss": 2.4161,
+      "step": 9776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8513454076980095e-05,
+      "loss": 1.9744,
+      "step": 9777
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8541605179438057e-05,
+      "loss": 2.6232,
+      "step": 9778
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.856977625868829e-05,
+      "loss": 2.7716,
+      "step": 9779
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8597967310449817e-05,
+      "loss": 2.216,
+      "step": 9780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8626178330438718e-05,
+      "loss": 2.076,
+      "step": 9781
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8654409314367835e-05,
+      "loss": 2.4619,
+      "step": 9782
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8682660257947167e-05,
+      "loss": 2.4871,
+      "step": 9783
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8710931156883636e-05,
+      "loss": 2.3881,
+      "step": 9784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8739222006881172e-05,
+      "loss": 2.5325,
+      "step": 9785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8767532803640483e-05,
+      "loss": 2.3525,
+      "step": 9786
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.879586354285943e-05,
+      "loss": 1.9932,
+      "step": 9787
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8824214220232835e-05,
+      "loss": 1.9272,
+      "step": 9788
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8852584831452467e-05,
+      "loss": 2.626,
+      "step": 9789
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.888097537220695e-05,
+      "loss": 2.1281,
+      "step": 9790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8909385838182027e-05,
+      "loss": 1.2389,
+      "step": 9791
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8937816225060365e-05,
+      "loss": 2.1989,
+      "step": 9792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8966266528521683e-05,
+      "loss": 3.1884,
+      "step": 9793
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8994736744242427e-05,
+      "loss": 2.0832,
+      "step": 9794
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9023226867896276e-05,
+      "loss": 2.3194,
+      "step": 9795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9051736895153768e-05,
+      "loss": 2.2062,
+      "step": 9796
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9080266821682462e-05,
+      "loss": 2.6763,
+      "step": 9797
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9108816643146902e-05,
+      "loss": 2.5264,
+      "step": 9798
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.913738635520848e-05,
+      "loss": 2.2597,
+      "step": 9799
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9165975953525715e-05,
+      "loss": 1.199,
+      "step": 9800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9194585433754035e-05,
+      "loss": 1.7656,
+      "step": 9801
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9223214791545917e-05,
+      "loss": 2.5835,
+      "step": 9802
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.925186402255069e-05,
+      "loss": 2.7488,
+      "step": 9803
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9280533122414787e-05,
+      "loss": 2.5088,
+      "step": 9804
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9309222086781512e-05,
+      "loss": 2.4747,
+      "step": 9805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337930911291284e-05,
+      "loss": 2.4255,
+      "step": 9806
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936665959158134e-05,
+      "loss": 2.1829,
+      "step": 9807
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395408123286044e-05,
+      "loss": 2.3112,
+      "step": 9808
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424176502036686e-05,
+      "loss": 1.2947,
+      "step": 9809
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9452964723461584e-05,
+      "loss": 2.3372,
+      "step": 9810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94817727831859e-05,
+      "loss": 2.6616,
+      "step": 9811
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9510600676831904e-05,
+      "loss": 2.1168,
+      "step": 9812
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.953944840001887e-05,
+      "loss": 2.0154,
+      "step": 9813
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.956831594836305e-05,
+      "loss": 2.4353,
+      "step": 9814
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9597203317477538e-05,
+      "loss": 2.2156,
+      "step": 9815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9626110502972564e-05,
+      "loss": 2.6701,
+      "step": 9816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9655037500455355e-05,
+      "loss": 1.9673,
+      "step": 9817
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.96839843055301e-05,
+      "loss": 2.2211,
+      "step": 9818
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.971295091379786e-05,
+      "loss": 1.9315,
+      "step": 9819
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9741937320856842e-05,
+      "loss": 2.231,
+      "step": 9820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.977094352230219e-05,
+      "loss": 2.5101,
+      "step": 9821
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.979996951372611e-05,
+      "loss": 2.8144,
+      "step": 9822
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9829015290717598e-05,
+      "loss": 2.4347,
+      "step": 9823
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9858080848862817e-05,
+      "loss": 2.2297,
+      "step": 9824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887166183744918e-05,
+      "loss": 1.889,
+      "step": 9825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9916271290944052e-05,
+      "loss": 2.3825,
+      "step": 9826
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9945396166037194e-05,
+      "loss": 2.6878,
+      "step": 9827
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9974540804598527e-05,
+      "loss": 2.3267,
+      "step": 9828
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0003705202199135e-05,
+      "loss": 2.1891,
+      "step": 9829
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.003288935440714e-05,
+      "loss": 2.7845,
+      "step": 9830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0062093256787575e-05,
+      "loss": 1.8765,
+      "step": 9831
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.009131690490261e-05,
+      "loss": 2.4974,
+      "step": 9832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.012056029431126e-05,
+      "loss": 2.3356,
+      "step": 9833
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0149823420569684e-05,
+      "loss": 2.485,
+      "step": 9834
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0179106279230917e-05,
+      "loss": 3.0599,
+      "step": 9835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0208408865845054e-05,
+      "loss": 2.7034,
+      "step": 9836
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0237731175959225e-05,
+      "loss": 3.2046,
+      "step": 9837
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.026707320511758e-05,
+      "loss": 2.311,
+      "step": 9838
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.0296434948861094e-05,
+      "loss": 2.4845,
+      "step": 9839
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.032581640272793e-05,
+      "loss": 2.5421,
+      "step": 9840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.03552175622532e-05,
+      "loss": 1.5476,
+      "step": 9841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0384638422969086e-05,
+      "loss": 2.1334,
+      "step": 9842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.041407898040459e-05,
+      "loss": 2.54,
+      "step": 9843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0443539230085897e-05,
+      "loss": 2.2672,
+      "step": 9844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0473019167536138e-05,
+      "loss": 2.478,
+      "step": 9845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.050251878827554e-05,
+      "loss": 2.4274,
+      "step": 9846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.053203808782111e-05,
+      "loss": 1.5711,
+      "step": 9847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0561577061687072e-05,
+      "loss": 2.3001,
+      "step": 9848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0591135705384633e-05,
+      "loss": 2.1004,
+      "step": 9849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0620714014421995e-05,
+      "loss": 2.233,
+      "step": 9850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0650311984304263e-05,
+      "loss": 2.1121,
+      "step": 9851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0679929610533696e-05,
+      "loss": 2.9329,
+      "step": 9852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0709566888609498e-05,
+      "loss": 2.2704,
+      "step": 9853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0739223814028e-05,
+      "loss": 2.4602,
+      "step": 9854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0768900382282293e-05,
+      "loss": 1.4114,
+      "step": 9855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0798596588862736e-05,
+      "loss": 2.0084,
+      "step": 9856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0828312429256587e-05,
+      "loss": 2.3234,
+      "step": 9857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0858047898948177e-05,
+      "loss": 2.5262,
+      "step": 9858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0887802993418783e-05,
+      "loss": 2.431,
+      "step": 9859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.091757770814671e-05,
+      "loss": 2.4615,
+      "step": 9860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.094737203860734e-05,
+      "loss": 1.9599,
+      "step": 9861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0977185980273082e-05,
+      "loss": 2.1259,
+      "step": 9862
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1007019528613255e-05,
+      "loss": 2.1499,
+      "step": 9863
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1036872679094275e-05,
+      "loss": 2.289,
+      "step": 9864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.106674542717965e-05,
+      "loss": 2.4271,
+      "step": 9865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.109663776832976e-05,
+      "loss": 1.6883,
+      "step": 9866
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1126549698002077e-05,
+      "loss": 2.7043,
+      "step": 9867
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1156481211651115e-05,
+      "loss": 1.0059,
+      "step": 9868
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1186432304728385e-05,
+      "loss": 2.7207,
+      "step": 9869
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1216402972682456e-05,
+      "loss": 2.305,
+      "step": 9870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1246393210958974e-05,
+      "loss": 2.3868,
+      "step": 9871
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1276403015000364e-05,
+      "loss": 2.1879,
+      "step": 9872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1306432380246345e-05,
+      "loss": 2.2728,
+      "step": 9873
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.133648130213357e-05,
+      "loss": 2.1421,
+      "step": 9874
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.136654977609578e-05,
+      "loss": 2.2496,
+      "step": 9875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.139663779756356e-05,
+      "loss": 2.2967,
+      "step": 9876
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1426745361964692e-05,
+      "loss": 2.8803,
+      "step": 9877
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1456872464723974e-05,
+      "loss": 2.4231,
+      "step": 9878
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1487019101263248e-05,
+      "loss": 2.4985,
+      "step": 9879
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.151718526700123e-05,
+      "loss": 1.9371,
+      "step": 9880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1547370957353823e-05,
+      "loss": 2.4924,
+      "step": 9881
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1577576167733956e-05,
+      "loss": 2.4238,
+      "step": 9882
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1607800893551607e-05,
+      "loss": 2.2806,
+      "step": 9883
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1638045130213614e-05,
+      "loss": 2.3947,
+      "step": 9884
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.166830887312404e-05,
+      "loss": 2.4514,
+      "step": 9885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1698592117683926e-05,
+      "loss": 2.648,
+      "step": 9886
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1728894859291406e-05,
+      "loss": 2.5611,
+      "step": 9887
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1759217093341462e-05,
+      "loss": 2.8053,
+      "step": 9888
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.17895588152263e-05,
+      "loss": 1.9741,
+      "step": 9889
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1819920020335102e-05,
+      "loss": 2.5507,
+      "step": 9890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.185030070405413e-05,
+      "loss": 2.4501,
+      "step": 9891
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1880700861766576e-05,
+      "loss": 2.0534,
+      "step": 9892
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1911120488852833e-05,
+      "loss": 2.3477,
+      "step": 9893
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1941559580690154e-05,
+      "loss": 1.1424,
+      "step": 9894
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.197201813265302e-05,
+      "loss": 2.6755,
+      "step": 9895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.200249614011278e-05,
+      "loss": 2.2097,
+      "step": 9896
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.203299359843796e-05,
+      "loss": 2.6065,
+      "step": 9897
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2063510502994046e-05,
+      "loss": 2.2435,
+      "step": 9898
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2094046849143704e-05,
+      "loss": 2.6564,
+      "step": 9899
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2124602632246375e-05,
+      "loss": 2.5628,
+      "step": 9900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.215517784765882e-05,
+      "loss": 2.3523,
+      "step": 9901
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2185772490734715e-05,
+      "loss": 1.9862,
+      "step": 9902
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.221638655682489e-05,
+      "loss": 3.0368,
+      "step": 9903
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2247020041277014e-05,
+      "loss": 2.4048,
+      "step": 9904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.227767293943598e-05,
+      "loss": 2.4951,
+      "step": 9905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2308345246643717e-05,
+      "loss": 2.2544,
+      "step": 9906
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.233903695823923e-05,
+      "loss": 2.6441,
+      "step": 9907
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2369748069558357e-05,
+      "loss": 1.6286,
+      "step": 9908
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2400478575934262e-05,
+      "loss": 2.1497,
+      "step": 9909
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2431228472697026e-05,
+      "loss": 2.3946,
+      "step": 9910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.246199775517388e-05,
+      "loss": 2.5653,
+      "step": 9911
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2492786418688885e-05,
+      "loss": 2.3981,
+      "step": 9912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2523594458563393e-05,
+      "loss": 2.2024,
+      "step": 9913
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.255442187011573e-05,
+      "loss": 2.0502,
+      "step": 9914
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2585268648661325e-05,
+      "loss": 3.0359,
+      "step": 9915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2616134789512497e-05,
+      "loss": 2.6107,
+      "step": 9916
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2647020287978778e-05,
+      "loss": 2.2304,
+      "step": 9917
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.267792513936676e-05,
+      "loss": 2.4037,
+      "step": 9918
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.270884933898006e-05,
+      "loss": 2.5366,
+      "step": 9919
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.273979288211933e-05,
+      "loss": 1.7103,
+      "step": 9920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2770755764082254e-05,
+      "loss": 2.308,
+      "step": 9921
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2801737980163653e-05,
+      "loss": 2.4338,
+      "step": 9922
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2832739525655443e-05,
+      "loss": 2.0934,
+      "step": 9923
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2863760395846438e-05,
+      "loss": 2.4724,
+      "step": 9924
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.289480058602268e-05,
+      "loss": 2.0917,
+      "step": 9925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2925860091467236e-05,
+      "loss": 2.651,
+      "step": 9926
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.295693890746017e-05,
+      "loss": 2.146,
+      "step": 9927
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2988037029278645e-05,
+      "loss": 2.6718,
+      "step": 9928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.301915445219692e-05,
+      "loss": 2.929,
+      "step": 9929
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3050291171486313e-05,
+      "loss": 2.3322,
+      "step": 9930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3081447182415258e-05,
+      "loss": 2.3906,
+      "step": 9931
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.311262248024906e-05,
+      "loss": 2.3091,
+      "step": 9932
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3143817060250324e-05,
+      "loss": 2.5035,
+      "step": 9933
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3175030917678595e-05,
+      "loss": 2.7059,
+      "step": 9934
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.320626404779063e-05,
+      "loss": 2.1455,
+      "step": 9935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3237516445839993e-05,
+      "loss": 2.2691,
+      "step": 9936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.326878810707757e-05,
+      "loss": 2.2589,
+      "step": 9937
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3300079026751205e-05,
+      "loss": 2.3574,
+      "step": 9938
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3331389200105928e-05,
+      "loss": 2.4068,
+      "step": 9939
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.336271862238362e-05,
+      "loss": 2.4469,
+      "step": 9940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3394067288823425e-05,
+      "loss": 2.2316,
+      "step": 9941
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3425435194661524e-05,
+      "loss": 1.5806,
+      "step": 9942
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3456822335131143e-05,
+      "loss": 2.0399,
+      "step": 9943
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3488228705462696e-05,
+      "loss": 2.4606,
+      "step": 9944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.351965430088344e-05,
+      "loss": 2.2374,
+      "step": 9945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3551099116617907e-05,
+      "loss": 2.7482,
+      "step": 9946
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3582563147887652e-05,
+      "loss": 2.5468,
+      "step": 9947
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.36140463899114e-05,
+      "loss": 2.1041,
+      "step": 9948
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3645548837904714e-05,
+      "loss": 2.2331,
+      "step": 9949
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3677070487080457e-05,
+      "loss": 2.815,
+      "step": 9950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3708611332648526e-05,
+      "loss": 2.4836,
+      "step": 9951
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3740171369815904e-05,
+      "loss": 1.3273,
+      "step": 9952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3771750593786577e-05,
+      "loss": 1.8936,
+      "step": 9953
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3803348999761733e-05,
+      "loss": 2.3872,
+      "step": 9954
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.383496658293955e-05,
+      "loss": 2.5819,
+      "step": 9955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3866603338515373e-05,
+      "loss": 2.3619,
+      "step": 9956
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3898259261681507e-05,
+      "loss": 1.8616,
+      "step": 9957
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3929934347627504e-05,
+      "loss": 2.5402,
+      "step": 9958
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.39616285915399e-05,
+      "loss": 1.918,
+      "step": 9959
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.399334198860241e-05,
+      "loss": 2.5515,
+      "step": 9960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.402507453399566e-05,
+      "loss": 2.0367,
+      "step": 9961
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4056826222897534e-05,
+      "loss": 2.9086,
+      "step": 9962
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4088597050482967e-05,
+      "loss": 2.0407,
+      "step": 9963
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4120387011924047e-05,
+      "loss": 1.5417,
+      "step": 9964
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4152196102389727e-05,
+      "loss": 2.5382,
+      "step": 9965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.418402431704627e-05,
+      "loss": 2.2972,
+      "step": 9966
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4215871651056995e-05,
+      "loss": 1.7012,
+      "step": 9967
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4247738099582342e-05,
+      "loss": 2.1106,
+      "step": 9968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.427962365777965e-05,
+      "loss": 2.346,
+      "step": 9969
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.431152832080358e-05,
+      "loss": 2.9209,
+      "step": 9970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4343452083805826e-05,
+      "loss": 2.1006,
+      "step": 9971
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.43753949419352e-05,
+      "loss": 2.7852,
+      "step": 9972
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4407356890337455e-05,
+      "loss": 2.2148,
+      "step": 9973
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4439337924155605e-05,
+      "loss": 2.3403,
+      "step": 9974
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4471338038529758e-05,
+      "loss": 2.2709,
+      "step": 9975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4503357228597136e-05,
+      "loss": 2.9289,
+      "step": 9976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4535395489491865e-05,
+      "loss": 1.6443,
+      "step": 9977
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4567452816345408e-05,
+      "loss": 2.3877,
+      "step": 9978
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4599529204286224e-05,
+      "loss": 2.3134,
+      "step": 9979
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4631624648439946e-05,
+      "loss": 2.2405,
+      "step": 9980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4663739143929213e-05,
+      "loss": 2.7626,
+      "step": 9981
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4695872685873774e-05,
+      "loss": 2.1096,
+      "step": 9982
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.472802526939057e-05,
+      "loss": 2.0599,
+      "step": 9983
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.476019688959365e-05,
+      "loss": 2.3115,
+      "step": 9984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.479238754159401e-05,
+      "loss": 2.9897,
+      "step": 9985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4824597220499943e-05,
+      "loss": 2.0724,
+      "step": 9986
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4856825921416806e-05,
+      "loss": 2.423,
+      "step": 9987
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4889073639446973e-05,
+      "loss": 2.1243,
+      "step": 9988
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4921340369689974e-05,
+      "loss": 2.1272,
+      "step": 9989
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.49536261072425e-05,
+      "loss": 1.1467,
+      "step": 9990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4985930847198315e-05,
+      "loss": 1.7751,
+      "step": 9991
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5018254584648373e-05,
+      "loss": 2.4717,
+      "step": 9992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5050597314680504e-05,
+      "loss": 1.9558,
+      "step": 9993
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5082959032379906e-05,
+      "loss": 2.3802,
+      "step": 9994
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.51153397328288e-05,
+      "loss": 2.7221,
+      "step": 9995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.51477394111066e-05,
+      "loss": 2.4345,
+      "step": 9996
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5180158062289578e-05,
+      "loss": 2.9058,
+      "step": 9997
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.521259568145143e-05,
+      "loss": 2.3642,
+      "step": 9998
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5245052263662795e-05,
+      "loss": 2.0754,
+      "step": 9999
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.527752780399159e-05,
+      "loss": 1.2312,
+      "step": 10000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.531002229750257e-05,
+      "loss": 2.4883,
+      "step": 10001
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.534253573925784e-05,
+      "loss": 2.6852,
+      "step": 10002
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5375068124316606e-05,
+      "loss": 2.5541,
+      "step": 10003
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5407619447735184e-05,
+      "loss": 2.4803,
+      "step": 10004
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5440189704566865e-05,
+      "loss": 2.1513,
+      "step": 10005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5472778889862244e-05,
+      "loss": 2.1255,
+      "step": 10006
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5505386998668965e-05,
+      "loss": 2.0043,
+      "step": 10007
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5538014026031857e-05,
+      "loss": 2.1969,
+      "step": 10008
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5570659966992786e-05,
+      "loss": 2.4274,
+      "step": 10009
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5603324816590754e-05,
+      "loss": 2.6407,
+      "step": 10010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5636008569861932e-05,
+      "loss": 2.5093,
+      "step": 10011
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5668711221839644e-05,
+      "loss": 2.3886,
+      "step": 10012
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.570143276755425e-05,
+      "loss": 2.4018,
+      "step": 10013
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5734173202033287e-05,
+      "loss": 2.2821,
+      "step": 10014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5766932520301507e-05,
+      "loss": 2.2505,
+      "step": 10015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.579971071738063e-05,
+      "loss": 2.1559,
+      "step": 10016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.583250778828963e-05,
+      "loss": 2.2219,
+      "step": 10017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5865323728044518e-05,
+      "loss": 2.7482,
+      "step": 10018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.589815853165853e-05,
+      "loss": 2.3677,
+      "step": 10019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5931012194141994e-05,
+      "loss": 2.2097,
+      "step": 10020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.596388471050243e-05,
+      "loss": 2.1055,
+      "step": 10021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5996776075744325e-05,
+      "loss": 2.2377,
+      "step": 10022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6029686284869456e-05,
+      "loss": 2.787,
+      "step": 10023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6062615332876692e-05,
+      "loss": 2.0753,
+      "step": 10024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.609556321476215e-05,
+      "loss": 1.239,
+      "step": 10025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6128529925518788e-05,
+      "loss": 2.7255,
+      "step": 10026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6161515460136996e-05,
+      "loss": 2.3984,
+      "step": 10027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6194519813604165e-05,
+      "loss": 2.5093,
+      "step": 10028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6227542980904957e-05,
+      "loss": 1.944,
+      "step": 10029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.626058495702094e-05,
+      "loss": 2.5656,
+      "step": 10030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6293645736931012e-05,
+      "loss": 2.047,
+      "step": 10031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.632672531561116e-05,
+      "loss": 2.6186,
+      "step": 10032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6359823688034615e-05,
+      "loss": 1.7169,
+      "step": 10033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.639294084917149e-05,
+      "loss": 2.4501,
+      "step": 10034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6426076793989276e-05,
+      "loss": 2.4599,
+      "step": 10035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.645923151745255e-05,
+      "loss": 2.5215,
+      "step": 10036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.649240501452311e-05,
+      "loss": 1.8479,
+      "step": 10037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6525597280159638e-05,
+      "loss": 2.1691,
+      "step": 10038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6558808309318254e-05,
+      "loss": 2.996,
+      "step": 10039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.659203809695207e-05,
+      "loss": 2.1227,
+      "step": 10040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6625286638011473e-05,
+      "loss": 2.1588,
+      "step": 10041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6658553927443876e-05,
+      "loss": 2.8897,
+      "step": 10042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.669183996019383e-05,
+      "loss": 2.4097,
+      "step": 10043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.672514473120314e-05,
+      "loss": 2.3433,
+      "step": 10044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6758468235410745e-05,
+      "loss": 1.5247,
+      "step": 10045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.679181046775266e-05,
+      "loss": 2.3616,
+      "step": 10046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.682517142316214e-05,
+      "loss": 2.6251,
+      "step": 10047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.685855109656957e-05,
+      "loss": 2.005,
+      "step": 10048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6891949482902475e-05,
+      "loss": 2.492,
+      "step": 10049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6925366577085466e-05,
+      "loss": 2.5634,
+      "step": 10050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6958802374040455e-05,
+      "loss": 1.5044,
+      "step": 10051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6992256868686442e-05,
+      "loss": 2.404,
+      "step": 10052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7025730055939638e-05,
+      "loss": 2.0214,
+      "step": 10053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.705922193071323e-05,
+      "loss": 1.7108,
+      "step": 10054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7092732487917778e-05,
+      "loss": 2.4437,
+      "step": 10055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.71262617224609e-05,
+      "loss": 2.6955,
+      "step": 10056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7159809629247493e-05,
+      "loss": 2.3303,
+      "step": 10057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7193376203179362e-05,
+      "loss": 1.7564,
+      "step": 10058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.722696143915571e-05,
+      "loss": 2.865,
+      "step": 10059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.726056533207283e-05,
+      "loss": 2.5443,
+      "step": 10060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7294187876824263e-05,
+      "loss": 2.2545,
+      "step": 10061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.732782906830046e-05,
+      "loss": 2.3132,
+      "step": 10062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.736148890138931e-05,
+      "loss": 1.9674,
+      "step": 10063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7395167370975756e-05,
+      "loss": 2.156,
+      "step": 10064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.742886447194199e-05,
+      "loss": 2.7329,
+      "step": 10065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7462580199167162e-05,
+      "loss": 2.9117,
+      "step": 10066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7496314547527814e-05,
+      "loss": 2.1005,
+      "step": 10067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7530067511897567e-05,
+      "loss": 2.5494,
+      "step": 10068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7563839087147278e-05,
+      "loss": 2.4058,
+      "step": 10069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7597629268144862e-05,
+      "loss": 2.3412,
+      "step": 10070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7631438049755462e-05,
+      "loss": 2.8971,
+      "step": 10071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7665265426841388e-05,
+      "loss": 2.4386,
+      "step": 10072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.769911139426222e-05,
+      "loss": 2.4566,
+      "step": 10073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7732975946874535e-05,
+      "loss": 2.6771,
+      "step": 10074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.776685907953222e-05,
+      "loss": 2.031,
+      "step": 10075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7800760787086312e-05,
+      "loss": 2.4287,
+      "step": 10076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.783468106438501e-05,
+      "loss": 1.6896,
+      "step": 10077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7868619906273644e-05,
+      "loss": 1.2589,
+      "step": 10078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.790257730759478e-05,
+      "loss": 2.6282,
+      "step": 10079
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7936553263188188e-05,
+      "loss": 3.1504,
+      "step": 10080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7970547767890862e-05,
+      "loss": 2.4569,
+      "step": 10081
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8004560816536715e-05,
+      "loss": 2.4792,
+      "step": 10082
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8038592403957132e-05,
+      "loss": 2.705,
+      "step": 10083
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.807264252498057e-05,
+      "loss": 2.1951,
+      "step": 10084
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.810671117443275e-05,
+      "loss": 2.5855,
+      "step": 10085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8140798347136346e-05,
+      "loss": 2.6789,
+      "step": 10086
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.817490403791145e-05,
+      "loss": 2.249,
+      "step": 10087
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8209028241575278e-05,
+      "loss": 2.5145,
+      "step": 10088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8243170952942205e-05,
+      "loss": 1.9606,
+      "step": 10089
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.827733216682389e-05,
+      "loss": 2.0652,
+      "step": 10090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8311511878028947e-05,
+      "loss": 2.2876,
+      "step": 10091
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8345710081363382e-05,
+      "loss": 2.4698,
+      "step": 10092
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8379926771630386e-05,
+      "loss": 2.0755,
+      "step": 10093
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8414161943630333e-05,
+      "loss": 2.1038,
+      "step": 10094
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8448415592160582e-05,
+      "loss": 2.3284,
+      "step": 10095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8482687712015972e-05,
+      "loss": 2.4869,
+      "step": 10096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8516978297988398e-05,
+      "loss": 1.4216,
+      "step": 10097
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8551287344867025e-05,
+      "loss": 2.1233,
+      "step": 10098
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.858561484743801e-05,
+      "loss": 2.2981,
+      "step": 10099
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8619960800484927e-05,
+      "loss": 2.4564,
+      "step": 10100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8654325198788464e-05,
+      "loss": 2.8618,
+      "step": 10101
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8688708037126552e-05,
+      "loss": 2.1518,
+      "step": 10102
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8723109310274234e-05,
+      "loss": 2.3855,
+      "step": 10103
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.875752901300376e-05,
+      "loss": 2.6243,
+      "step": 10104
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8791967140084648e-05,
+      "loss": 2.2973,
+      "step": 10105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8826423686283616e-05,
+      "loss": 2.5562,
+      "step": 10106
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8860898646364474e-05,
+      "loss": 2.0648,
+      "step": 10107
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.889539201508837e-05,
+      "loss": 1.8432,
+      "step": 10108
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8929903787213587e-05,
+      "loss": 2.2891,
+      "step": 10109
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.896443395749562e-05,
+      "loss": 2.6367,
+      "step": 10110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8998982520687113e-05,
+      "loss": 2.3926,
+      "step": 10111
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9033549471538003e-05,
+      "loss": 2.2143,
+      "step": 10112
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9068134804795405e-05,
+      "loss": 2.8538,
+      "step": 10113
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.910273851520371e-05,
+      "loss": 2.3421,
+      "step": 10114
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9137360597504272e-05,
+      "loss": 1.3249,
+      "step": 10115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9172001046435915e-05,
+      "loss": 2.5313,
+      "step": 10116
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9206659856734555e-05,
+      "loss": 2.5991,
+      "step": 10117
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9241337023133448e-05,
+      "loss": 2.4937,
+      "step": 10118
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.927603254036278e-05,
+      "loss": 2.3276,
+      "step": 10119
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9310746403150203e-05,
+      "loss": 2.4499,
+      "step": 10120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9345478606220486e-05,
+      "loss": 2.5561,
+      "step": 10121
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.938022914429573e-05,
+      "loss": 2.3144,
+      "step": 10122
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9414998012094963e-05,
+      "loss": 2.3398,
+      "step": 10123
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9449785204334697e-05,
+      "loss": 2.3455,
+      "step": 10124
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.948459071572859e-05,
+      "loss": 1.8129,
+      "step": 10125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9519414540987542e-05,
+      "loss": 2.0522,
+      "step": 10126
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9554256674819466e-05,
+      "loss": 2.0031,
+      "step": 10127
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9589117111929767e-05,
+      "loss": 2.5275,
+      "step": 10128
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9623995847020932e-05,
+      "loss": 2.4785,
+      "step": 10129
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9658892874792734e-05,
+      "loss": 1.9679,
+      "step": 10130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9693808189942065e-05,
+      "loss": 2.5998,
+      "step": 10131
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9728741787163084e-05,
+      "loss": 2.507,
+      "step": 10132
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.976369366114719e-05,
+      "loss": 1.7297,
+      "step": 10133
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9798663806583067e-05,
+      "loss": 2.0951,
+      "step": 10134
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9833652218156447e-05,
+      "loss": 2.2432,
+      "step": 10135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9868658890550447e-05,
+      "loss": 2.1314,
+      "step": 10136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.990368381844539e-05,
+      "loss": 2.6786,
+      "step": 10137
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9938726996518738e-05,
+      "loss": 2.5748,
+      "step": 10138
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9973788419445194e-05,
+      "loss": 2.9981,
+      "step": 10139
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0008868081896775e-05,
+      "loss": 2.6332,
+      "step": 10140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.004396597854268e-05,
+      "loss": 2.5011,
+      "step": 10141
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0079082104049384e-05,
+      "loss": 1.6089,
+      "step": 10142
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0114216453080415e-05,
+      "loss": 2.1408,
+      "step": 10143
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0149369020296703e-05,
+      "loss": 2.1869,
+      "step": 10144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0184539800356396e-05,
+      "loss": 2.3833,
+      "step": 10145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0219728787914894e-05,
+      "loss": 2.4969,
+      "step": 10146
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0254935977624628e-05,
+      "loss": 2.6395,
+      "step": 10147
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0290161364135514e-05,
+      "loss": 3.1765,
+      "step": 10148
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.032540494209456e-05,
+      "loss": 2.4238,
+      "step": 10149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0360666706146177e-05,
+      "loss": 2.3671,
+      "step": 10150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.039594665093169e-05,
+      "loss": 2.4875,
+      "step": 10151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0431244771089963e-05,
+      "loss": 2.3563,
+      "step": 10152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0466561061256983e-05,
+      "loss": 2.1643,
+      "step": 10153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.050189551606606e-05,
+      "loss": 2.3604,
+      "step": 10154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.053724813014752e-05,
+      "loss": 1.6078,
+      "step": 10155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.057261889812917e-05,
+      "loss": 2.4993,
+      "step": 10156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.060800781463593e-05,
+      "loss": 1.8869,
+      "step": 10157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.064341487429008e-05,
+      "loss": 2.2471,
+      "step": 10158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.067884007171098e-05,
+      "loss": 2.5205,
+      "step": 10159
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.071428340151531e-05,
+      "loss": 2.2321,
+      "step": 10160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.074974485831704e-05,
+      "loss": 1.9376,
+      "step": 10161
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.078522443672734e-05,
+      "loss": 2.3114,
+      "step": 10162
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0820722131354645e-05,
+      "loss": 2.0161,
+      "step": 10163
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0856237936804595e-05,
+      "loss": 1.7932,
+      "step": 10164
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.0891771847680137e-05,
+      "loss": 2.959,
+      "step": 10165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.092732385858139e-05,
+      "loss": 2.2579,
+      "step": 10166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.096289396410584e-05,
+      "loss": 2.565,
+      "step": 10167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.099848215884808e-05,
+      "loss": 2.3212,
+      "step": 10168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.103408843740004e-05,
+      "loss": 2.3923,
+      "step": 10169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.106971279435096e-05,
+      "loss": 2.4092,
+      "step": 10170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.11053552242872e-05,
+      "loss": 2.3874,
+      "step": 10171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.1141015721792404e-05,
+      "loss": 1.1819,
+      "step": 10172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.117669428144753e-05,
+      "loss": 2.6151,
+      "step": 10173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.121239089783079e-05,
+      "loss": 2.7615,
+      "step": 10174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.124810556551768e-05,
+      "loss": 2.3171,
+      "step": 10175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.128383827908074e-05,
+      "loss": 2.2242,
+      "step": 10176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.131958903309001e-05,
+      "loss": 2.0179,
+      "step": 10177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.13553578221127e-05,
+      "loss": 2.3232,
+      "step": 10178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.1391144640713365e-05,
+      "loss": 2.5164,
+      "step": 10179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.1426949483453585e-05,
+      "loss": 2.58,
+      "step": 10180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.146277234489242e-05,
+      "loss": 2.7056,
+      "step": 10181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.149861321958612e-05,
+      "loss": 2.2965,
+      "step": 10182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.153447210208832e-05,
+      "loss": 2.6304,
+      "step": 10183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.1570348986949606e-05,
+      "loss": 2.529,
+      "step": 10184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.160624386871812e-05,
+      "loss": 2.1355,
+      "step": 10185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.1642156741939184e-05,
+      "loss": 2.2487,
+      "step": 10186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.167808760115542e-05,
+      "loss": 2.2559,
+      "step": 10187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.1714036440906555e-05,
+      "loss": 2.6564,
+      "step": 10188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.175000325572973e-05,
+      "loss": 2.232,
+      "step": 10189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.178598804015938e-05,
+      "loss": 1.8429,
+      "step": 10190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.182199078872718e-05,
+      "loss": 2.3662,
+      "step": 10191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.185801149596197e-05,
+      "loss": 2.4205,
+      "step": 10192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.189405015638995e-05,
+      "loss": 2.3862,
+      "step": 10193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.193010676453459e-05,
+      "loss": 2.4298,
+      "step": 10194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.196618131491669e-05,
+      "loss": 1.9362,
+      "step": 10195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.200227380205416e-05,
+      "loss": 2.8353,
+      "step": 10196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.203838422046233e-05,
+      "loss": 1.8835,
+      "step": 10197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.207451256465381e-05,
+      "loss": 2.8429,
+      "step": 10198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.211065882913837e-05,
+      "loss": 2.3089,
+      "step": 10199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.214682300842309e-05,
+      "loss": 2.5278,
+      "step": 10200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.21830050970124e-05,
+      "loss": 2.6866,
+      "step": 10201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.2219205089407965e-05,
+      "loss": 2.4306,
+      "step": 10202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.225542298010881e-05,
+      "loss": 2.4871,
+      "step": 10203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.229165876361102e-05,
+      "loss": 2.3038,
+      "step": 10204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.232791243440816e-05,
+      "loss": 2.3289,
+      "step": 10205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.236418398699102e-05,
+      "loss": 2.3425,
+      "step": 10206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.240047341584775e-05,
+      "loss": 2.7456,
+      "step": 10207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.243678071546357e-05,
+      "loss": 2.3086,
+      "step": 10208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.2473105880321165e-05,
+      "loss": 2.2698,
+      "step": 10209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.250944890490048e-05,
+      "loss": 2.063,
+      "step": 10210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.254580978367881e-05,
+      "loss": 2.2058,
+      "step": 10211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.2582188511130473e-05,
+      "loss": 2.1523,
+      "step": 10212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.261858508172734e-05,
+      "loss": 2.7197,
+      "step": 10213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.265499948993849e-05,
+      "loss": 2.0046,
+      "step": 10214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.2691431730230365e-05,
+      "loss": 2.5469,
+      "step": 10215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.272788179706647e-05,
+      "loss": 2.6124,
+      "step": 10216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.276434968490779e-05,
+      "loss": 2.4402,
+      "step": 10217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.28008353882126e-05,
+      "loss": 2.7979,
+      "step": 10218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.283733890143645e-05,
+      "loss": 2.3549,
+      "step": 10219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.287386021903215e-05,
+      "loss": 1.8398,
+      "step": 10220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.2910399335449726e-05,
+      "loss": 2.5321,
+      "step": 10221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.294695624513668e-05,
+      "loss": 2.6637,
+      "step": 10222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.298353094253773e-05,
+      "loss": 2.487,
+      "step": 10223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.30201234220948e-05,
+      "loss": 2.2754,
+      "step": 10224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.305673367824725e-05,
+      "loss": 2.4839,
+      "step": 10225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.309336170543171e-05,
+      "loss": 2.4562,
+      "step": 10226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.313000749808203e-05,
+      "loss": 2.4198,
+      "step": 10227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.31666710506294e-05,
+      "loss": 2.3516,
+      "step": 10228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.3203352357502334e-05,
+      "loss": 2.6311,
+      "step": 10229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.324005141312666e-05,
+      "loss": 2.5172,
+      "step": 10230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.327676821192554e-05,
+      "loss": 2.5574,
+      "step": 10231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.331350274831925e-05,
+      "loss": 2.4656,
+      "step": 10232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.3350255016725555e-05,
+      "loss": 2.1085,
+      "step": 10233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.338702501155952e-05,
+      "loss": 2.3179,
+      "step": 10234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.342381272723343e-05,
+      "loss": 2.1083,
+      "step": 10235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.3460618158157e-05,
+      "loss": 1.9976,
+      "step": 10236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.3497441298737036e-05,
+      "loss": 2.9078,
+      "step": 10237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.353428214337787e-05,
+      "loss": 1.8098,
+      "step": 10238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.3571140686481034e-05,
+      "loss": 2.1829,
+      "step": 10239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.360801692244551e-05,
+      "loss": 2.2699,
+      "step": 10240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.364491084566729e-05,
+      "loss": 2.6596,
+      "step": 10241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.368182245053998e-05,
+      "loss": 2.5269,
+      "step": 10242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.371875173145435e-05,
+      "loss": 2.1662,
+      "step": 10243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.375569868279864e-05,
+      "loss": 1.8435,
+      "step": 10244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.379266329895811e-05,
+      "loss": 2.0144,
+      "step": 10245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.382964557431558e-05,
+      "loss": 1.9259,
+      "step": 10246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.386664550325112e-05,
+      "loss": 2.6408,
+      "step": 10247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.390366308014222e-05,
+      "loss": 2.372,
+      "step": 10248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.3940698299363405e-05,
+      "loss": 2.3985,
+      "step": 10249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.397775115528677e-05,
+      "loss": 2.1965,
+      "step": 10250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.401482164228167e-05,
+      "loss": 1.867,
+      "step": 10251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.405190975471481e-05,
+      "loss": 2.3877,
+      "step": 10252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.408901548695012e-05,
+      "loss": 2.3115,
+      "step": 10253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4126138833348895e-05,
+      "loss": 2.1906,
+      "step": 10254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4163279788269774e-05,
+      "loss": 2.6528,
+      "step": 10255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.420043834606877e-05,
+      "loss": 2.5181,
+      "step": 10256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.423761450109906e-05,
+      "loss": 2.4925,
+      "step": 10257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4274808247711324e-05,
+      "loss": 2.4936,
+      "step": 10258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4312019580253515e-05,
+      "loss": 2.2492,
+      "step": 10259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.434924849307086e-05,
+      "loss": 2.0153,
+      "step": 10260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4386494980505896e-05,
+      "loss": 2.478,
+      "step": 10261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4423759036898593e-05,
+      "loss": 2.0733,
+      "step": 10262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4461040656586185e-05,
+      "loss": 2.036,
+      "step": 10263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.449833983390335e-05,
+      "loss": 2.3027,
+      "step": 10264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.453565656318183e-05,
+      "loss": 1.8588,
+      "step": 10265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.457299083875093e-05,
+      "loss": 1.5095,
+      "step": 10266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.461034265493726e-05,
+      "loss": 2.0365,
+      "step": 10267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.464771200606478e-05,
+      "loss": 2.2866,
+      "step": 10268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.46850988864546e-05,
+      "loss": 2.2624,
+      "step": 10269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.472250329042537e-05,
+      "loss": 2.2207,
+      "step": 10270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.475992521229302e-05,
+      "loss": 2.272,
+      "step": 10271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.479736464637089e-05,
+      "loss": 2.3282,
+      "step": 10272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.483482158696939e-05,
+      "loss": 2.145,
+      "step": 10273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.487229602839658e-05,
+      "loss": 2.254,
+      "step": 10274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.4909787964957694e-05,
+      "loss": 2.4914,
+      "step": 10275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.494729739095547e-05,
+      "loss": 1.7833,
+      "step": 10276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.498482430068967e-05,
+      "loss": 2.5529,
+      "step": 10277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.502236868845771e-05,
+      "loss": 2.4559,
+      "step": 10278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.505993054855422e-05,
+      "loss": 1.8729,
+      "step": 10279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.509750987527125e-05,
+      "loss": 2.328,
+      "step": 10280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.513510666289807e-05,
+      "loss": 1.5714,
+      "step": 10281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.517272090572134e-05,
+      "loss": 1.9527,
+      "step": 10282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.521035259802514e-05,
+      "loss": 2.3069,
+      "step": 10283
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.524800173409088e-05,
+      "loss": 2.2626,
+      "step": 10284
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5285668308197226e-05,
+      "loss": 1.077,
+      "step": 10285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.532335231462027e-05,
+      "loss": 2.3437,
+      "step": 10286
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5361053747633516e-05,
+      "loss": 2.2824,
+      "step": 10287
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.539877260150767e-05,
+      "loss": 2.4133,
+      "step": 10288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.543650887051086e-05,
+      "loss": 2.3259,
+      "step": 10289
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.547426254890861e-05,
+      "loss": 2.1761,
+      "step": 10290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5512033630963745e-05,
+      "loss": 2.4098,
+      "step": 10291
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5549822110936566e-05,
+      "loss": 2.6079,
+      "step": 10292
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5587627983084435e-05,
+      "loss": 2.3634,
+      "step": 10293
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.562545124166239e-05,
+      "loss": 2.536,
+      "step": 10294
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5663291880922664e-05,
+      "loss": 2.38,
+      "step": 10295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5701149895114986e-05,
+      "loss": 2.2454,
+      "step": 10296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.573902527848617e-05,
+      "loss": 2.5317,
+      "step": 10297
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5776918025280656e-05,
+      "loss": 2.2658,
+      "step": 10298
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.581482812974014e-05,
+      "loss": 2.4188,
+      "step": 10299
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.585275558610379e-05,
+      "loss": 2.6238,
+      "step": 10300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.5890700388607886e-05,
+      "loss": 2.481,
+      "step": 10301
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.592866253148628e-05,
+      "loss": 2.0003,
+      "step": 10302
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.596664200897016e-05,
+      "loss": 2.4055,
+      "step": 10303
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.600463881528814e-05,
+      "loss": 2.2915,
+      "step": 10304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.604265294466594e-05,
+      "loss": 1.3841,
+      "step": 10305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6080684391326916e-05,
+      "loss": 2.0819,
+      "step": 10306
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.611873314949169e-05,
+      "loss": 2.5909,
+      "step": 10307
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.615679921337826e-05,
+      "loss": 2.5536,
+      "step": 10308
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6194882577202124e-05,
+      "loss": 1.9539,
+      "step": 10309
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.623298323517581e-05,
+      "loss": 2.3643,
+      "step": 10310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6271101181509527e-05,
+      "loss": 2.0518,
+      "step": 10311
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6309236410410787e-05,
+      "loss": 2.5309,
+      "step": 10312
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.634738891608447e-05,
+      "loss": 2.3528,
+      "step": 10313
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.638555869273278e-05,
+      "loss": 2.3124,
+      "step": 10314
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.642374573455528e-05,
+      "loss": 3.0022,
+      "step": 10315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.646195003574902e-05,
+      "loss": 2.5131,
+      "step": 10316
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6500171590508376e-05,
+      "loss": 2.6058,
+      "step": 10317
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.653841039302504e-05,
+      "loss": 2.4174,
+      "step": 10318
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.657666643748815e-05,
+      "loss": 2.0871,
+      "step": 10319
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.661493971808427e-05,
+      "loss": 2.099,
+      "step": 10320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6653230228997226e-05,
+      "loss": 2.2718,
+      "step": 10321
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.669153796440826e-05,
+      "loss": 2.2751,
+      "step": 10322
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.672986291849603e-05,
+      "loss": 2.3995,
+      "step": 10323
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.67682050854366e-05,
+      "loss": 2.6568,
+      "step": 10324
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6806564459403466e-05,
+      "loss": 1.969,
+      "step": 10325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6844941034567234e-05,
+      "loss": 2.157,
+      "step": 10326
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.688333480509618e-05,
+      "loss": 2.4618,
+      "step": 10327
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6921745765155916e-05,
+      "loss": 2.9969,
+      "step": 10328
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.696017390890945e-05,
+      "loss": 2.2112,
+      "step": 10329
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.6998619230516954e-05,
+      "loss": 2.1581,
+      "step": 10330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.703708172413627e-05,
+      "loss": 2.492,
+      "step": 10331
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.707556138392254e-05,
+      "loss": 2.0593,
+      "step": 10332
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7114058204028366e-05,
+      "loss": 2.4774,
+      "step": 10333
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.715257217860346e-05,
+      "loss": 2.3574,
+      "step": 10334
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7191103301795256e-05,
+      "loss": 2.2376,
+      "step": 10335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7229651567748424e-05,
+      "loss": 1.8665,
+      "step": 10336
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.726821697060516e-05,
+      "loss": 2.0472,
+      "step": 10337
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7306799504504774e-05,
+      "loss": 2.2191,
+      "step": 10338
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.734539916358427e-05,
+      "loss": 2.2884,
+      "step": 10339
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7384015941977874e-05,
+      "loss": 2.8817,
+      "step": 10340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7422649833817374e-05,
+      "loss": 2.6723,
+      "step": 10341
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7461300833231785e-05,
+      "loss": 2.1272,
+      "step": 10342
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7499968934347546e-05,
+      "loss": 2.2292,
+      "step": 10343
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.75386541312886e-05,
+      "loss": 2.2031,
+      "step": 10344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.757735641817624e-05,
+      "loss": 2.9112,
+      "step": 10345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7616075789129117e-05,
+      "loss": 2.3423,
+      "step": 10346
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.765481223826334e-05,
+      "loss": 2.9609,
+      "step": 10347
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.769356575969247e-05,
+      "loss": 2.0434,
+      "step": 10348
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.773233634752735e-05,
+      "loss": 1.7449,
+      "step": 10349
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.777112399587624e-05,
+      "loss": 2.5364,
+      "step": 10350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7809928698844935e-05,
+      "loss": 2.3342,
+      "step": 10351
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7848750450536526e-05,
+      "loss": 2.5477,
+      "step": 10352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.788758924505167e-05,
+      "loss": 2.2875,
+      "step": 10353
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.7926445076488105e-05,
+      "loss": 2.1036,
+      "step": 10354
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.79653179389413e-05,
+      "loss": 2.7319,
+      "step": 10355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.800420782650402e-05,
+      "loss": 1.7301,
+      "step": 10356
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.804311473326653e-05,
+      "loss": 2.3279,
+      "step": 10357
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8082038653316244e-05,
+      "loss": 2.8236,
+      "step": 10358
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.812097958073827e-05,
+      "loss": 2.3627,
+      "step": 10359
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.815993750961504e-05,
+      "loss": 2.5501,
+      "step": 10360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8198912434026456e-05,
+      "loss": 2.2473,
+      "step": 10361
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.823790434804964e-05,
+      "loss": 2.357,
+      "step": 10362
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8276913245759334e-05,
+      "loss": 2.4992,
+      "step": 10363
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.831593912122762e-05,
+      "loss": 2.5565,
+      "step": 10364
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.835498196852414e-05,
+      "loss": 2.53,
+      "step": 10365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.839404178171563e-05,
+      "loss": 2.2341,
+      "step": 10366
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8433118554866536e-05,
+      "loss": 2.7347,
+      "step": 10367
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.847221228203867e-05,
+      "loss": 2.0758,
+      "step": 10368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8511322957291286e-05,
+      "loss": 1.4971,
+      "step": 10369
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8550450574680866e-05,
+      "loss": 2.3717,
+      "step": 10370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8589595128261524e-05,
+      "loss": 2.6679,
+      "step": 10371
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8628756612084776e-05,
+      "loss": 2.5808,
+      "step": 10372
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8667935020199566e-05,
+      "loss": 2.6389,
+      "step": 10373
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8707130346652146e-05,
+      "loss": 2.6418,
+      "step": 10374
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8746342585486344e-05,
+      "loss": 2.5869,
+      "step": 10375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8785571730743303e-05,
+      "loss": 2.4205,
+      "step": 10376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8824817776461673e-05,
+      "loss": 2.7586,
+      "step": 10377
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.8864080716677583e-05,
+      "loss": 2.3413,
+      "step": 10378
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.89033605454244e-05,
+      "loss": 2.4539,
+      "step": 10379
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.894265725673311e-05,
+      "loss": 2.3555,
+      "step": 10380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.898197084463211e-05,
+      "loss": 2.9184,
+      "step": 10381
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.902130130314717e-05,
+      "loss": 2.5812,
+      "step": 10382
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.906064862630146e-05,
+      "loss": 2.4188,
+      "step": 10383
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.910001280811571e-05,
+      "loss": 2.4244,
+      "step": 10384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.9139393842608e-05,
+      "loss": 2.0942,
+      "step": 10385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.917879172379399e-05,
+      "loss": 2.3745,
+      "step": 10386
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.921820644568649e-05,
+      "loss": 2.4183,
+      "step": 10387
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.925763800229598e-05,
+      "loss": 2.5245,
+      "step": 10388
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.929708638763036e-05,
+      "loss": 1.2377,
+      "step": 10389
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.9336551595695034e-05,
+      "loss": 2.3013,
+      "step": 10390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.9376033620492556e-05,
+      "loss": 2.5654,
+      "step": 10391
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.941553245602322e-05,
+      "loss": 2.4871,
+      "step": 10392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.945504809628469e-05,
+      "loss": 1.7322,
+      "step": 10393
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.949458053527211e-05,
+      "loss": 2.3045,
+      "step": 10394
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.953412976697786e-05,
+      "loss": 2.3667,
+      "step": 10395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.957369578539204e-05,
+      "loss": 2.1289,
+      "step": 10396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.961327858450203e-05,
+      "loss": 2.6629,
+      "step": 10397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.9652878158292854e-05,
+      "loss": 2.5798,
+      "step": 10398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.969249450074665e-05,
+      "loss": 2.0056,
+      "step": 10399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.973212760584329e-05,
+      "loss": 2.0417,
+      "step": 10400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.9771777467560033e-05,
+      "loss": 1.218,
+      "step": 10401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.981144407987159e-05,
+      "loss": 2.3103,
+      "step": 10402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.985112743675008e-05,
+      "loss": 2.093,
+      "step": 10403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.989082753216506e-05,
+      "loss": 2.348,
+      "step": 10404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.993054436008364e-05,
+      "loss": 2.5058,
+      "step": 10405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.997027791447039e-05,
+      "loss": 2.3734,
+      "step": 10406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.001002818928717e-05,
+      "loss": 2.5571,
+      "step": 10407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0049795178493475e-05,
+      "loss": 2.6754,
+      "step": 10408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0089578876046255e-05,
+      "loss": 2.5235,
+      "step": 10409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0129379275899805e-05,
+      "loss": 2.3528,
+      "step": 10410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0169196372005904e-05,
+      "loss": 2.6164,
+      "step": 10411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.020903015831389e-05,
+      "loss": 2.5935,
+      "step": 10412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0248880628770486e-05,
+      "loss": 2.5635,
+      "step": 10413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.028874777731999e-05,
+      "loss": 2.7472,
+      "step": 10414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.032863159790392e-05,
+      "loss": 2.2613,
+      "step": 10415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.036853208446149e-05,
+      "loss": 2.4866,
+      "step": 10416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.04084492309293e-05,
+      "loss": 2.4164,
+      "step": 10417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.044838303124155e-05,
+      "loss": 1.9011,
+      "step": 10418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.048833347932957e-05,
+      "loss": 2.4087,
+      "step": 10419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.052830056912248e-05,
+      "loss": 2.0451,
+      "step": 10420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.056828429454677e-05,
+      "loss": 2.2033,
+      "step": 10421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.060828464952649e-05,
+      "loss": 1.8205,
+      "step": 10422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.064830162798288e-05,
+      "loss": 2.3167,
+      "step": 10423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.068833522383496e-05,
+      "loss": 2.4006,
+      "step": 10424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0728385430999075e-05,
+      "loss": 2.6981,
+      "step": 10425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.076845224338921e-05,
+      "loss": 2.0741,
+      "step": 10426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.080853565491649e-05,
+      "loss": 2.4384,
+      "step": 10427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.084863565948983e-05,
+      "loss": 2.7827,
+      "step": 10428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.0888752251015506e-05,
+      "loss": 2.1802,
+      "step": 10429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.092888542339739e-05,
+      "loss": 2.7344,
+      "step": 10430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.096903517053654e-05,
+      "loss": 2.2409,
+      "step": 10431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.100920148633179e-05,
+      "loss": 2.7869,
+      "step": 10432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.104938436467935e-05,
+      "loss": 2.2943,
+      "step": 10433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.108958379947294e-05,
+      "loss": 2.3869,
+      "step": 10434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.112979978460367e-05,
+      "loss": 1.992,
+      "step": 10435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1170032313960285e-05,
+      "loss": 2.0761,
+      "step": 10436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.121028138142886e-05,
+      "loss": 2.4552,
+      "step": 10437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1250546980893094e-05,
+      "loss": 2.3806,
+      "step": 10438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1290829106234046e-05,
+      "loss": 2.2615,
+      "step": 10439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.133112775133036e-05,
+      "loss": 2.0936,
+      "step": 10440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.137144291005816e-05,
+      "loss": 2.2945,
+      "step": 10441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1411774576291094e-05,
+      "loss": 2.4675,
+      "step": 10442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.14521227439001e-05,
+      "loss": 2.723,
+      "step": 10443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.149248740675384e-05,
+      "loss": 2.4757,
+      "step": 10444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1532868558718364e-05,
+      "loss": 2.4682,
+      "step": 10445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.157326619365735e-05,
+      "loss": 0.7612,
+      "step": 10446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.161368030543167e-05,
+      "loss": 1.989,
+      "step": 10447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.165411088789996e-05,
+      "loss": 1.148,
+      "step": 10448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1694557934918295e-05,
+      "loss": 2.5109,
+      "step": 10449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.173502144034022e-05,
+      "loss": 2.7313,
+      "step": 10450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.177550139801686e-05,
+      "loss": 2.1185,
+      "step": 10451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.181599780179659e-05,
+      "loss": 2.2543,
+      "step": 10452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.185651064552557e-05,
+      "loss": 2.0939,
+      "step": 10453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1897039923047326e-05,
+      "loss": 1.5998,
+      "step": 10454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.193758562820304e-05,
+      "loss": 2.1554,
+      "step": 10455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.1978147754831045e-05,
+      "loss": 2.8314,
+      "step": 10456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.201872629676753e-05,
+      "loss": 2.3007,
+      "step": 10457
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2059321247846045e-05,
+      "loss": 2.4052,
+      "step": 10458
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.209993260189777e-05,
+      "loss": 2.0984,
+      "step": 10459
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2140560352751104e-05,
+      "loss": 2.2199,
+      "step": 10460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.218120449423222e-05,
+      "loss": 2.7985,
+      "step": 10461
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.222186502016471e-05,
+      "loss": 2.4698,
+      "step": 10462
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.226254192436976e-05,
+      "loss": 2.34,
+      "step": 10463
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2303235200665914e-05,
+      "loss": 2.4152,
+      "step": 10464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.234394484286929e-05,
+      "loss": 2.7364,
+      "step": 10465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.238467084479357e-05,
+      "loss": 2.0365,
+      "step": 10466
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2425413200249937e-05,
+      "loss": 2.0935,
+      "step": 10467
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2466171903047017e-05,
+      "loss": 2.6433,
+      "step": 10468
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2506946946991004e-05,
+      "loss": 2.8136,
+      "step": 10469
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2547738325885684e-05,
+      "loss": 2.161,
+      "step": 10470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2588546033532225e-05,
+      "loss": 2.4847,
+      "step": 10471
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2629370063729336e-05,
+      "loss": 2.258,
+      "step": 10472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.26702104102733e-05,
+      "loss": 2.8438,
+      "step": 10473
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2711067066957936e-05,
+      "loss": 2.5516,
+      "step": 10474
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.2751940027574604e-05,
+      "loss": 2.0537,
+      "step": 10475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.279282928591198e-05,
+      "loss": 2.3226,
+      "step": 10476
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.283373483575649e-05,
+      "loss": 2.7806,
+      "step": 10477
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.287465667089203e-05,
+      "loss": 2.4499,
+      "step": 10478
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.291559478510007e-05,
+      "loss": 2.4711,
+      "step": 10479
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.295654917215938e-05,
+      "loss": 2.75,
+      "step": 10480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.299751982584647e-05,
+      "loss": 2.4328,
+      "step": 10481
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.303850673993535e-05,
+      "loss": 2.7442,
+      "step": 10482
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.307950990819762e-05,
+      "loss": 1.8506,
+      "step": 10483
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3120529324402134e-05,
+      "loss": 2.1107,
+      "step": 10484
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.316156498231556e-05,
+      "loss": 1.34,
+      "step": 10485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3202616875702e-05,
+      "loss": 2.2266,
+      "step": 10486
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.32436849983232e-05,
+      "loss": 2.3113,
+      "step": 10487
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3284769343938106e-05,
+      "loss": 2.5156,
+      "step": 10488
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.332586990630354e-05,
+      "loss": 2.1043,
+      "step": 10489
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.336698667917375e-05,
+      "loss": 2.4683,
+      "step": 10490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3408119656300596e-05,
+      "loss": 2.7074,
+      "step": 10491
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.344926883143321e-05,
+      "loss": 2.6358,
+      "step": 10492
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3490434198318554e-05,
+      "loss": 2.9203,
+      "step": 10493
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.353161575070102e-05,
+      "loss": 3.124,
+      "step": 10494
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.357281348232254e-05,
+      "loss": 2.8028,
+      "step": 10495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.361402738692256e-05,
+      "loss": 2.8016,
+      "step": 10496
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.365525745823816e-05,
+      "loss": 1.9871,
+      "step": 10497
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.369650369000381e-05,
+      "loss": 1.6914,
+      "step": 10498
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.373776607595172e-05,
+      "loss": 2.5938,
+      "step": 10499
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.3779044609811425e-05,
+      "loss": 1.6587,
+      "step": 10500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.382033928531018e-05,
+      "loss": 2.3594,
+      "step": 10501
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.386165009617273e-05,
+      "loss": 1.6605,
+      "step": 10502
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.390297703612146e-05,
+      "loss": 2.3207,
+      "step": 10503
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.394432009887602e-05,
+      "loss": 2.4527,
+      "step": 10504
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.398567927815391e-05,
+      "loss": 1.9748,
+      "step": 10505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.402705456767004e-05,
+      "loss": 2.2903,
+      "step": 10506
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.406844596113701e-05,
+      "loss": 2.688,
+      "step": 10507
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4109853452264705e-05,
+      "loss": 2.3219,
+      "step": 10508
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4151277034760784e-05,
+      "loss": 2.527,
+      "step": 10509
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4192716702330404e-05,
+      "loss": 2.8161,
+      "step": 10510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.423417244867639e-05,
+      "loss": 2.2105,
+      "step": 10511
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.42756442674988e-05,
+      "loss": 3.2364,
+      "step": 10512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4317132152495574e-05,
+      "loss": 2.4979,
+      "step": 10513
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.435863609736208e-05,
+      "loss": 1.7172,
+      "step": 10514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4400156095791344e-05,
+      "loss": 2.7343,
+      "step": 10515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.44416921414737e-05,
+      "loss": 2.6733,
+      "step": 10516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.448324422809731e-05,
+      "loss": 2.5153,
+      "step": 10517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.452481234934777e-05,
+      "loss": 2.1977,
+      "step": 10518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.45663964989084e-05,
+      "loss": 1.1938,
+      "step": 10519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.4607996670459754e-05,
+      "loss": 2.2111,
+      "step": 10520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4649612857680226e-05,
+      "loss": 1.7007,
+      "step": 10521
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4691245054245724e-05,
+      "loss": 2.4735,
+      "step": 10522
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.473289325382968e-05,
+      "loss": 0.9103,
+      "step": 10523
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.47745574501032e-05,
+      "loss": 2.5091,
+      "step": 10524
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4816237636734786e-05,
+      "loss": 2.3639,
+      "step": 10525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.485793380739058e-05,
+      "loss": 2.3955,
+      "step": 10526
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.489964595573433e-05,
+      "loss": 2.8322,
+      "step": 10527
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4941374075427415e-05,
+      "loss": 2.262,
+      "step": 10528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4983118160128606e-05,
+      "loss": 1.8865,
+      "step": 10529
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.502487820349438e-05,
+      "loss": 2.3473,
+      "step": 10530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.506665419917886e-05,
+      "loss": 2.0802,
+      "step": 10531
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5108446140833555e-05,
+      "loss": 2.2797,
+      "step": 10532
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.51502540221076e-05,
+      "loss": 2.3633,
+      "step": 10533
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.519207783664782e-05,
+      "loss": 2.2717,
+      "step": 10534
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5233917578098517e-05,
+      "loss": 2.693,
+      "step": 10535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.527577324010174e-05,
+      "loss": 2.5498,
+      "step": 10536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.531764481629675e-05,
+      "loss": 2.6883,
+      "step": 10537
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.535953230032074e-05,
+      "loss": 2.4415,
+      "step": 10538
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.540143568580838e-05,
+      "loss": 2.5096,
+      "step": 10539
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5443354966392006e-05,
+      "loss": 2.3723,
+      "step": 10540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.548529013570121e-05,
+      "loss": 1.9702,
+      "step": 10541
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.552724118736356e-05,
+      "loss": 1.6747,
+      "step": 10542
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5569208115004014e-05,
+      "loss": 0.8635,
+      "step": 10543
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.561119091224527e-05,
+      "loss": 2.4914,
+      "step": 10544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5653189572707316e-05,
+      "loss": 2.1579,
+      "step": 10545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5695204090007984e-05,
+      "loss": 2.5563,
+      "step": 10546
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.573723445776266e-05,
+      "loss": 2.4655,
+      "step": 10547
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5779280669584365e-05,
+      "loss": 2.4074,
+      "step": 10548
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.582134271908347e-05,
+      "loss": 2.0399,
+      "step": 10549
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.586342059986818e-05,
+      "loss": 2.9496,
+      "step": 10550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.590551430554422e-05,
+      "loss": 2.6791,
+      "step": 10551
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.594762382971496e-05,
+      "loss": 2.3046,
+      "step": 10552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.598974916598127e-05,
+      "loss": 2.5082,
+      "step": 10553
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6031890307941635e-05,
+      "loss": 2.6594,
+      "step": 10554
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.607404724919219e-05,
+      "loss": 2.4598,
+      "step": 10555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.611621998332671e-05,
+      "loss": 2.109,
+      "step": 10556
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.615840850393639e-05,
+      "loss": 2.3202,
+      "step": 10557
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6200612804610256e-05,
+      "loss": 2.3543,
+      "step": 10558
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.624283287893471e-05,
+      "loss": 1.885,
+      "step": 10559
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6285068720494005e-05,
+      "loss": 2.5455,
+      "step": 10560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6327320322869716e-05,
+      "loss": 2.286,
+      "step": 10561
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.636958767964125e-05,
+      "loss": 2.1319,
+      "step": 10562
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6411870784385516e-05,
+      "loss": 1.8003,
+      "step": 10563
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.645416963067717e-05,
+      "loss": 2.206,
+      "step": 10564
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6496484212088153e-05,
+      "loss": 2.4939,
+      "step": 10565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.653881452218833e-05,
+      "loss": 2.7146,
+      "step": 10566
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.658116055454504e-05,
+      "loss": 1.9232,
+      "step": 10567
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.66235223027234e-05,
+      "loss": 2.8368,
+      "step": 10568
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.666589976028576e-05,
+      "loss": 2.3508,
+      "step": 10569
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.670829292079244e-05,
+      "loss": 2.4075,
+      "step": 10570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.675070177780125e-05,
+      "loss": 2.1856,
+      "step": 10571
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6793126324867705e-05,
+      "loss": 2.4752,
+      "step": 10572
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.683556655554468e-05,
+      "loss": 2.7181,
+      "step": 10573
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.687802246338292e-05,
+      "loss": 2.1204,
+      "step": 10574
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.692049404193071e-05,
+      "loss": 2.3646,
+      "step": 10575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6962981284734034e-05,
+      "loss": 2.4369,
+      "step": 10576
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.700548418533622e-05,
+      "loss": 2.1722,
+      "step": 10577
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7048002737278504e-05,
+      "loss": 2.5482,
+      "step": 10578
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.709053693409965e-05,
+      "loss": 2.1295,
+      "step": 10579
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.713308676933614e-05,
+      "loss": 2.5468,
+      "step": 10580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.717565223652177e-05,
+      "loss": 2.2055,
+      "step": 10581
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.721823332918828e-05,
+      "loss": 2.1559,
+      "step": 10582
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.726083004086493e-05,
+      "loss": 1.9054,
+      "step": 10583
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7303442365078634e-05,
+      "loss": 2.4035,
+      "step": 10584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7346070295353824e-05,
+      "loss": 2.42,
+      "step": 10585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.738871382521271e-05,
+      "loss": 1.6659,
+      "step": 10586
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.743137294817498e-05,
+      "loss": 1.9376,
+      "step": 10587
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7474047657758145e-05,
+      "loss": 2.7018,
+      "step": 10588
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.751673794747711e-05,
+      "loss": 2.1216,
+      "step": 10589
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.755944381084459e-05,
+      "loss": 2.3219,
+      "step": 10590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7602165241370896e-05,
+      "loss": 2.2842,
+      "step": 10591
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.764490223256405e-05,
+      "loss": 2.3195,
+      "step": 10592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.76876547779294e-05,
+      "loss": 2.6106,
+      "step": 10593
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.773042287097026e-05,
+      "loss": 2.204,
+      "step": 10594
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.777320650518748e-05,
+      "loss": 2.0022,
+      "step": 10595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.781600567407954e-05,
+      "loss": 2.7108,
+      "step": 10596
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.785882037114263e-05,
+      "loss": 2.4695,
+      "step": 10597
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7901650589870315e-05,
+      "loss": 2.2655,
+      "step": 10598
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.79444963237541e-05,
+      "loss": 2.2709,
+      "step": 10599
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.798735756628303e-05,
+      "loss": 2.9271,
+      "step": 10600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.803023431094388e-05,
+      "loss": 2.1854,
+      "step": 10601
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.807312655122079e-05,
+      "loss": 2.4309,
+      "step": 10602
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.811603428059581e-05,
+      "loss": 2.4183,
+      "step": 10603
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8158957492548575e-05,
+      "loss": 2.4531,
+      "step": 10604
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8201896180556436e-05,
+      "loss": 2.5986,
+      "step": 10605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.824485033809414e-05,
+      "loss": 2.3115,
+      "step": 10606
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.828781995863433e-05,
+      "loss": 2.6215,
+      "step": 10607
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8330805035647214e-05,
+      "loss": 2.9651,
+      "step": 10608
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.837380556260076e-05,
+      "loss": 2.3306,
+      "step": 10609
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.841682153296029e-05,
+      "loss": 2.4407,
+      "step": 10610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.845985294018908e-05,
+      "loss": 2.5608,
+      "step": 10611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.850289977774795e-05,
+      "loss": 2.3089,
+      "step": 10612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.854596203909543e-05,
+      "loss": 2.7448,
+      "step": 10613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8589039717687586e-05,
+      "loss": 2.2927,
+      "step": 10614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8632132806978215e-05,
+      "loss": 2.7936,
+      "step": 10615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.867524130041876e-05,
+      "loss": 2.3395,
+      "step": 10616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.87183651914584e-05,
+      "loss": 2.5101,
+      "step": 10617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.876150447354381e-05,
+      "loss": 2.2286,
+      "step": 10618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.880465914011954e-05,
+      "loss": 2.3998,
+      "step": 10619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.884782918462755e-05,
+      "loss": 2.8824,
+      "step": 10620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.889101460050773e-05,
+      "loss": 2.2825,
+      "step": 10621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.893421538119737e-05,
+      "loss": 2.6006,
+      "step": 10622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.897743152013162e-05,
+      "loss": 2.3224,
+      "step": 10623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.902066301074325e-05,
+      "loss": 2.5734,
+      "step": 10624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9063909846462746e-05,
+      "loss": 2.2887,
+      "step": 10625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.910717202071802e-05,
+      "loss": 2.2291,
+      "step": 10626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.915044952693492e-05,
+      "loss": 1.9078,
+      "step": 10627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.919374235853687e-05,
+      "loss": 2.5558,
+      "step": 10628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.923705050894507e-05,
+      "loss": 2.2391,
+      "step": 10629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9280373971578086e-05,
+      "loss": 2.3211,
+      "step": 10630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9323712739852484e-05,
+      "loss": 2.0148,
+      "step": 10631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9367066807182355e-05,
+      "loss": 2.2238,
+      "step": 10632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.94104361669796e-05,
+      "loss": 2.5792,
+      "step": 10633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.945382081265349e-05,
+      "loss": 2.2697,
+      "step": 10634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.949722073761126e-05,
+      "loss": 2.0093,
+      "step": 10635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.954063593525776e-05,
+      "loss": 2.1646,
+      "step": 10636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.958406639899556e-05,
+      "loss": 2.5438,
+      "step": 10637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9627512122224664e-05,
+      "loss": 2.6762,
+      "step": 10638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9670973098343024e-05,
+      "loss": 2.7728,
+      "step": 10639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9714449320746204e-05,
+      "loss": 2.0035,
+      "step": 10640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9757940782827514e-05,
+      "loss": 2.2383,
+      "step": 10641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9801447477977686e-05,
+      "loss": 2.6556,
+      "step": 10642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.98449693995854e-05,
+      "loss": 2.4313,
+      "step": 10643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.988850654103695e-05,
+      "loss": 2.349,
+      "step": 10644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.993205889571636e-05,
+      "loss": 2.4082,
+      "step": 10645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9975626457005176e-05,
+      "loss": 2.4331,
+      "step": 10646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.001920921828286e-05,
+      "loss": 2.0735,
+      "step": 10647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0062807172926354e-05,
+      "loss": 2.0211,
+      "step": 10648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.010642031431049e-05,
+      "loss": 2.4867,
+      "step": 10649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.015004863580758e-05,
+      "loss": 2.2957,
+      "step": 10650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.01936921307878e-05,
+      "loss": 2.3559,
+      "step": 10651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.023735079261894e-05,
+      "loss": 1.6795,
+      "step": 10652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.028102461466663e-05,
+      "loss": 2.8441,
+      "step": 10653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0324713590293864e-05,
+      "loss": 2.5885,
+      "step": 10654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.036841771286163e-05,
+      "loss": 2.621,
+      "step": 10655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.041213697572855e-05,
+      "loss": 2.5316,
+      "step": 10656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.045587137225099e-05,
+      "loss": 2.2487,
+      "step": 10657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.049962089578276e-05,
+      "loss": 2.1555,
+      "step": 10658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0543385539675665e-05,
+      "loss": 2.4053,
+      "step": 10659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0587165297279095e-05,
+      "loss": 2.7766,
+      "step": 10660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.063096016194025e-05,
+      "loss": 2.3484,
+      "step": 10661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.067477012700376e-05,
+      "loss": 1.659,
+      "step": 10662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0718595185812214e-05,
+      "loss": 2.5222,
+      "step": 10663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.076243533170583e-05,
+      "loss": 2.408,
+      "step": 10664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0806290558022644e-05,
+      "loss": 2.2035,
+      "step": 10665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.08501608580981e-05,
+      "loss": 1.9927,
+      "step": 10666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.089404622526563e-05,
+      "loss": 2.8342,
+      "step": 10667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.09379466528563e-05,
+      "loss": 1.8746,
+      "step": 10668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.098186213419885e-05,
+      "loss": 2.8439,
+      "step": 10669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.102579266261988e-05,
+      "loss": 2.4871,
+      "step": 10670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1069738231443385e-05,
+      "loss": 1.871,
+      "step": 10671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1113698833991357e-05,
+      "loss": 2.0732,
+      "step": 10672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.115767446358344e-05,
+      "loss": 2.459,
+      "step": 10673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1201665113536986e-05,
+      "loss": 2.6171,
+      "step": 10674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.124567077716703e-05,
+      "loss": 2.6222,
+      "step": 10675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.128969144778628e-05,
+      "loss": 2.7383,
+      "step": 10676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1333727118705284e-05,
+      "loss": 2.4257,
+      "step": 10677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1377777783232317e-05,
+      "loss": 2.2955,
+      "step": 10678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1421843434673186e-05,
+      "loss": 2.504,
+      "step": 10679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.146592406633164e-05,
+      "loss": 2.5632,
+      "step": 10680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.151001967150909e-05,
+      "loss": 2.1272,
+      "step": 10681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.155413024350457e-05,
+      "loss": 2.3207,
+      "step": 10682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.15982557756149e-05,
+      "loss": 2.64,
+      "step": 10683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1642396261134675e-05,
+      "loss": 2.1919,
+      "step": 10684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.168655169335615e-05,
+      "loss": 2.259,
+      "step": 10685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1730722065569494e-05,
+      "loss": 1.4978,
+      "step": 10686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.177490737106222e-05,
+      "loss": 2.532,
+      "step": 10687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.181910760311991e-05,
+      "loss": 2.3789,
+      "step": 10688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.186332275502576e-05,
+      "loss": 2.363,
+      "step": 10689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1907552820060816e-05,
+      "loss": 2.2718,
+      "step": 10690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.195179779150355e-05,
+      "loss": 2.4222,
+      "step": 10691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.199605766263049e-05,
+      "loss": 2.1563,
+      "step": 10692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2040332426715796e-05,
+      "loss": 2.7285,
+      "step": 10693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.208462207703132e-05,
+      "loss": 2.6563,
+      "step": 10694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.212892660684659e-05,
+      "loss": 2.4538,
+      "step": 10695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2173246009429126e-05,
+      "loss": 2.4889,
+      "step": 10696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.221758027804397e-05,
+      "loss": 2.4336,
+      "step": 10697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2261929405953924e-05,
+      "loss": 2.7649,
+      "step": 10698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.23062933864195e-05,
+      "loss": 2.3447,
+      "step": 10699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.23506722126992e-05,
+      "loss": 2.5078,
+      "step": 10700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.239506587804889e-05,
+      "loss": 2.7052,
+      "step": 10701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.243947437572268e-05,
+      "loss": 2.5164,
+      "step": 10702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2483897698971736e-05,
+      "loss": 2.444,
+      "step": 10703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2528335841045684e-05,
+      "loss": 2.0738,
+      "step": 10704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2572788795191325e-05,
+      "loss": 1.8258,
+      "step": 10705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.261725655465383e-05,
+      "loss": 1.8877,
+      "step": 10706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2661739112675295e-05,
+      "loss": 2.377,
+      "step": 10707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.2706236462496346e-05,
+      "loss": 2.5811,
+      "step": 10708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.275074859735482e-05,
+      "loss": 2.6059,
+      "step": 10709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.279527551048685e-05,
+      "loss": 2.6063,
+      "step": 10710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.28398171951256e-05,
+      "loss": 2.7141,
+      "step": 10711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.288437364450268e-05,
+      "loss": 2.034,
+      "step": 10712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.292894485184696e-05,
+      "loss": 2.2093,
+      "step": 10713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.297353081038559e-05,
+      "loss": 2.3992,
+      "step": 10714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.301813151334274e-05,
+      "loss": 2.0506,
+      "step": 10715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.306274695394106e-05,
+      "loss": 2.6247,
+      "step": 10716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.310737712540048e-05,
+      "loss": 1.5259,
+      "step": 10717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3152022020939056e-05,
+      "loss": 2.4835,
+      "step": 10718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.319668163377232e-05,
+      "loss": 2.0352,
+      "step": 10719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3241355957113686e-05,
+      "loss": 2.1852,
+      "step": 10720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3286044984174213e-05,
+      "loss": 2.274,
+      "step": 10721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.333074870816302e-05,
+      "loss": 2.351,
+      "step": 10722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3375467122286696e-05,
+      "loss": 1.7869,
+      "step": 10723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.342020021974962e-05,
+      "loss": 2.1894,
+      "step": 10724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.346494799375421e-05,
+      "loss": 2.5293,
+      "step": 10725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3509710437500386e-05,
+      "loss": 1.9499,
+      "step": 10726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.355448754418589e-05,
+      "loss": 2.8425,
+      "step": 10727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.359927930700621e-05,
+      "loss": 2.4567,
+      "step": 10728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.36440857191548e-05,
+      "loss": 2.1427,
+      "step": 10729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.368890677382271e-05,
+      "loss": 2.162,
+      "step": 10730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3733742464198767e-05,
+      "loss": 2.5661,
+      "step": 10731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.377859278346951e-05,
+      "loss": 2.2309,
+      "step": 10732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.382345772481958e-05,
+      "loss": 2.718,
+      "step": 10733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.3868337281431076e-05,
+      "loss": 2.5361,
+      "step": 10734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.391323144648395e-05,
+      "loss": 2.2525,
+      "step": 10735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.395814021315585e-05,
+      "loss": 2.3229,
+      "step": 10736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.400306357462255e-05,
+      "loss": 2.4799,
+      "step": 10737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4048001524057236e-05,
+      "loss": 2.3678,
+      "step": 10738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.409295405463103e-05,
+      "loss": 1.9494,
+      "step": 10739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4137921159512695e-05,
+      "loss": 2.4929,
+      "step": 10740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.418290283186912e-05,
+      "loss": 0.9184,
+      "step": 10741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4227899064864537e-05,
+      "loss": 2.7944,
+      "step": 10742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.427290985166152e-05,
+      "loss": 2.5329,
+      "step": 10743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.431793518541969e-05,
+      "loss": 2.23,
+      "step": 10744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.436297505929717e-05,
+      "loss": 2.4222,
+      "step": 10745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.440802946644938e-05,
+      "loss": 2.7605,
+      "step": 10746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.445309840003e-05,
+      "loss": 2.8345,
+      "step": 10747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.449818185318985e-05,
+      "loss": 2.2391,
+      "step": 10748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.454327981907821e-05,
+      "loss": 2.0263,
+      "step": 10749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4588392290841656e-05,
+      "loss": 2.2266,
+      "step": 10750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4633519261624994e-05,
+      "loss": 2.5021,
+      "step": 10751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4678660724570475e-05,
+      "loss": 2.5071,
+      "step": 10752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.472381667281829e-05,
+      "loss": 2.6658,
+      "step": 10753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4768987099506295e-05,
+      "loss": 2.5253,
+      "step": 10754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.481417199777049e-05,
+      "loss": 2.1333,
+      "step": 10755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.485937136074433e-05,
+      "loss": 2.2969,
+      "step": 10756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.490458518155908e-05,
+      "loss": 2.1489,
+      "step": 10757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.494981345334415e-05,
+      "loss": 1.8917,
+      "step": 10758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.4995056169226406e-05,
+      "loss": 2.5163,
+      "step": 10759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.504031332233063e-05,
+      "loss": 2.473,
+      "step": 10760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.508558490577932e-05,
+      "loss": 2.5267,
+      "step": 10761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.513087091269308e-05,
+      "loss": 2.2311,
+      "step": 10762
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.517617133619002e-05,
+      "loss": 2.3687,
+      "step": 10763
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.522148616938617e-05,
+      "loss": 2.5134,
+      "step": 10764
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.526681540539523e-05,
+      "loss": 2.1698,
+      "step": 10765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.531215903732907e-05,
+      "loss": 1.8755,
+      "step": 10766
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5357517058297015e-05,
+      "loss": 2.1724,
+      "step": 10767
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5402889461406366e-05,
+      "loss": 2.5894,
+      "step": 10768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.544827623976207e-05,
+      "loss": 2.2046,
+      "step": 10769
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5493677386467236e-05,
+      "loss": 2.6965,
+      "step": 10770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.553909289462246e-05,
+      "loss": 2.3274,
+      "step": 10771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5584522757326307e-05,
+      "loss": 2.4664,
+      "step": 10772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5629966967675004e-05,
+      "loss": 2.0453,
+      "step": 10773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.567542551876292e-05,
+      "loss": 2.3987,
+      "step": 10774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5720898403681935e-05,
+      "loss": 2.3854,
+      "step": 10775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.576638561552188e-05,
+      "loss": 2.6476,
+      "step": 10776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.581188714737028e-05,
+      "loss": 2.2486,
+      "step": 10777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.585740299231279e-05,
+      "loss": 2.8974,
+      "step": 10778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.590293314343259e-05,
+      "loss": 2.1324,
+      "step": 10779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.59484775938108e-05,
+      "loss": 2.0487,
+      "step": 10780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.5994036336526235e-05,
+      "loss": 2.5505,
+      "step": 10781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.60396093646559e-05,
+      "loss": 2.3905,
+      "step": 10782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.608519667127423e-05,
+      "loss": 2.4015,
+      "step": 10783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.61307982494536e-05,
+      "loss": 2.2034,
+      "step": 10784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.6176414092264424e-05,
+      "loss": 1.9615,
+      "step": 10785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.62220441927747e-05,
+      "loss": 2.4466,
+      "step": 10786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.6267688544050384e-05,
+      "loss": 2.2965,
+      "step": 10787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.631334713915508e-05,
+      "loss": 2.0151,
+      "step": 10788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.6359019971150595e-05,
+      "loss": 2.7758,
+      "step": 10789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.640470703309616e-05,
+      "loss": 2.6422,
+      "step": 10790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.6450408318049345e-05,
+      "loss": 2.4407,
+      "step": 10791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.649612381906484e-05,
+      "loss": 2.0966,
+      "step": 10792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.654185352919591e-05,
+      "loss": 2.8644,
+      "step": 10793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.65875974414931e-05,
+      "loss": 1.247,
+      "step": 10794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.663335554900536e-05,
+      "loss": 2.9986,
+      "step": 10795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.6679127844778753e-05,
+      "loss": 1.853,
+      "step": 10796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.67249143218579e-05,
+      "loss": 2.2907,
+      "step": 10797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.677071497328471e-05,
+      "loss": 2.451,
+      "step": 10798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.681652979209956e-05,
+      "loss": 2.2068,
+      "step": 10799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.686235877133986e-05,
+      "loss": 2.3875,
+      "step": 10800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.690820190404159e-05,
+      "loss": 2.2438,
+      "step": 10801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.6954059183238135e-05,
+      "loss": 2.7968,
+      "step": 10802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.699993060196118e-05,
+      "loss": 2.212,
+      "step": 10803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.704581615323955e-05,
+      "loss": 2.1673,
+      "step": 10804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.7091715830100686e-05,
+      "loss": 2.1756,
+      "step": 10805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.713762962556929e-05,
+      "loss": 2.5726,
+      "step": 10806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.718355753266843e-05,
+      "loss": 2.4068,
+      "step": 10807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.7229499544418625e-05,
+      "loss": 2.8567,
+      "step": 10808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.727545565383841e-05,
+      "loss": 2.1716,
+      "step": 10809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.732142585394408e-05,
+      "loss": 2.6895,
+      "step": 10810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.7367410137750044e-05,
+      "loss": 2.9289,
+      "step": 10811
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.7413408498268334e-05,
+      "loss": 2.6921,
+      "step": 10812
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.745942092850887e-05,
+      "loss": 2.3296,
+      "step": 10813
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.750544742147939e-05,
+      "loss": 2.2531,
+      "step": 10814
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.755148797018578e-05,
+      "loss": 2.4159,
+      "step": 10815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.759754256763148e-05,
+      "loss": 2.4984,
+      "step": 10816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.76436112068178e-05,
+      "loss": 2.4769,
+      "step": 10817
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.768969388074424e-05,
+      "loss": 2.6713,
+      "step": 10818
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.773579058240782e-05,
+      "loss": 2.2394,
+      "step": 10819
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.778190130480357e-05,
+      "loss": 2.8018,
+      "step": 10820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.7828026040924255e-05,
+      "loss": 1.2381,
+      "step": 10821
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.7874164783760864e-05,
+      "loss": 2.462,
+      "step": 10822
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.792031752630178e-05,
+      "loss": 2.7059,
+      "step": 10823
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.796648426153382e-05,
+      "loss": 2.1828,
+      "step": 10824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.801266498244096e-05,
+      "loss": 2.3201,
+      "step": 10825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.805885968200575e-05,
+      "loss": 1.9655,
+      "step": 10826
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.810506835320808e-05,
+      "loss": 2.2349,
+      "step": 10827
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.8151290989026284e-05,
+      "loss": 2.6492,
+      "step": 10828
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.819752758243581e-05,
+      "loss": 3.0572,
+      "step": 10829
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.824377812641077e-05,
+      "loss": 2.2242,
+      "step": 10830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.829004261392251e-05,
+      "loss": 2.0584,
+      "step": 10831
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.833632103794093e-05,
+      "loss": 1.8863,
+      "step": 10832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.838261339143299e-05,
+      "loss": 2.343,
+      "step": 10833
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.842891966736425e-05,
+      "loss": 2.3133,
+      "step": 10834
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.8475239858697735e-05,
+      "loss": 2.6715,
+      "step": 10835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.8521573958394746e-05,
+      "loss": 2.2381,
+      "step": 10836
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.8567921959413846e-05,
+      "loss": 2.4956,
+      "step": 10837
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.861428385471216e-05,
+      "loss": 2.2827,
+      "step": 10838
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.866065963724419e-05,
+      "loss": 2.0018,
+      "step": 10839
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.870704929996275e-05,
+      "loss": 1.586,
+      "step": 10840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.875345283581822e-05,
+      "loss": 1.7708,
+      "step": 10841
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.879987023775901e-05,
+      "loss": 2.275,
+      "step": 10842
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.884630149873128e-05,
+      "loss": 3.1072,
+      "step": 10843
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.8892746611679414e-05,
+      "loss": 2.7941,
+      "step": 10844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.893920556954538e-05,
+      "loss": 2.1468,
+      "step": 10845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.898567836526903e-05,
+      "loss": 2.6691,
+      "step": 10846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9032164991788434e-05,
+      "loss": 2.2538,
+      "step": 10847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.907866544203926e-05,
+      "loss": 2.5433,
+      "step": 10848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.912517970895516e-05,
+      "loss": 2.7116,
+      "step": 10849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.91717077854676e-05,
+      "loss": 2.6633,
+      "step": 10850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9218249664506226e-05,
+      "loss": 2.49,
+      "step": 10851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9264805338998295e-05,
+      "loss": 2.454,
+      "step": 10852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.931137480186911e-05,
+      "loss": 2.3347,
+      "step": 10853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.93579580460417e-05,
+      "loss": 2.2244,
+      "step": 10854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.940455506443739e-05,
+      "loss": 2.8104,
+      "step": 10855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.945116584997502e-05,
+      "loss": 2.2236,
+      "step": 10856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.949779039557151e-05,
+      "loss": 1.437,
+      "step": 10857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.954442869414153e-05,
+      "loss": 2.2064,
+      "step": 10858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9591080738598044e-05,
+      "loss": 2.4127,
+      "step": 10859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.963774652185152e-05,
+      "loss": 2.8134,
+      "step": 10860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9684426036810526e-05,
+      "loss": 2.2001,
+      "step": 10861
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.973111927638139e-05,
+      "loss": 2.4441,
+      "step": 10862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.977782623346868e-05,
+      "loss": 2.3748,
+      "step": 10863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.98245469009746e-05,
+      "loss": 2.2519,
+      "step": 10864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9871281271799305e-05,
+      "loss": 2.3432,
+      "step": 10865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.9918029338840816e-05,
+      "loss": 2.5868,
+      "step": 10866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.996479109499539e-05,
+      "loss": 2.3272,
+      "step": 10867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.001156653315684e-05,
+      "loss": 2.3074,
+      "step": 10868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.005835564621707e-05,
+      "loss": 2.7278,
+      "step": 10869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.010515842706575e-05,
+      "loss": 2.0883,
+      "step": 10870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.015197486859081e-05,
+      "loss": 2.7399,
+      "step": 10871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0198804963677795e-05,
+      "loss": 1.1386,
+      "step": 10872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.024564870521014e-05,
+      "loss": 2.4412,
+      "step": 10873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.029250608606956e-05,
+      "loss": 2.2379,
+      "step": 10874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.033937709913537e-05,
+      "loss": 2.3796,
+      "step": 10875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.038626173728494e-05,
+      "loss": 2.2629,
+      "step": 10876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.043315999339341e-05,
+      "loss": 1.8014,
+      "step": 10877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0480071860334195e-05,
+      "loss": 2.3119,
+      "step": 10878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0526997330978236e-05,
+      "loss": 2.5287,
+      "step": 10879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.057393639819492e-05,
+      "loss": 2.2306,
+      "step": 10880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0620889054850836e-05,
+      "loss": 2.0175,
+      "step": 10881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0667855293811246e-05,
+      "loss": 2.4064,
+      "step": 10882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.07148351079388e-05,
+      "loss": 2.0541,
+      "step": 10883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.076182849009461e-05,
+      "loss": 2.7453,
+      "step": 10884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.080883543313706e-05,
+      "loss": 1.1571,
+      "step": 10885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.085585592992312e-05,
+      "loss": 2.3357,
+      "step": 10886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.090288997330721e-05,
+      "loss": 1.8731,
+      "step": 10887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0949937556142126e-05,
+      "loss": 2.3437,
+      "step": 10888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.0996998671278286e-05,
+      "loss": 2.7031,
+      "step": 10889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.104407331156412e-05,
+      "loss": 2.6308,
+      "step": 10890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.109116146984595e-05,
+      "loss": 2.1494,
+      "step": 10891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.113826313896831e-05,
+      "loss": 2.4989,
+      "step": 10892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.118537831177342e-05,
+      "loss": 2.4187,
+      "step": 10893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.123250698110152e-05,
+      "loss": 2.0926,
+      "step": 10894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.127964913979069e-05,
+      "loss": 2.8138,
+      "step": 10895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.132680478067727e-05,
+      "loss": 2.2871,
+      "step": 10896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.137397389659528e-05,
+      "loss": 2.3585,
+      "step": 10897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.142115648037673e-05,
+      "loss": 1.1377,
+      "step": 10898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.146835252485157e-05,
+      "loss": 2.7387,
+      "step": 10899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.151556202284791e-05,
+      "loss": 2.503,
+      "step": 10900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.156278496719157e-05,
+      "loss": 2.4231,
+      "step": 10901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.161002135070643e-05,
+      "loss": 2.1767,
+      "step": 10902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.16572711662142e-05,
+      "loss": 2.2497,
+      "step": 10903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.170453440653486e-05,
+      "loss": 2.4292,
+      "step": 10904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.175181106448608e-05,
+      "loss": 2.3119,
+      "step": 10905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.179910113288345e-05,
+      "loss": 2.2453,
+      "step": 10906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.184640460454082e-05,
+      "loss": 1.7959,
+      "step": 10907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.189372147226974e-05,
+      "loss": 2.5317,
+      "step": 10908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.19410517288798e-05,
+      "loss": 2.1476,
+      "step": 10909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.198839536717847e-05,
+      "loss": 2.8708,
+      "step": 10910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.203575237997143e-05,
+      "loss": 2.749,
+      "step": 10911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.208312276006203e-05,
+      "loss": 1.8664,
+      "step": 10912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.213050650025201e-05,
+      "loss": 2.664,
+      "step": 10913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.21779035933404e-05,
+      "loss": 2.4514,
+      "step": 10914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.222531403212488e-05,
+      "loss": 2.6873,
+      "step": 10915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.227273780940067e-05,
+      "loss": 2.4294,
+      "step": 10916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.23201749179614e-05,
+      "loss": 2.3019,
+      "step": 10917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.236762535059797e-05,
+      "loss": 2.7924,
+      "step": 10918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.24150891001e-05,
+      "loss": 2.2212,
+      "step": 10919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.246256615925455e-05,
+      "loss": 2.1927,
+      "step": 10920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.25100565208472e-05,
+      "loss": 1.8269,
+      "step": 10921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.255756017766073e-05,
+      "loss": 2.303,
+      "step": 10922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.26050771224767e-05,
+      "loss": 2.7466,
+      "step": 10923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.265260734807406e-05,
+      "loss": 2.4881,
+      "step": 10924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.270015084723033e-05,
+      "loss": 2.1341,
+      "step": 10925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.27477076127202e-05,
+      "loss": 2.1315,
+      "step": 10926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.279527763731715e-05,
+      "loss": 2.1528,
+      "step": 10927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.284286091379207e-05,
+      "loss": 2.2914,
+      "step": 10928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.289045743491433e-05,
+      "loss": 2.6876,
+      "step": 10929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.293806719345088e-05,
+      "loss": 2.6331,
+      "step": 10930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.29856901821668e-05,
+      "loss": 2.5015,
+      "step": 10931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.30333263938251e-05,
+      "loss": 2.3096,
+      "step": 10932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.308097582118701e-05,
+      "loss": 2.1746,
+      "step": 10933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.312863845701153e-05,
+      "loss": 2.4149,
+      "step": 10934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.317631429405555e-05,
+      "loss": 1.8887,
+      "step": 10935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.322400332507436e-05,
+      "loss": 2.2967,
+      "step": 10936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.327170554282091e-05,
+      "loss": 2.6966,
+      "step": 10937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.33194209400462e-05,
+      "loss": 2.701,
+      "step": 10938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.336714950949917e-05,
+      "loss": 2.7907,
+      "step": 10939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.341489124392706e-05,
+      "loss": 2.5709,
+      "step": 10940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.346264613607478e-05,
+      "loss": 2.1122,
+      "step": 10941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.351041417868539e-05,
+      "loss": 2.0187,
+      "step": 10942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.355819536449978e-05,
+      "loss": 2.0587,
+      "step": 10943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.360598968625725e-05,
+      "loss": 2.336,
+      "step": 10944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.365379713669469e-05,
+      "loss": 1.9247,
+      "step": 10945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.370161770854715e-05,
+      "loss": 2.7116,
+      "step": 10946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.374945139454758e-05,
+      "loss": 2.5585,
+      "step": 10947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.379729818742724e-05,
+      "loss": 1.9976,
+      "step": 10948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.384515807991512e-05,
+      "loss": 2.6404,
+      "step": 10949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.389303106473823e-05,
+      "loss": 2.7458,
+      "step": 10950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.394091713462163e-05,
+      "loss": 2.5347,
+      "step": 10951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.398881628228856e-05,
+      "loss": 2.3878,
+      "step": 10952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.403672850046007e-05,
+      "loss": 2.6064,
+      "step": 10953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.408465378185525e-05,
+      "loss": 2.8963,
+      "step": 10954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.413259211919115e-05,
+      "loss": 2.5271,
+      "step": 10955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.418054350518313e-05,
+      "loss": 2.2379,
+      "step": 10956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.422850793254426e-05,
+      "loss": 1.979,
+      "step": 10957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.427648539398573e-05,
+      "loss": 2.3363,
+      "step": 10958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.432447588221661e-05,
+      "loss": 2.1745,
+      "step": 10959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.437247938994438e-05,
+      "loss": 2.217,
+      "step": 10960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.442049590987407e-05,
+      "loss": 2.6488,
+      "step": 10961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.446852543470916e-05,
+      "loss": 2.2679,
+      "step": 10962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.451656795715085e-05,
+      "loss": 2.4238,
+      "step": 10963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.456462346989844e-05,
+      "loss": 2.7413,
+      "step": 10964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.461269196564918e-05,
+      "loss": 2.4144,
+      "step": 10965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.466077343709869e-05,
+      "loss": 2.73,
+      "step": 10966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.47088678769402e-05,
+      "loss": 2.1679,
+      "step": 10967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.475697527786507e-05,
+      "loss": 2.4866,
+      "step": 10968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.4805095632563e-05,
+      "loss": 2.3494,
+      "step": 10969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.485322893372136e-05,
+      "loss": 2.3427,
+      "step": 10970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.490137517402566e-05,
+      "loss": 2.1025,
+      "step": 10971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.494953434615937e-05,
+      "loss": 2.2555,
+      "step": 10972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.499770644280428e-05,
+      "loss": 2.5166,
+      "step": 10973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.504589145663994e-05,
+      "loss": 2.644,
+      "step": 10974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.509408938034398e-05,
+      "loss": 2.6047,
+      "step": 10975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.514230020659207e-05,
+      "loss": 2.7335,
+      "step": 10976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.519052392805807e-05,
+      "loss": 2.3512,
+      "step": 10977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.523876053741376e-05,
+      "loss": 2.3972,
+      "step": 10978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.528701002732889e-05,
+      "loss": 2.415,
+      "step": 10979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.533527239047127e-05,
+      "loss": 2.7643,
+      "step": 10980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.5383547619507e-05,
+      "loss": 2.4231,
+      "step": 10981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.543183570709995e-05,
+      "loss": 2.2976,
+      "step": 10982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.548013664591212e-05,
+      "loss": 2.0779,
+      "step": 10983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.552845042860345e-05,
+      "loss": 2.7061,
+      "step": 10984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.557677704783223e-05,
+      "loss": 2.4235,
+      "step": 10985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.562511649625456e-05,
+      "loss": 2.512,
+      "step": 10986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.567346876652459e-05,
+      "loss": 0.8704,
+      "step": 10987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.572183385129448e-05,
+      "loss": 2.7734,
+      "step": 10988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.577021174321477e-05,
+      "loss": 1.9969,
+      "step": 10989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.581860243493368e-05,
+      "loss": 2.4675,
+      "step": 10990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.586700591909762e-05,
+      "loss": 1.9821,
+      "step": 10991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.591542218835098e-05,
+      "loss": 1.9113,
+      "step": 10992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.596385123533648e-05,
+      "loss": 2.6418,
+      "step": 10993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.601229305269461e-05,
+      "loss": 2.42,
+      "step": 10994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.606074763306388e-05,
+      "loss": 2.3748,
+      "step": 10995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.610921496908123e-05,
+      "loss": 2.1946,
+      "step": 10996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.615769505338133e-05,
+      "loss": 2.9606,
+      "step": 10997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.620618787859698e-05,
+      "loss": 2.8738,
+      "step": 10998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.625469343735899e-05,
+      "loss": 2.6782,
+      "step": 10999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.63032117222965e-05,
+      "loss": 2.7916,
+      "step": 11000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.635174272603634e-05,
+      "loss": 2.2415,
+      "step": 11001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.640028644120393e-05,
+      "loss": 1.839,
+      "step": 11002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.644884286042198e-05,
+      "loss": 2.5309,
+      "step": 11003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.649741197631206e-05,
+      "loss": 2.558,
+      "step": 11004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.654599378149323e-05,
+      "loss": 1.902,
+      "step": 11005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.659458826858323e-05,
+      "loss": 1.9185,
+      "step": 11006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.664319543019699e-05,
+      "loss": 2.8357,
+      "step": 11007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.669181525894843e-05,
+      "loss": 2.0344,
+      "step": 11008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.674044774744889e-05,
+      "loss": 2.1328,
+      "step": 11009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.678909288830837e-05,
+      "loss": 2.2904,
+      "step": 11010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.683775067413419e-05,
+      "loss": 2.0984,
+      "step": 11011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.688642109753248e-05,
+      "loss": 2.2344,
+      "step": 11012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.693510415110694e-05,
+      "loss": 2.5594,
+      "step": 11013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.698379982745988e-05,
+      "loss": 3.0164,
+      "step": 11014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.703250811919092e-05,
+      "loss": 2.8498,
+      "step": 11015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.708122901889855e-05,
+      "loss": 2.2202,
+      "step": 11016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.712996251917878e-05,
+      "loss": 2.2926,
+      "step": 11017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.717870861262627e-05,
+      "loss": 2.4334,
+      "step": 11018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.722746729183295e-05,
+      "loss": 2.6757,
+      "step": 11019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.72762385493897e-05,
+      "loss": 2.1847,
+      "step": 11020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.732502237788482e-05,
+      "loss": 2.8058,
+      "step": 11021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.737381876990524e-05,
+      "loss": 2.2801,
+      "step": 11022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.742262771803557e-05,
+      "loss": 2.0187,
+      "step": 11023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.74714492148587e-05,
+      "loss": 2.2235,
+      "step": 11024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.752028325295548e-05,
+      "loss": 2.606,
+      "step": 11025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.756912982490513e-05,
+      "loss": 2.1405,
+      "step": 11026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.76179889232847e-05,
+      "loss": 2.3952,
+      "step": 11027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.766686054066928e-05,
+      "loss": 2.097,
+      "step": 11028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.771574466963247e-05,
+      "loss": 2.7755,
+      "step": 11029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.776464130274556e-05,
+      "loss": 2.4422,
+      "step": 11030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.78135504325781e-05,
+      "loss": 2.2037,
+      "step": 11031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.786247205169756e-05,
+      "loss": 2.9102,
+      "step": 11032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.791140615266993e-05,
+      "loss": 2.2266,
+      "step": 11033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.796035272805882e-05,
+      "loss": 2.0533,
+      "step": 11034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.800931177042648e-05,
+      "loss": 2.3612,
+      "step": 11035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.805828327233256e-05,
+      "loss": 2.568,
+      "step": 11036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.810726722633549e-05,
+      "loss": 2.212,
+      "step": 11037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.815626362499135e-05,
+      "loss": 2.738,
+      "step": 11038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.820527246085478e-05,
+      "loss": 1.9025,
+      "step": 11039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.825429372647789e-05,
+      "loss": 2.4172,
+      "step": 11040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.830332741441155e-05,
+      "loss": 1.9289,
+      "step": 11041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.835237351720425e-05,
+      "loss": 2.5132,
+      "step": 11042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.840143202740316e-05,
+      "loss": 2.3557,
+      "step": 11043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.845050293755274e-05,
+      "loss": 2.5949,
+      "step": 11044
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.849958624019636e-05,
+      "loss": 2.3367,
+      "step": 11045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.854868192787499e-05,
+      "loss": 2.4732,
+      "step": 11046
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.859778999312821e-05,
+      "loss": 2.4788,
+      "step": 11047
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.864691042849302e-05,
+      "loss": 2.2918,
+      "step": 11048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.86960432265052e-05,
+      "loss": 2.3567,
+      "step": 11049
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.874518837969824e-05,
+      "loss": 2.2832,
+      "step": 11050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.87943458806041e-05,
+      "loss": 1.6215,
+      "step": 11051
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.884351572175253e-05,
+      "loss": 2.5367,
+      "step": 11052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.889269789567157e-05,
+      "loss": 2.4376,
+      "step": 11053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.894189239488722e-05,
+      "loss": 2.5769,
+      "step": 11054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.8991099211924e-05,
+      "loss": 2.3236,
+      "step": 11055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.904031833930417e-05,
+      "loss": 2.8149,
+      "step": 11056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.908954976954814e-05,
+      "loss": 1.9661,
+      "step": 11057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.913879349517479e-05,
+      "loss": 2.5048,
+      "step": 11058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.91880495087008e-05,
+      "loss": 2.6345,
+      "step": 11059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.923731780264106e-05,
+      "loss": 2.4536,
+      "step": 11060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.928659836950852e-05,
+      "loss": 2.4803,
+      "step": 11061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.93358912018146e-05,
+      "loss": 2.9838,
+      "step": 11062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.93851962920685e-05,
+      "loss": 1.7778,
+      "step": 11063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.943451363277768e-05,
+      "loss": 3.1995,
+      "step": 11064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.94838432164476e-05,
+      "loss": 2.0435,
+      "step": 11065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.953318503558227e-05,
+      "loss": 1.2437,
+      "step": 11066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.958253908268341e-05,
+      "loss": 1.7189,
+      "step": 11067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.963190535025106e-05,
+      "loss": 2.6439,
+      "step": 11068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.968128383078326e-05,
+      "loss": 1.5355,
+      "step": 11069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.973067451677655e-05,
+      "loss": 1.9428,
+      "step": 11070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.978007740072526e-05,
+      "loss": 2.6456,
+      "step": 11071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.982949247512199e-05,
+      "loss": 2.0024,
+      "step": 11072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.987891973245737e-05,
+      "loss": 2.712,
+      "step": 11073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.992835916522053e-05,
+      "loss": 2.5129,
+      "step": 11074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.997781076589839e-05,
+      "loss": 1.8383,
+      "step": 11075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.002727452697613e-05,
+      "loss": 2.2366,
+      "step": 11076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.007675044093698e-05,
+      "loss": 2.1124,
+      "step": 11077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.012623850026271e-05,
+      "loss": 1.7682,
+      "step": 11078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.017573869743284e-05,
+      "loss": 2.1639,
+      "step": 11079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.022525102492514e-05,
+      "loss": 2.3254,
+      "step": 11080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.027477547521552e-05,
+      "loss": 2.5015,
+      "step": 11081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.032431204077828e-05,
+      "loss": 2.4239,
+      "step": 11082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.037386071408563e-05,
+      "loss": 2.1597,
+      "step": 11083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.042342148760789e-05,
+      "loss": 1.9813,
+      "step": 11084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.047299435381386e-05,
+      "loss": 2.5812,
+      "step": 11085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.052257930517021e-05,
+      "loss": 2.6567,
+      "step": 11086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.05721763341419e-05,
+      "loss": 2.4744,
+      "step": 11087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.062178543319189e-05,
+      "loss": 2.636,
+      "step": 11088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.067140659478164e-05,
+      "loss": 2.3266,
+      "step": 11089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.072103981137038e-05,
+      "loss": 2.3069,
+      "step": 11090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.077068507541602e-05,
+      "loss": 2.3694,
+      "step": 11091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.08203423793739e-05,
+      "loss": 2.4423,
+      "step": 11092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.08700117156983e-05,
+      "loss": 2.2966,
+      "step": 11093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.091969307684105e-05,
+      "loss": 2.422,
+      "step": 11094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.096938645525281e-05,
+      "loss": 2.6478,
+      "step": 11095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.101909184338159e-05,
+      "loss": 2.7507,
+      "step": 11096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.106880923367432e-05,
+      "loss": 2.5515,
+      "step": 11097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.11185386185756e-05,
+      "loss": 2.3627,
+      "step": 11098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.116827999052875e-05,
+      "loss": 2.4821,
+      "step": 11099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.121803334197447e-05,
+      "loss": 1.2296,
+      "step": 11100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.126779866535242e-05,
+      "loss": 1.6417,
+      "step": 11101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.131757595309993e-05,
+      "loss": 2.5087,
+      "step": 11102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.13673651976529e-05,
+      "loss": 2.3729,
+      "step": 11103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.141716639144513e-05,
+      "loss": 2.564,
+      "step": 11104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.146697952690868e-05,
+      "loss": 2.7506,
+      "step": 11105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.151680459647365e-05,
+      "loss": 2.1858,
+      "step": 11106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.156664159256877e-05,
+      "loss": 2.364,
+      "step": 11107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.16164905076205e-05,
+      "loss": 2.5261,
+      "step": 11108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.16663513340537e-05,
+      "loss": 1.9342,
+      "step": 11109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.171622406429124e-05,
+      "loss": 2.1332,
+      "step": 11110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.176610869075453e-05,
+      "loss": 2.7374,
+      "step": 11111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.181600520586287e-05,
+      "loss": 2.5644,
+      "step": 11112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.186591360203386e-05,
+      "loss": 2.1762,
+      "step": 11113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.191583387168313e-05,
+      "loss": 2.6606,
+      "step": 11114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.19657660072249e-05,
+      "loss": 2.4625,
+      "step": 11115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.201571000107121e-05,
+      "loss": 2.4955,
+      "step": 11116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.206566584563233e-05,
+      "loss": 2.3717,
+      "step": 11117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.211563353331706e-05,
+      "loss": 2.9586,
+      "step": 11118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.216561305653202e-05,
+      "loss": 2.5986,
+      "step": 11119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.221560440768224e-05,
+      "loss": 2.4781,
+      "step": 11120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.22656075791707e-05,
+      "loss": 2.3245,
+      "step": 11121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.231562256339907e-05,
+      "loss": 2.6681,
+      "step": 11122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.236564935276667e-05,
+      "loss": 2.3627,
+      "step": 11123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.241568793967166e-05,
+      "loss": 2.137,
+      "step": 11124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.246573831650952e-05,
+      "loss": 2.7156,
+      "step": 11125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.251580047567485e-05,
+      "loss": 2.5849,
+      "step": 11126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.256587440955981e-05,
+      "loss": 2.1592,
+      "step": 11127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.261596011055534e-05,
+      "loss": 2.0933,
+      "step": 11128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.266605757104987e-05,
+      "loss": 2.4609,
+      "step": 11129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.271616678343075e-05,
+      "loss": 1.626,
+      "step": 11130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.276628774008304e-05,
+      "loss": 2.2232,
+      "step": 11131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.281642043339054e-05,
+      "loss": 2.3881,
+      "step": 11132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.286656485573448e-05,
+      "loss": 2.3808,
+      "step": 11133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.291672099949514e-05,
+      "loss": 2.4736,
+      "step": 11134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.29668888570504e-05,
+      "loss": 2.4012,
+      "step": 11135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.301706842077695e-05,
+      "loss": 2.4261,
+      "step": 11136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.306725968304893e-05,
+      "loss": 1.9978,
+      "step": 11137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.311746263623944e-05,
+      "loss": 2.2997,
+      "step": 11138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.31676772727193e-05,
+      "loss": 3.0216,
+      "step": 11139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.321790358485807e-05,
+      "loss": 2.192,
+      "step": 11140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.326814156502277e-05,
+      "loss": 2.2926,
+      "step": 11141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.331839120557945e-05,
+      "loss": 2.6881,
+      "step": 11142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.336865249889176e-05,
+      "loss": 1.9157,
+      "step": 11143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.341892543732212e-05,
+      "loss": 2.6042,
+      "step": 11144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.346921001323078e-05,
+      "loss": 1.8603,
+      "step": 11145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.351950621897635e-05,
+      "loss": 2.3373,
+      "step": 11146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.356981404691557e-05,
+      "loss": 2.5607,
+      "step": 11147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.362013348940377e-05,
+      "loss": 2.7097,
+      "step": 11148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.367046453879409e-05,
+      "loss": 2.3868,
+      "step": 11149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.3720807187438e-05,
+      "loss": 2.6108,
+      "step": 11150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.377116142768553e-05,
+      "loss": 2.6167,
+      "step": 11151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.382152725188458e-05,
+      "loss": 2.2125,
+      "step": 11152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.387190465238142e-05,
+      "loss": 1.6288,
+      "step": 11153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.392229362152042e-05,
+      "loss": 2.5291,
+      "step": 11154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.397269415164457e-05,
+      "loss": 2.7966,
+      "step": 11155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.402310623509476e-05,
+      "loss": 2.4867,
+      "step": 11156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.407352986421023e-05,
+      "loss": 2.7139,
+      "step": 11157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.412396503132832e-05,
+      "loss": 2.5829,
+      "step": 11158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.417441172878502e-05,
+      "loss": 2.4541,
+      "step": 11159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.422486994891415e-05,
+      "loss": 2.7314,
+      "step": 11160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.427533968404797e-05,
+      "loss": 1.6829,
+      "step": 11161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.432582092651682e-05,
+      "loss": 2.2664,
+      "step": 11162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.437631366864967e-05,
+      "loss": 2.6505,
+      "step": 11163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.442681790277338e-05,
+      "loss": 2.8672,
+      "step": 11164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.447733362121319e-05,
+      "loss": 2.4761,
+      "step": 11165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.452786081629243e-05,
+      "loss": 1.7827,
+      "step": 11166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.457839948033316e-05,
+      "loss": 2.0542,
+      "step": 11167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.462894960565518e-05,
+      "loss": 1.4087,
+      "step": 11168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.467951118457679e-05,
+      "loss": 2.1886,
+      "step": 11169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.473008420941438e-05,
+      "loss": 2.0138,
+      "step": 11170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.478066867248298e-05,
+      "loss": 2.1018,
+      "step": 11171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.48312645660955e-05,
+      "loss": 2.5371,
+      "step": 11172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.488187188256311e-05,
+      "loss": 2.6768,
+      "step": 11173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.493249061419565e-05,
+      "loss": 2.4637,
+      "step": 11174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.49831207533008e-05,
+      "loss": 2.4236,
+      "step": 11175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.503376229218457e-05,
+      "loss": 2.5089,
+      "step": 11176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.508441522315154e-05,
+      "loss": 2.2599,
+      "step": 11177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.513507953850425e-05,
+      "loss": 2.4839,
+      "step": 11178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.518575523054347e-05,
+      "loss": 1.5832,
+      "step": 11179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.523644229156863e-05,
+      "loss": 2.2057,
+      "step": 11180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.528714071387705e-05,
+      "loss": 2.2703,
+      "step": 11181
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.533785048976444e-05,
+      "loss": 2.2242,
+      "step": 11182
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.538857161152472e-05,
+      "loss": 2.8085,
+      "step": 11183
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.543930407145036e-05,
+      "loss": 1.9147,
+      "step": 11184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.54900478618318e-05,
+      "loss": 2.349,
+      "step": 11185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.554080297495791e-05,
+      "loss": 2.2382,
+      "step": 11186
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.559156940311563e-05,
+      "loss": 2.4867,
+      "step": 11187
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.564234713859061e-05,
+      "loss": 2.2271,
+      "step": 11188
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.569313617366639e-05,
+      "loss": 1.9262,
+      "step": 11189
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.574393650062492e-05,
+      "loss": 2.1833,
+      "step": 11190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.579474811174634e-05,
+      "loss": 2.7808,
+      "step": 11191
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.58455709993094e-05,
+      "loss": 1.9082,
+      "step": 11192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.589640515559078e-05,
+      "loss": 2.5629,
+      "step": 11193
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.59472505728656e-05,
+      "loss": 2.5409,
+      "step": 11194
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.59981072434071e-05,
+      "loss": 2.5276,
+      "step": 11195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.604897515948719e-05,
+      "loss": 2.2959,
+      "step": 11196
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.609985431337576e-05,
+      "loss": 2.4577,
+      "step": 11197
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.615074469734103e-05,
+      "loss": 2.485,
+      "step": 11198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.620164630364944e-05,
+      "loss": 2.2232,
+      "step": 11199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.625255912456609e-05,
+      "loss": 2.5178,
+      "step": 11200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.6303483152354e-05,
+      "loss": 2.5302,
+      "step": 11201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.635441837927459e-05,
+      "loss": 2.6848,
+      "step": 11202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.640536479758751e-05,
+      "loss": 2.5936,
+      "step": 11203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.6456322399551e-05,
+      "loss": 2.2416,
+      "step": 11204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.650729117742128e-05,
+      "loss": 2.0953,
+      "step": 11205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.655827112345293e-05,
+      "loss": 2.0271,
+      "step": 11206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.660926222989903e-05,
+      "loss": 1.6526,
+      "step": 11207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.66602644890108e-05,
+      "loss": 2.4807,
+      "step": 11208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.671127789303776e-05,
+      "loss": 2.4618,
+      "step": 11209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.676230243422764e-05,
+      "loss": 2.3254,
+      "step": 11210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.681333810482685e-05,
+      "loss": 2.3607,
+      "step": 11211
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.686438489707961e-05,
+      "loss": 2.2294,
+      "step": 11212
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.691544280322907e-05,
+      "loss": 2.1743,
+      "step": 11213
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.696651181551585e-05,
+      "loss": 2.2968,
+      "step": 11214
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.701759192617974e-05,
+      "loss": 2.5112,
+      "step": 11215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.706868312745818e-05,
+      "loss": 2.8946,
+      "step": 11216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.711978541158759e-05,
+      "loss": 2.2953,
+      "step": 11217
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.717089877080184e-05,
+      "loss": 2.4534,
+      "step": 11218
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.722202319733395e-05,
+      "loss": 2.2232,
+      "step": 11219
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.727315868341467e-05,
+      "loss": 2.7452,
+      "step": 11220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.732430522127369e-05,
+      "loss": 2.4581,
+      "step": 11221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.737546280313811e-05,
+      "loss": 2.5423,
+      "step": 11222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.742663142123427e-05,
+      "loss": 2.2213,
+      "step": 11223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.747781106778617e-05,
+      "loss": 2.7227,
+      "step": 11224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.75290017350168e-05,
+      "loss": 2.0585,
+      "step": 11225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.75802034151466e-05,
+      "loss": 2.264,
+      "step": 11226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.763141610039515e-05,
+      "loss": 2.0353,
+      "step": 11227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.76826397829798e-05,
+      "loss": 2.3589,
+      "step": 11228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.77338744551168e-05,
+      "loss": 1.9059,
+      "step": 11229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.778512010901995e-05,
+      "loss": 2.5135,
+      "step": 11230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.783637673690211e-05,
+      "loss": 2.132,
+      "step": 11231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.788764433097399e-05,
+      "loss": 2.3066,
+      "step": 11232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.793892288344501e-05,
+      "loss": 2.13,
+      "step": 11233
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.799021238652265e-05,
+      "loss": 1.8376,
+      "step": 11234
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.80415128324128e-05,
+      "loss": 2.9578,
+      "step": 11235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.80928242133196e-05,
+      "loss": 2.0336,
+      "step": 11236
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.814414652144583e-05,
+      "loss": 2.3132,
+      "step": 11237
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.81954797489923e-05,
+      "loss": 2.627,
+      "step": 11238
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.824682388815817e-05,
+      "loss": 2.217,
+      "step": 11239
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.829817893114124e-05,
+      "loss": 2.7171,
+      "step": 11240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.834954487013736e-05,
+      "loss": 2.6602,
+      "step": 11241
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.840092169734079e-05,
+      "loss": 2.5964,
+      "step": 11242
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.845230940494406e-05,
+      "loss": 2.7181,
+      "step": 11243
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.850370798513837e-05,
+      "loss": 2.5476,
+      "step": 11244
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.855511743011297e-05,
+      "loss": 1.7437,
+      "step": 11245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.860653773205548e-05,
+      "loss": 2.3792,
+      "step": 11246
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.865796888315184e-05,
+      "loss": 2.5582,
+      "step": 11247
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.870941087558665e-05,
+      "loss": 1.9363,
+      "step": 11248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.87608637015424e-05,
+      "loss": 2.7777,
+      "step": 11249
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.881232735320053e-05,
+      "loss": 2.7498,
+      "step": 11250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.886380182273999e-05,
+      "loss": 1.7346,
+      "step": 11251
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.891528710233896e-05,
+      "loss": 2.5453,
+      "step": 11252
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.896678318417333e-05,
+      "loss": 2.5197,
+      "step": 11253
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.901829006041795e-05,
+      "loss": 2.3545,
+      "step": 11254
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.906980772324526e-05,
+      "loss": 2.5254,
+      "step": 11255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.91213361648268e-05,
+      "loss": 2.1975,
+      "step": 11256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.917287537733195e-05,
+      "loss": 2.7652,
+      "step": 11257
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.9224425352929e-05,
+      "loss": 2.2473,
+      "step": 11258
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.927598608378383e-05,
+      "loss": 3.0215,
+      "step": 11259
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.932755756206146e-05,
+      "loss": 2.3359,
+      "step": 11260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.937913977992467e-05,
+      "loss": 2.7215,
+      "step": 11261
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.943073272953527e-05,
+      "loss": 2.2142,
+      "step": 11262
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.948233640305257e-05,
+      "loss": 2.4239,
+      "step": 11263
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.95339507926351e-05,
+      "loss": 2.3166,
+      "step": 11264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.958557589043913e-05,
+      "loss": 2.8611,
+      "step": 11265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.96372116886198e-05,
+      "loss": 2.1182,
+      "step": 11266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.968885817933029e-05,
+      "loss": 2.7671,
+      "step": 11267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.974051535472224e-05,
+      "loss": 2.4229,
+      "step": 11268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.979218320694553e-05,
+      "loss": 1.9343,
+      "step": 11269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.984386172814885e-05,
+      "loss": 1.6367,
+      "step": 11270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.989555091047885e-05,
+      "loss": 2.5146,
+      "step": 11271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.994725074608055e-05,
+      "loss": 2.4629,
+      "step": 11272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.999896122709778e-05,
+      "loss": 2.4448,
+      "step": 11273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.005068234567234e-05,
+      "loss": 2.4245,
+      "step": 11274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.010241409394448e-05,
+      "loss": 2.264,
+      "step": 11275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.015415646405286e-05,
+      "loss": 2.2147,
+      "step": 11276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.020590944813476e-05,
+      "loss": 2.3295,
+      "step": 11277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.025767303832551e-05,
+      "loss": 2.0902,
+      "step": 11278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.030944722675901e-05,
+      "loss": 2.296,
+      "step": 11279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.036123200556737e-05,
+      "loss": 2.601,
+      "step": 11280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.041302736688144e-05,
+      "loss": 2.6642,
+      "step": 11281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.046483330283019e-05,
+      "loss": 2.3248,
+      "step": 11282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.0516649805541e-05,
+      "loss": 2.9971,
+      "step": 11283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.056847686713956e-05,
+      "loss": 2.237,
+      "step": 11284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.062031447975034e-05,
+      "loss": 3.0275,
+      "step": 11285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.067216263549581e-05,
+      "loss": 2.3253,
+      "step": 11286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.072402132649703e-05,
+      "loss": 2.5346,
+      "step": 11287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.07758905448732e-05,
+      "loss": 2.5134,
+      "step": 11288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.082777028274246e-05,
+      "loss": 2.5473,
+      "step": 11289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.087966053222081e-05,
+      "loss": 1.9153,
+      "step": 11290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.093156128542291e-05,
+      "loss": 2.8261,
+      "step": 11291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.098347253446164e-05,
+      "loss": 2.4571,
+      "step": 11292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.103539427144869e-05,
+      "loss": 2.2633,
+      "step": 11293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.108732648849369e-05,
+      "loss": 1.1855,
+      "step": 11294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.113926917770484e-05,
+      "loss": 1.267,
+      "step": 11295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.119122233118894e-05,
+      "loss": 2.4505,
+      "step": 11296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.124318594105098e-05,
+      "loss": 2.4231,
+      "step": 11297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.129515999939442e-05,
+      "loss": 1.9406,
+      "step": 11298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.134714449832099e-05,
+      "loss": 2.2501,
+      "step": 11299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.13991394299312e-05,
+      "loss": 2.6845,
+      "step": 11300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.145114478632356e-05,
+      "loss": 2.4906,
+      "step": 11301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.15031605595955e-05,
+      "loss": 2.1588,
+      "step": 11302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.155518674184211e-05,
+      "loss": 2.3131,
+      "step": 11303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.160722332515769e-05,
+      "loss": 1.7563,
+      "step": 11304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.16592703016344e-05,
+      "loss": 1.3746,
+      "step": 11305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.171132766336334e-05,
+      "loss": 1.003,
+      "step": 11306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.176339540243326e-05,
+      "loss": 2.2856,
+      "step": 11307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.18154735109322e-05,
+      "loss": 2.6283,
+      "step": 11308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.186756198094593e-05,
+      "loss": 2.4279,
+      "step": 11309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.191966080455933e-05,
+      "loss": 2.1199,
+      "step": 11310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.197176997385482e-05,
+      "loss": 2.6388,
+      "step": 11311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.202388948091409e-05,
+      "loss": 2.7248,
+      "step": 11312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.207601931781673e-05,
+      "loss": 2.6817,
+      "step": 11313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.212815947664121e-05,
+      "loss": 2.2927,
+      "step": 11314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.218030994946378e-05,
+      "loss": 2.1247,
+      "step": 11315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.22324707283598e-05,
+      "loss": 2.7372,
+      "step": 11316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.228464180540254e-05,
+      "loss": 2.2879,
+      "step": 11317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.233682317266429e-05,
+      "loss": 2.1525,
+      "step": 11318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.238901482221498e-05,
+      "loss": 2.5734,
+      "step": 11319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.244121674612373e-05,
+      "loss": 2.3501,
+      "step": 11320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.249342893645757e-05,
+      "loss": 2.0409,
+      "step": 11321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.254565138528244e-05,
+      "loss": 2.8569,
+      "step": 11322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.259788408466231e-05,
+      "loss": 2.3338,
+      "step": 11323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.26501270266598e-05,
+      "loss": 2.6348,
+      "step": 11324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.270238020333574e-05,
+      "loss": 2.663,
+      "step": 11325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.275464360674989e-05,
+      "loss": 1.9384,
+      "step": 11326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.280691722896e-05,
+      "loss": 2.2617,
+      "step": 11327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.285920106202232e-05,
+      "loss": 2.2843,
+      "step": 11328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.291149509799187e-05,
+      "loss": 2.2877,
+      "step": 11329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.296379932892179e-05,
+      "loss": 2.5063,
+      "step": 11330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.301611374686378e-05,
+      "loss": 1.8413,
+      "step": 11331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.306843834386786e-05,
+      "loss": 2.6302,
+      "step": 11332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.31207731119829e-05,
+      "loss": 2.1602,
+      "step": 11333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.31731180432557e-05,
+      "loss": 2.5387,
+      "step": 11334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.322547312973215e-05,
+      "loss": 2.1515,
+      "step": 11335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.327783836345572e-05,
+      "loss": 2.4022,
+      "step": 11336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.333021373646921e-05,
+      "loss": 2.4099,
+      "step": 11337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.338259924081328e-05,
+      "loss": 2.7455,
+      "step": 11338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.343499486852763e-05,
+      "loss": 2.1143,
+      "step": 11339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.348740061164959e-05,
+      "loss": 2.7078,
+      "step": 11340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.353981646221582e-05,
+      "loss": 2.2551,
+      "step": 11341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.359224241226075e-05,
+      "loss": 1.2481,
+      "step": 11342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.364467845381802e-05,
+      "loss": 1.7778,
+      "step": 11343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.369712457891876e-05,
+      "loss": 2.1941,
+      "step": 11344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.374958077959353e-05,
+      "loss": 2.6489,
+      "step": 11345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.380204704787064e-05,
+      "loss": 2.5005,
+      "step": 11346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.385452337577755e-05,
+      "loss": 2.2976,
+      "step": 11347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.390700975533931e-05,
+      "loss": 2.3091,
+      "step": 11348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.395950617858033e-05,
+      "loss": 1.9907,
+      "step": 11349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.401201263752285e-05,
+      "loss": 1.9231,
+      "step": 11350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.406452912418819e-05,
+      "loss": 2.0002,
+      "step": 11351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.411705563059533e-05,
+      "loss": 2.1424,
+      "step": 11352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.416959214876255e-05,
+      "loss": 2.327,
+      "step": 11353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.422213867070599e-05,
+      "loss": 2.3058,
+      "step": 11354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.427469518844082e-05,
+      "loss": 1.9293,
+      "step": 11355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.43272616939802e-05,
+      "loss": 2.4699,
+      "step": 11356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.437983817933605e-05,
+      "loss": 2.2771,
+      "step": 11357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.443242463651854e-05,
+      "loss": 2.2879,
+      "step": 11358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.448502105753677e-05,
+      "loss": 2.9182,
+      "step": 11359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.453762743439782e-05,
+      "loss": 2.2836,
+      "step": 11360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.459024375910746e-05,
+      "loss": 2.3648,
+      "step": 11361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.464287002367015e-05,
+      "loss": 2.5045,
+      "step": 11362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.469550622008854e-05,
+      "loss": 2.1442,
+      "step": 11363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.474815234036388e-05,
+      "loss": 2.3281,
+      "step": 11364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.48008083764958e-05,
+      "loss": 2.1732,
+      "step": 11365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.485347432048279e-05,
+      "loss": 2.0325,
+      "step": 11366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.490615016432145e-05,
+      "loss": 2.563,
+      "step": 11367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.495883590000699e-05,
+      "loss": 2.4948,
+      "step": 11368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.501153151953306e-05,
+      "loss": 2.6628,
+      "step": 11369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.506423701489211e-05,
+      "loss": 2.3573,
+      "step": 11370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.511695237807472e-05,
+      "loss": 2.8528,
+      "step": 11371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.516967760107011e-05,
+      "loss": 2.3223,
+      "step": 11372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.522241267586592e-05,
+      "loss": 2.2727,
+      "step": 11373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.52751575944486e-05,
+      "loss": 2.5647,
+      "step": 11374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.532791234880275e-05,
+      "loss": 2.2216,
+      "step": 11375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.538067693091163e-05,
+      "loss": 2.3192,
+      "step": 11376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.543345133275685e-05,
+      "loss": 2.3376,
+      "step": 11377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.548623554631889e-05,
+      "loss": 2.2319,
+      "step": 11378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.553902956357641e-05,
+      "loss": 2.0041,
+      "step": 11379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.559183337650668e-05,
+      "loss": 2.9146,
+      "step": 11380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.564464697708534e-05,
+      "loss": 2.4748,
+      "step": 11381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.569747035728698e-05,
+      "loss": 2.4954,
+      "step": 11382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.575030350908422e-05,
+      "loss": 1.4713,
+      "step": 11383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.580314642444831e-05,
+      "loss": 2.34,
+      "step": 11384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.58559990953493e-05,
+      "loss": 2.3406,
+      "step": 11385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.590886151375544e-05,
+      "loss": 2.4208,
+      "step": 11386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.596173367163362e-05,
+      "loss": 2.3155,
+      "step": 11387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.601461556094909e-05,
+      "loss": 1.957,
+      "step": 11388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.606750717366603e-05,
+      "loss": 2.1918,
+      "step": 11389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.612040850174663e-05,
+      "loss": 1.8781,
+      "step": 11390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.617331953715222e-05,
+      "loss": 1.786,
+      "step": 11391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.622624027184178e-05,
+      "loss": 2.5049,
+      "step": 11392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.627917069777372e-05,
+      "loss": 2.1765,
+      "step": 11393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.633211080690434e-05,
+      "loss": 2.7473,
+      "step": 11394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.638506059118893e-05,
+      "loss": 2.3873,
+      "step": 11395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.643802004258099e-05,
+      "loss": 2.0342,
+      "step": 11396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.649098915303264e-05,
+      "loss": 2.1028,
+      "step": 11397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.654396791449439e-05,
+      "loss": 2.2306,
+      "step": 11398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.659695631891575e-05,
+      "loss": 2.4722,
+      "step": 11399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.664995435824426e-05,
+      "loss": 2.6574,
+      "step": 11400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.670296202442624e-05,
+      "loss": 2.348,
+      "step": 11401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.675597930940633e-05,
+      "loss": 2.3513,
+      "step": 11402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.680900620512816e-05,
+      "loss": 2.1856,
+      "step": 11403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.686204270353344e-05,
+      "loss": 1.3404,
+      "step": 11404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.691508879656264e-05,
+      "loss": 2.7539,
+      "step": 11405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.696814447615456e-05,
+      "loss": 2.3214,
+      "step": 11406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.702120973424697e-05,
+      "loss": 2.4892,
+      "step": 11407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.707428456277584e-05,
+      "loss": 2.3988,
+      "step": 11408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.71273689536757e-05,
+      "loss": 2.4877,
+      "step": 11409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.718046289887962e-05,
+      "loss": 2.3942,
+      "step": 11410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.723356639031949e-05,
+      "loss": 2.0953,
+      "step": 11411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.728667941992546e-05,
+      "loss": 1.3969,
+      "step": 11412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.733980197962637e-05,
+      "loss": 2.1822,
+      "step": 11413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.739293406134934e-05,
+      "loss": 2.2116,
+      "step": 11414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.744607565702061e-05,
+      "loss": 2.4036,
+      "step": 11415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.749922675856444e-05,
+      "loss": 2.0079,
+      "step": 11416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.755238735790375e-05,
+      "loss": 2.4542,
+      "step": 11417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.760555744696032e-05,
+      "loss": 2.6543,
+      "step": 11418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.76587370176542e-05,
+      "loss": 2.3907,
+      "step": 11419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.771192606190403e-05,
+      "loss": 2.3672,
+      "step": 11420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.776512457162694e-05,
+      "loss": 2.1436,
+      "step": 11421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.781833253873898e-05,
+      "loss": 2.8674,
+      "step": 11422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.787154995515426e-05,
+      "loss": 2.5197,
+      "step": 11423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.792477681278607e-05,
+      "loss": 2.0745,
+      "step": 11424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.797801310354545e-05,
+      "loss": 2.8436,
+      "step": 11425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.80312588193428e-05,
+      "loss": 2.0831,
+      "step": 11426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.808451395208652e-05,
+      "loss": 2.36,
+      "step": 11427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.813777849368418e-05,
+      "loss": 2.5553,
+      "step": 11428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.819105243604102e-05,
+      "loss": 2.2721,
+      "step": 11429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.824433577106178e-05,
+      "loss": 2.5172,
+      "step": 11430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.829762849064913e-05,
+      "loss": 2.3861,
+      "step": 11431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.835093058670491e-05,
+      "loss": 2.2184,
+      "step": 11432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.840424205112871e-05,
+      "loss": 2.3206,
+      "step": 11433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.84575628758195e-05,
+      "loss": 2.2011,
+      "step": 11434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.85108930526743e-05,
+      "loss": 1.8702,
+      "step": 11435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.856423257358924e-05,
+      "loss": 2.3338,
+      "step": 11436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.86175814304582e-05,
+      "loss": 2.06,
+      "step": 11437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.867093961517452e-05,
+      "loss": 2.4406,
+      "step": 11438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.872430711962948e-05,
+      "loss": 2.3314,
+      "step": 11439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.877768393571359e-05,
+      "loss": 1.567,
+      "step": 11440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.883107005531504e-05,
+      "loss": 2.1742,
+      "step": 11441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.88844654703215e-05,
+      "loss": 1.9814,
+      "step": 11442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.893787017261863e-05,
+      "loss": 1.8766,
+      "step": 11443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.899128415409111e-05,
+      "loss": 2.7183,
+      "step": 11444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.904470740662188e-05,
+      "loss": 2.2401,
+      "step": 11445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.909813992209261e-05,
+      "loss": 2.3359,
+      "step": 11446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.915158169238342e-05,
+      "loss": 1.9084,
+      "step": 11447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.920503270937337e-05,
+      "loss": 2.299,
+      "step": 11448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.925849296493978e-05,
+      "loss": 2.0692,
+      "step": 11449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.931196245095856e-05,
+      "loss": 2.4127,
+      "step": 11450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.936544115930456e-05,
+      "loss": 2.357,
+      "step": 11451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.941892908185091e-05,
+      "loss": 2.4531,
+      "step": 11452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.947242621046942e-05,
+      "loss": 2.1266,
+      "step": 11453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.952593253703039e-05,
+      "loss": 2.1137,
+      "step": 11454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.957944805340309e-05,
+      "loss": 2.9174,
+      "step": 11455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.963297275145503e-05,
+      "loss": 2.2097,
+      "step": 11456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.968650662305242e-05,
+      "loss": 2.8866,
+      "step": 11457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.974004966006003e-05,
+      "loss": 2.1602,
+      "step": 11458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.979360185434151e-05,
+      "loss": 2.2703,
+      "step": 11459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.984716319775881e-05,
+      "loss": 2.5535,
+      "step": 11460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.99007336821726e-05,
+      "loss": 2.637,
+      "step": 11461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.9954313299442e-05,
+      "loss": 2.1681,
+      "step": 11462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.000790204142519e-05,
+      "loss": 2.6361,
+      "step": 11463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.006149989997851e-05,
+      "loss": 2.3659,
+      "step": 11464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.011510686695711e-05,
+      "loss": 2.5712,
+      "step": 11465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.01687229342146e-05,
+      "loss": 2.373,
+      "step": 11466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.022234809360353e-05,
+      "loss": 2.1649,
+      "step": 11467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.027598233697468e-05,
+      "loss": 2.3958,
+      "step": 11468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.032962565617796e-05,
+      "loss": 2.4875,
+      "step": 11469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.038327804306113e-05,
+      "loss": 2.8845,
+      "step": 11470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.043693948947134e-05,
+      "loss": 2.5022,
+      "step": 11471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.049060998725383e-05,
+      "loss": 2.1579,
+      "step": 11472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.054428952825305e-05,
+      "loss": 2.3616,
+      "step": 11473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.05979781043112e-05,
+      "loss": 2.3804,
+      "step": 11474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.065167570726996e-05,
+      "loss": 2.2209,
+      "step": 11475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.070538232896908e-05,
+      "loss": 2.2262,
+      "step": 11476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.07590979612474e-05,
+      "loss": 2.2634,
+      "step": 11477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.081282259594196e-05,
+      "loss": 1.8467,
+      "step": 11478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.086655622488867e-05,
+      "loss": 2.1982,
+      "step": 11479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.092029883992187e-05,
+      "loss": 2.6469,
+      "step": 11480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.097405043287492e-05,
+      "loss": 2.7097,
+      "step": 11481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.102781099557947e-05,
+      "loss": 2.4526,
+      "step": 11482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.108158051986576e-05,
+      "loss": 2.4113,
+      "step": 11483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.113535899756307e-05,
+      "loss": 2.6306,
+      "step": 11484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.118914642049897e-05,
+      "loss": 2.1833,
+      "step": 11485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.124294278049976e-05,
+      "loss": 2.6726,
+      "step": 11486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.129674806939027e-05,
+      "loss": 2.2175,
+      "step": 11487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.135056227899433e-05,
+      "loss": 2.0671,
+      "step": 11488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.140438540113408e-05,
+      "loss": 2.4015,
+      "step": 11489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.145821742763036e-05,
+      "loss": 2.5165,
+      "step": 11490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.151205835030264e-05,
+      "loss": 2.5648,
+      "step": 11491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.156590816096929e-05,
+      "loss": 2.1382,
+      "step": 11492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.161976685144704e-05,
+      "loss": 2.6647,
+      "step": 11493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.167363441355137e-05,
+      "loss": 2.6995,
+      "step": 11494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.17275108390963e-05,
+      "loss": 1.4259,
+      "step": 11495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.178139611989483e-05,
+      "loss": 2.5062,
+      "step": 11496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.183529024775833e-05,
+      "loss": 2.3822,
+      "step": 11497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.188919321449683e-05,
+      "loss": 2.0664,
+      "step": 11498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.194310501191901e-05,
+      "loss": 2.4375,
+      "step": 11499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.199702563183249e-05,
+      "loss": 2.0723,
+      "step": 11500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.205095506604323e-05,
+      "loss": 2.112,
+      "step": 11501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.210489330635595e-05,
+      "loss": 3.0255,
+      "step": 11502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.215884034457395e-05,
+      "loss": 1.9857,
+      "step": 11503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.22127961724995e-05,
+      "loss": 1.7601,
+      "step": 11504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.226676078193322e-05,
+      "loss": 2.3032,
+      "step": 11505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.232073416467432e-05,
+      "loss": 2.3318,
+      "step": 11506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.237471631252115e-05,
+      "loss": 2.639,
+      "step": 11507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.24287072172703e-05,
+      "loss": 1.8893,
+      "step": 11508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.248270687071714e-05,
+      "loss": 2.5982,
+      "step": 11509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.25367152646556e-05,
+      "loss": 2.5133,
+      "step": 11510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.259073239087868e-05,
+      "loss": 2.234,
+      "step": 11511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.264475824117752e-05,
+      "loss": 2.7293,
+      "step": 11512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.269879280734257e-05,
+      "loss": 2.7819,
+      "step": 11513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275283608116212e-05,
+      "loss": 2.3479,
+      "step": 11514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280688805442392e-05,
+      "loss": 2.8055,
+      "step": 11515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286094871891386e-05,
+      "loss": 2.4316,
+      "step": 11516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29150180664171e-05,
+      "loss": 1.8392,
+      "step": 11517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296909608871661e-05,
+      "loss": 2.4574,
+      "step": 11518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302318277759488e-05,
+      "loss": 2.186,
+      "step": 11519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307727812483256e-05,
+      "loss": 2.4069,
+      "step": 11520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.313138212220947e-05,
+      "loss": 2.3877,
+      "step": 11521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318549476150338e-05,
+      "loss": 2.3406,
+      "step": 11522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323961603449149e-05,
+      "loss": 2.7182,
+      "step": 11523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329374593294917e-05,
+      "loss": 1.9791,
+      "step": 11524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.334788444865105e-05,
+      "loss": 1.1032,
+      "step": 11525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.34020315733696e-05,
+      "loss": 1.9837,
+      "step": 11526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.34561872988768e-05,
+      "loss": 1.9028,
+      "step": 11527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.351035161694277e-05,
+      "loss": 2.3372,
+      "step": 11528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.356452451933694e-05,
+      "loss": 2.3045,
+      "step": 11529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.361870599782652e-05,
+      "loss": 2.1359,
+      "step": 11530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.367289604417832e-05,
+      "loss": 2.6625,
+      "step": 11531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.372709465015722e-05,
+      "loss": 2.8823,
+      "step": 11532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.378130180752727e-05,
+      "loss": 2.4748,
+      "step": 11533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.38355175080509e-05,
+      "loss": 2.641,
+      "step": 11534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.38897417434893e-05,
+      "loss": 2.4081,
+      "step": 11535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.394397450560233e-05,
+      "loss": 1.8026,
+      "step": 11536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.399821578614883e-05,
+      "loss": 1.8569,
+      "step": 11537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.405246557688604e-05,
+      "loss": 2.3682,
+      "step": 11538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.410672386956987e-05,
+      "loss": 2.3128,
+      "step": 11539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.416099065595535e-05,
+      "loss": 1.9922,
+      "step": 11540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.421526592779578e-05,
+      "loss": 3.1253,
+      "step": 11541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.426954967684335e-05,
+      "loss": 2.7336,
+      "step": 11542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432384189484885e-05,
+      "loss": 2.6753,
+      "step": 11543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437814257356208e-05,
+      "loss": 2.3073,
+      "step": 11544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443245170473115e-05,
+      "loss": 3.0114,
+      "step": 11545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448676928010343e-05,
+      "loss": 2.044,
+      "step": 11546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.454109529142418e-05,
+      "loss": 1.7755,
+      "step": 11547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459542973043824e-05,
+      "loss": 2.9894,
+      "step": 11548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464977258888857e-05,
+      "loss": 1.9355,
+      "step": 11549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470412385851739e-05,
+      "loss": 2.9031,
+      "step": 11550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.475848353106488e-05,
+      "loss": 1.0836,
+      "step": 11551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481285159827076e-05,
+      "loss": 2.4456,
+      "step": 11552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486722805187281e-05,
+      "loss": 2.0303,
+      "step": 11553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492161288360824e-05,
+      "loss": 2.3593,
+      "step": 11554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497600608521207e-05,
+      "loss": 2.7866,
+      "step": 11555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503040764841892e-05,
+      "loss": 2.4099,
+      "step": 11556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508481756496153e-05,
+      "loss": 2.2811,
+      "step": 11557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5139235826572e-05,
+      "loss": 2.3071,
+      "step": 11558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519366242498027e-05,
+      "loss": 2.2467,
+      "step": 11559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524809735191587e-05,
+      "loss": 2.5522,
+      "step": 11560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.530254059910651e-05,
+      "loss": 2.5085,
+      "step": 11561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.535699215827917e-05,
+      "loss": 2.4405,
+      "step": 11562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.541145202115876e-05,
+      "loss": 2.3715,
+      "step": 11563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.546592017946975e-05,
+      "loss": 2.4877,
+      "step": 11564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.55203966249348e-05,
+      "loss": 1.3568,
+      "step": 11565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.557488134927575e-05,
+      "loss": 2.2631,
+      "step": 11566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.562937434421279e-05,
+      "loss": 2.4748,
+      "step": 11567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.568387560146503e-05,
+      "loss": 2.4695,
+      "step": 11568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.573838511275019e-05,
+      "loss": 2.1122,
+      "step": 11569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.579290286978509e-05,
+      "loss": 2.5293,
+      "step": 11570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.584742886428491e-05,
+      "loss": 2.8417,
+      "step": 11571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.590196308796364e-05,
+      "loss": 2.6407,
+      "step": 11572
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.595650553253432e-05,
+      "loss": 2.6206,
+      "step": 11573
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.60110561897084e-05,
+      "loss": 1.6942,
+      "step": 11574
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.606561505119625e-05,
+      "loss": 1.6581,
+      "step": 11575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.612018210870678e-05,
+      "loss": 2.2387,
+      "step": 11576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.617475735394807e-05,
+      "loss": 1.9163,
+      "step": 11577
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.622934077862663e-05,
+      "loss": 2.1744,
+      "step": 11578
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.628393237444776e-05,
+      "loss": 2.7086,
+      "step": 11579
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.633853213311548e-05,
+      "loss": 2.3958,
+      "step": 11580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.639314004633288e-05,
+      "loss": 2.3167,
+      "step": 11581
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.644775610580146e-05,
+      "loss": 2.3617,
+      "step": 11582
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.650238030322162e-05,
+      "loss": 2.5364,
+      "step": 11583
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.655701263029239e-05,
+      "loss": 2.7701,
+      "step": 11584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.661165307871193e-05,
+      "loss": 2.1777,
+      "step": 11585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.666630164017678e-05,
+      "loss": 2.5379,
+      "step": 11586
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.672095830638242e-05,
+      "loss": 2.039,
+      "step": 11587
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.67756230690229e-05,
+      "loss": 2.6191,
+      "step": 11588
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.683029591979151e-05,
+      "loss": 2.2375,
+      "step": 11589
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.688497685037984e-05,
+      "loss": 1.9821,
+      "step": 11590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.693966585247843e-05,
+      "loss": 2.2728,
+      "step": 11591
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.699436291777649e-05,
+      "loss": 2.2246,
+      "step": 11592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.704906803796228e-05,
+      "loss": 2.0486,
+      "step": 11593
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.710378120472262e-05,
+      "loss": 2.2911,
+      "step": 11594
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.715850240974306e-05,
+      "loss": 2.1026,
+      "step": 11595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.721323164470792e-05,
+      "loss": 2.7586,
+      "step": 11596
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.726796890130064e-05,
+      "loss": 2.6021,
+      "step": 11597
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.732271417120307e-05,
+      "loss": 2.3125,
+      "step": 11598
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.737746744609584e-05,
+      "loss": 2.0034,
+      "step": 11599
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.743222871765869e-05,
+      "loss": 2.6072,
+      "step": 11600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.748699797756987e-05,
+      "loss": 2.6339,
+      "step": 11601
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.754177521750647e-05,
+      "loss": 3.0058,
+      "step": 11602
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.759656042914426e-05,
+      "loss": 2.2932,
+      "step": 11603
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.765135360415819e-05,
+      "loss": 2.5202,
+      "step": 11604
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.770615473422144e-05,
+      "loss": 2.249,
+      "step": 11605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.776096381100668e-05,
+      "loss": 2.9085,
+      "step": 11606
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.781578082618447e-05,
+      "loss": 2.3458,
+      "step": 11607
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.787060577142503e-05,
+      "loss": 1.8585,
+      "step": 11608
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.792543863839673e-05,
+      "loss": 1.8942,
+      "step": 11609
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.798027941876741e-05,
+      "loss": 2.6851,
+      "step": 11610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.803512810420282e-05,
+      "loss": 2.1812,
+      "step": 11611
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.808998468636835e-05,
+      "loss": 2.8589,
+      "step": 11612
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.814484915692761e-05,
+      "loss": 2.8186,
+      "step": 11613
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.819972150754343e-05,
+      "loss": 2.585,
+      "step": 11614
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.825460172987721e-05,
+      "loss": 2.3659,
+      "step": 11615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.830948981558917e-05,
+      "loss": 2.3406,
+      "step": 11616
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.836438575633818e-05,
+      "loss": 2.7994,
+      "step": 11617
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.841928954378242e-05,
+      "loss": 1.8488,
+      "step": 11618
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.84742011695784e-05,
+      "loss": 2.7343,
+      "step": 11619
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.852912062538157e-05,
+      "loss": 2.6905,
+      "step": 11620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.858404790284611e-05,
+      "loss": 1.5807,
+      "step": 11621
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.863898299362537e-05,
+      "loss": 2.1729,
+      "step": 11622
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.869392588937111e-05,
+      "loss": 2.5157,
+      "step": 11623
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.874887658173406e-05,
+      "loss": 2.2142,
+      "step": 11624
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.880383506236367e-05,
+      "loss": 2.0897,
+      "step": 11625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.885880132290848e-05,
+      "loss": 2.7044,
+      "step": 11626
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.891377535501556e-05,
+      "loss": 2.5177,
+      "step": 11627
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.89687571503308e-05,
+      "loss": 2.5119,
+      "step": 11628
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.902374670049923e-05,
+      "loss": 2.3901,
+      "step": 11629
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.907874399716436e-05,
+      "loss": 2.633,
+      "step": 11630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.913374903196867e-05,
+      "loss": 2.272,
+      "step": 11631
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.918876179655325e-05,
+      "loss": 2.1235,
+      "step": 11632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.924378228255854e-05,
+      "loss": 1.6453,
+      "step": 11633
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.929881048162315e-05,
+      "loss": 2.7375,
+      "step": 11634
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.935384638538524e-05,
+      "loss": 2.6106,
+      "step": 11635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.94088899854809e-05,
+      "loss": 1.56,
+      "step": 11636
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.946394127354592e-05,
+      "loss": 2.5956,
+      "step": 11637
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.95190002412143e-05,
+      "loss": 2.1276,
+      "step": 11638
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.957406688011949e-05,
+      "loss": 2.5898,
+      "step": 11639
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.962914118189291e-05,
+      "loss": 2.8548,
+      "step": 11640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.96842231381657e-05,
+      "loss": 2.4001,
+      "step": 11641
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.973931274056718e-05,
+      "loss": 2.7663,
+      "step": 11642
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.979440998072617e-05,
+      "loss": 2.5736,
+      "step": 11643
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.984951485026942e-05,
+      "loss": 2.3219,
+      "step": 11644
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.990462734082341e-05,
+      "loss": 2.3526,
+      "step": 11645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.995974744401285e-05,
+      "loss": 2.7193,
+      "step": 11646
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010001487515146193,
+      "loss": 2.4468,
+      "step": 11647
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010007001045479272,
+      "loss": 2.1358,
+      "step": 11648
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010012515334562715,
+      "loss": 2.0701,
+      "step": 11649
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010018030381558526,
+      "loss": 2.2949,
+      "step": 11650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001002354618562866,
+      "loss": 2.449,
+      "step": 11651
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010029062745934871,
+      "loss": 2.0729,
+      "step": 11652
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010034580061638885,
+      "loss": 2.1085,
+      "step": 11653
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001004009813190225,
+      "loss": 1.83,
+      "step": 11654
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010045616955886446,
+      "loss": 2.4346,
+      "step": 11655
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010051136532752806,
+      "loss": 2.3921,
+      "step": 11656
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010056656861662561,
+      "loss": 2.5014,
+      "step": 11657
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010062177941776813,
+      "loss": 2.5035,
+      "step": 11658
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010067699772256587,
+      "loss": 2.4908,
+      "step": 11659
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010073222352262757,
+      "loss": 1.8867,
+      "step": 11660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010078745680956088,
+      "loss": 2.3858,
+      "step": 11661
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010084269757497258,
+      "loss": 1.941,
+      "step": 11662
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010089794581046801,
+      "loss": 2.6157,
+      "step": 11663
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010095320150765151,
+      "loss": 2.0706,
+      "step": 11664
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010100846465812615,
+      "loss": 2.6744,
+      "step": 11665
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010106373525349424,
+      "loss": 2.3228,
+      "step": 11666
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010111901328535656,
+      "loss": 2.5505,
+      "step": 11667
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010117429874531289,
+      "loss": 2.4961,
+      "step": 11668
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010122959162496178,
+      "loss": 2.7202,
+      "step": 11669
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010128489191590103,
+      "loss": 2.2417,
+      "step": 11670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001013401996097269,
+      "loss": 2.0728,
+      "step": 11671
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010139551469803471,
+      "loss": 2.5541,
+      "step": 11672
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010145083717241846,
+      "loss": 2.5698,
+      "step": 11673
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010150616702447146,
+      "loss": 2.4204,
+      "step": 11674
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010156150424578549,
+      "loss": 1.9654,
+      "step": 11675
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010161684882795138,
+      "loss": 2.0969,
+      "step": 11676
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010167220076255865,
+      "loss": 2.6612,
+      "step": 11677
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010172756004119611,
+      "loss": 2.4309,
+      "step": 11678
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010178292665545111,
+      "loss": 2.4583,
+      "step": 11679
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010183830059690996,
+      "loss": 2.7464,
+      "step": 11680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010189368185715777,
+      "loss": 2.8494,
+      "step": 11681
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010194907042777889,
+      "loss": 2.0741,
+      "step": 11682
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010200446630035616,
+      "loss": 2.138,
+      "step": 11683
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010205986946647152,
+      "loss": 2.2392,
+      "step": 11684
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010211527991770557,
+      "loss": 2.4504,
+      "step": 11685
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010217069764563826,
+      "loss": 2.3378,
+      "step": 11686
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010222612264184791,
+      "loss": 2.2815,
+      "step": 11687
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001022815548979122,
+      "loss": 1.5978,
+      "step": 11688
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010233699440540737,
+      "loss": 2.4612,
+      "step": 11689
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010239244115590871,
+      "loss": 2.6849,
+      "step": 11690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010244789514099021,
+      "loss": 2.4331,
+      "step": 11691
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010250335635222517,
+      "loss": 2.8907,
+      "step": 11692
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010255882478118545,
+      "loss": 2.5967,
+      "step": 11693
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010261430041944178,
+      "loss": 2.2277,
+      "step": 11694
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010266978325856416,
+      "loss": 2.0637,
+      "step": 11695
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010272527329012114,
+      "loss": 2.0499,
+      "step": 11696
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010278077050568028,
+      "loss": 2.6167,
+      "step": 11697
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010283627489680798,
+      "loss": 2.4795,
+      "step": 11698
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010289178645506985,
+      "loss": 2.423,
+      "step": 11699
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010294730517203009,
+      "loss": 2.387,
+      "step": 11700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001030028310392519,
+      "loss": 2.6079,
+      "step": 11701
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010305836404829729,
+      "loss": 2.4476,
+      "step": 11702
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010311390419072758,
+      "loss": 2.0517,
+      "step": 11703
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010316945145810258,
+      "loss": 2.1678,
+      "step": 11704
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010322500584198118,
+      "loss": 2.1335,
+      "step": 11705
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010328056733392105,
+      "loss": 2.2411,
+      "step": 11706
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010333613592547915,
+      "loss": 2.8549,
+      "step": 11707
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000103391711608211,
+      "loss": 1.9928,
+      "step": 11708
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010344729437367118,
+      "loss": 2.4616,
+      "step": 11709
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010350288421341298,
+      "loss": 2.2734,
+      "step": 11710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010355848111898916,
+      "loss": 2.5712,
+      "step": 11711
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010361408508195087,
+      "loss": 2.3673,
+      "step": 11712
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010366969609384839,
+      "loss": 2.454,
+      "step": 11713
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010372531414623076,
+      "loss": 2.1015,
+      "step": 11714
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010378093923064638,
+      "loss": 2.629,
+      "step": 11715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010383657133864218,
+      "loss": 2.8269,
+      "step": 11716
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010389221046176413,
+      "loss": 2.2949,
+      "step": 11717
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010394785659155704,
+      "loss": 2.0861,
+      "step": 11718
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010400350971956502,
+      "loss": 2.5943,
+      "step": 11719
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010405916983733071,
+      "loss": 1.0131,
+      "step": 11720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010411483693639574,
+      "loss": 2.3992,
+      "step": 11721
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010417051100830102,
+      "loss": 2.2945,
+      "step": 11722
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010422619204458605,
+      "loss": 2.6493,
+      "step": 11723
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010428188003678935,
+      "loss": 2.3368,
+      "step": 11724
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010433757497644829,
+      "loss": 1.8543,
+      "step": 11725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010439327685509958,
+      "loss": 2.5708,
+      "step": 11726
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010444898566427836,
+      "loss": 3.0569,
+      "step": 11727
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010450470139551933,
+      "loss": 2.5049,
+      "step": 11728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010456042404035527,
+      "loss": 2.2309,
+      "step": 11729
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010461615359031873,
+      "loss": 2.388,
+      "step": 11730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010467189003694072,
+      "loss": 2.6458,
+      "step": 11731
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010472763337175171,
+      "loss": 2.4217,
+      "step": 11732
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010478338358628027,
+      "loss": 2.6488,
+      "step": 11733
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010483914067205482,
+      "loss": 2.1823,
+      "step": 11734
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010489490462060212,
+      "loss": 1.9598,
+      "step": 11735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001049506754234485,
+      "loss": 2.1185,
+      "step": 11736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010500645307211833,
+      "loss": 2.3341,
+      "step": 11737
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001050622375581359,
+      "loss": 2.5179,
+      "step": 11738
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010511802887302377,
+      "loss": 2.0259,
+      "step": 11739
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010517382700830413,
+      "loss": 2.4,
+      "step": 11740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010522963195549719,
+      "loss": 2.2799,
+      "step": 11741
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010528544370612306,
+      "loss": 2.2558,
+      "step": 11742
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010534126225170022,
+      "loss": 2.4118,
+      "step": 11743
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010539708758374654,
+      "loss": 2.5891,
+      "step": 11744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010545291969377851,
+      "loss": 2.6522,
+      "step": 11745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010550875857331174,
+      "loss": 2.2663,
+      "step": 11746
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010556460421386069,
+      "loss": 2.2333,
+      "step": 11747
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001056204566069391,
+      "loss": 2.2057,
+      "step": 11748
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001056763157440594,
+      "loss": 2.9452,
+      "step": 11749
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010573218161673292,
+      "loss": 2.122,
+      "step": 11750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001057880542164704,
+      "loss": 2.2983,
+      "step": 11751
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010584393353478112,
+      "loss": 2.0195,
+      "step": 11752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010589981956317352,
+      "loss": 2.7594,
+      "step": 11753
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010595571229315488,
+      "loss": 2.3296,
+      "step": 11754
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010601161171623181,
+      "loss": 2.9991,
+      "step": 11755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010606751782390959,
+      "loss": 2.9426,
+      "step": 11756
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010612343060769256,
+      "loss": 2.1258,
+      "step": 11757
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010617935005908387,
+      "loss": 2.4682,
+      "step": 11758
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010623527616958614,
+      "loss": 2.0942,
+      "step": 11759
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010629120893070046,
+      "loss": 2.5581,
+      "step": 11760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010634714833392742,
+      "loss": 1.9727,
+      "step": 11761
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010640309437076589,
+      "loss": 3.2634,
+      "step": 11762
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010645904703271448,
+      "loss": 2.3492,
+      "step": 11763
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010651500631127022,
+      "loss": 2.0735,
+      "step": 11764
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010657097219792975,
+      "loss": 2.5227,
+      "step": 11765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010662694468418786,
+      "loss": 2.5776,
+      "step": 11766
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010668292376153915,
+      "loss": 1.4017,
+      "step": 11767
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010673890942147668,
+      "loss": 2.0633,
+      "step": 11768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010679490165549303,
+      "loss": 2.4308,
+      "step": 11769
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010685090045507899,
+      "loss": 2.5606,
+      "step": 11770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001069069058117252,
+      "loss": 2.2602,
+      "step": 11771
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010696291771692064,
+      "loss": 1.9458,
+      "step": 11772
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000107018936162154,
+      "loss": 2.65,
+      "step": 11773
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010707496113891207,
+      "loss": 2.4518,
+      "step": 11774
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010713099263868146,
+      "loss": 2.7853,
+      "step": 11775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010718703065294724,
+      "loss": 2.6233,
+      "step": 11776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010724307517319394,
+      "loss": 2.6032,
+      "step": 11777
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010729912619090479,
+      "loss": 2.5733,
+      "step": 11778
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010735518369756209,
+      "loss": 2.1688,
+      "step": 11779
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010741124768464706,
+      "loss": 1.958,
+      "step": 11780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010746731814364032,
+      "loss": 2.5725,
+      "step": 11781
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010752339506602112,
+      "loss": 2.1838,
+      "step": 11782
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010757947844326769,
+      "loss": 2.4996,
+      "step": 11783
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010763556826685773,
+      "loss": 2.5499,
+      "step": 11784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010769166452826752,
+      "loss": 1.8938,
+      "step": 11785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010774776721897255,
+      "loss": 1.5818,
+      "step": 11786
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001078038763304471,
+      "loss": 1.7731,
+      "step": 11787
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000107859991854165,
+      "loss": 2.4515,
+      "step": 11788
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010791611378159858,
+      "loss": 2.501,
+      "step": 11789
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010797224210421946,
+      "loss": 3.177,
+      "step": 11790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010802837681349804,
+      "loss": 2.4016,
+      "step": 11791
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001080845179009042,
+      "loss": 2.4905,
+      "step": 11792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010814066535790646,
+      "loss": 2.3421,
+      "step": 11793
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010819681917597249,
+      "loss": 2.5505,
+      "step": 11794
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010825297934656884,
+      "loss": 2.5412,
+      "step": 11795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010830914586116156,
+      "loss": 2.1541,
+      "step": 11796
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010836531871121527,
+      "loss": 2.4607,
+      "step": 11797
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010842149788819379,
+      "loss": 3.0112,
+      "step": 11798
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010847768338355982,
+      "loss": 1.5764,
+      "step": 11799
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010853387518877554,
+      "loss": 2.3968,
+      "step": 11800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010859007329530171,
+      "loss": 2.448,
+      "step": 11801
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010864627769459837,
+      "loss": 2.8079,
+      "step": 11802
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010870248837812436,
+      "loss": 2.6408,
+      "step": 11803
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010875870533733802,
+      "loss": 2.7272,
+      "step": 11804
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010881492856369634,
+      "loss": 2.0268,
+      "step": 11805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010887115804865546,
+      "loss": 2.5659,
+      "step": 11806
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010892739378367049,
+      "loss": 2.2539,
+      "step": 11807
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010898363576019591,
+      "loss": 2.4971,
+      "step": 11808
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001090398839696849,
+      "loss": 2.0922,
+      "step": 11809
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010909613840358972,
+      "loss": 2.4811,
+      "step": 11810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010915239905336202,
+      "loss": 2.2641,
+      "step": 11811
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010920866591045215,
+      "loss": 2.2002,
+      "step": 11812
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010926493896630966,
+      "loss": 2.1355,
+      "step": 11813
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010932121821238298,
+      "loss": 2.5292,
+      "step": 11814
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010937750364012003,
+      "loss": 2.3628,
+      "step": 11815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010943379524096723,
+      "loss": 2.3742,
+      "step": 11816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010949009300637076,
+      "loss": 2.3706,
+      "step": 11817
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010954639692777492,
+      "loss": 2.6014,
+      "step": 11818
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000109602706996624,
+      "loss": 2.1714,
+      "step": 11819
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010965902320436073,
+      "loss": 2.4433,
+      "step": 11820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010971534554242746,
+      "loss": 2.1161,
+      "step": 11821
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010977167400226486,
+      "loss": 2.4195,
+      "step": 11822
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010982800857531341,
+      "loss": 2.3785,
+      "step": 11823
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010988434925301211,
+      "loss": 2.2487,
+      "step": 11824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010994069602679967,
+      "loss": 2.1477,
+      "step": 11825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00010999704888811297,
+      "loss": 2.7977,
+      "step": 11826
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001100534078283888,
+      "loss": 2.7483,
+      "step": 11827
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001101097728390625,
+      "loss": 2.2002,
+      "step": 11828
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001101661439115689,
+      "loss": 2.8452,
+      "step": 11829
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011022252103734157,
+      "loss": 2.8042,
+      "step": 11830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001102789042078133,
+      "loss": 3.0479,
+      "step": 11831
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011033529341441579,
+      "loss": 1.9669,
+      "step": 11832
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011039168864858025,
+      "loss": 2.5321,
+      "step": 11833
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011044808990173656,
+      "loss": 2.4021,
+      "step": 11834
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011050449716531383,
+      "loss": 2.609,
+      "step": 11835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011056091043074011,
+      "loss": 2.3439,
+      "step": 11836
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011061732968944296,
+      "loss": 2.3173,
+      "step": 11837
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011067375493284862,
+      "loss": 2.2245,
+      "step": 11838
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001107301861523824,
+      "loss": 1.6548,
+      "step": 11839
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001107866233394691,
+      "loss": 2.2718,
+      "step": 11840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011084306648553227,
+      "loss": 2.321,
+      "step": 11841
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011089951558199463,
+      "loss": 2.4262,
+      "step": 11842
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001109559706202779,
+      "loss": 2.6056,
+      "step": 11843
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011101243159180323,
+      "loss": 1.1567,
+      "step": 11844
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011106889848799058,
+      "loss": 1.3017,
+      "step": 11845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011112537130025902,
+      "loss": 2.7133,
+      "step": 11846
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011118185002002667,
+      "loss": 2.0422,
+      "step": 11847
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011123833463871116,
+      "loss": 2.4795,
+      "step": 11848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011129482514772862,
+      "loss": 2.2256,
+      "step": 11849
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011135132153849502,
+      "loss": 2.7758,
+      "step": 11850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011140782380242443,
+      "loss": 2.3276,
+      "step": 11851
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011146433193093105,
+      "loss": 2.3073,
+      "step": 11852
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011152084591542747,
+      "loss": 2.627,
+      "step": 11853
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011157736574732603,
+      "loss": 2.5685,
+      "step": 11854
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011163389141803735,
+      "loss": 1.9487,
+      "step": 11855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011169042291897196,
+      "loss": 2.4756,
+      "step": 11856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011174696024153894,
+      "loss": 2.1218,
+      "step": 11857
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011180350337714707,
+      "loss": 2.601,
+      "step": 11858
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011186005231720343,
+      "loss": 2.1263,
+      "step": 11859
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011191660705311504,
+      "loss": 2.3813,
+      "step": 11860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011197316757628745,
+      "loss": 2.0949,
+      "step": 11861
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001120297338781259,
+      "loss": 2.3044,
+      "step": 11862
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011208630595003393,
+      "loss": 2.2901,
+      "step": 11863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011214288378341507,
+      "loss": 1.1917,
+      "step": 11864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011219946736967135,
+      "loss": 2.6368,
+      "step": 11865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011225605670020439,
+      "loss": 2.3524,
+      "step": 11866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001123126517664146,
+      "loss": 2.3437,
+      "step": 11867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011236925255970165,
+      "loss": 0.9636,
+      "step": 11868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011242585907146417,
+      "loss": 1.9538,
+      "step": 11869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011248247129310036,
+      "loss": 2.2011,
+      "step": 11870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001125390892160071,
+      "loss": 2.7768,
+      "step": 11871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011259571283158044,
+      "loss": 2.7735,
+      "step": 11872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011265234213121598,
+      "loss": 2.4465,
+      "step": 11873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011270897710630803,
+      "loss": 1.5679,
+      "step": 11874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011276561774825019,
+      "loss": 2.1459,
+      "step": 11875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011282226404843502,
+      "loss": 2.6418,
+      "step": 11876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00011287891599825468,
+      "loss": 2.0151,
+      "step": 11877
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011293557358910005,
+      "loss": 2.0141,
+      "step": 11878
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011299223681236126,
+      "loss": 2.4109,
+      "step": 11879
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011304890565942749,
+      "loss": 2.4545,
+      "step": 11880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011310558012168743,
+      "loss": 2.3458,
+      "step": 11881
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011316226019052853,
+      "loss": 2.4195,
+      "step": 11882
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011321894585733756,
+      "loss": 2.3575,
+      "step": 11883
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011327563711350025,
+      "loss": 2.4947,
+      "step": 11884
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001133323339504019,
+      "loss": 2.2421,
+      "step": 11885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011338903635942656,
+      "loss": 2.31,
+      "step": 11886
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011344574433195756,
+      "loss": 2.7027,
+      "step": 11887
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011350245785937731,
+      "loss": 2.1688,
+      "step": 11888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011355917693306766,
+      "loss": 2.551,
+      "step": 11889
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011361590154440933,
+      "loss": 2.4053,
+      "step": 11890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011367263168478229,
+      "loss": 2.4557,
+      "step": 11891
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001137293673455655,
+      "loss": 2.5821,
+      "step": 11892
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011378610851813753,
+      "loss": 2.6138,
+      "step": 11893
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011384285519387572,
+      "loss": 2.4644,
+      "step": 11894
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011389960736415665,
+      "loss": 2.5546,
+      "step": 11895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011395636502035597,
+      "loss": 1.6981,
+      "step": 11896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011401312815384891,
+      "loss": 2.9622,
+      "step": 11897
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011406989675600942,
+      "loss": 2.6414,
+      "step": 11898
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001141266708182107,
+      "loss": 2.3505,
+      "step": 11899
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001141834503318254,
+      "loss": 2.5555,
+      "step": 11900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011424023528822506,
+      "loss": 2.5006,
+      "step": 11901
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011429702567878034,
+      "loss": 2.6739,
+      "step": 11902
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011435382149486144,
+      "loss": 2.1115,
+      "step": 11903
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001144106227278374,
+      "loss": 2.7069,
+      "step": 11904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011446742936907643,
+      "loss": 2.3477,
+      "step": 11905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011452424140994626,
+      "loss": 2.4763,
+      "step": 11906
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011458105884181344,
+      "loss": 2.3734,
+      "step": 11907
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011463788165604383,
+      "loss": 2.5703,
+      "step": 11908
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011469470984400238,
+      "loss": 2.3688,
+      "step": 11909
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011475154339705356,
+      "loss": 1.8792,
+      "step": 11910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011480838230656062,
+      "loss": 2.5975,
+      "step": 11911
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001148652265638862,
+      "loss": 2.3734,
+      "step": 11912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011492207616039193,
+      "loss": 2.1308,
+      "step": 11913
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011497893108743905,
+      "loss": 2.3756,
+      "step": 11914
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011503579133638758,
+      "loss": 2.2965,
+      "step": 11915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011509265689859691,
+      "loss": 2.2384,
+      "step": 11916
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001151495277654254,
+      "loss": 2.529,
+      "step": 11917
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011520640392823111,
+      "loss": 2.3817,
+      "step": 11918
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011526328537837081,
+      "loss": 2.5598,
+      "step": 11919
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011532017210720066,
+      "loss": 2.1267,
+      "step": 11920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011537706410607581,
+      "loss": 2.6379,
+      "step": 11921
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011543396136635111,
+      "loss": 2.5677,
+      "step": 11922
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011549086387938015,
+      "loss": 2.0904,
+      "step": 11923
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011554777163651582,
+      "loss": 2.6148,
+      "step": 11924
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011560468462911014,
+      "loss": 2.569,
+      "step": 11925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011566160284851475,
+      "loss": 2.3021,
+      "step": 11926
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011571852628608004,
+      "loss": 2.5835,
+      "step": 11927
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011577545493315573,
+      "loss": 2.5773,
+      "step": 11928
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001158323887810907,
+      "loss": 2.2972,
+      "step": 11929
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011588932782123334,
+      "loss": 2.2718,
+      "step": 11930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011594627204493094,
+      "loss": 2.2255,
+      "step": 11931
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011600322144352994,
+      "loss": 2.4506,
+      "step": 11932
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011606017600837641,
+      "loss": 2.3807,
+      "step": 11933
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011611713573081523,
+      "loss": 2.2208,
+      "step": 11934
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011617410060219066,
+      "loss": 2.6437,
+      "step": 11935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011623107061384602,
+      "loss": 2.5301,
+      "step": 11936
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011628804575712423,
+      "loss": 2.5629,
+      "step": 11937
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011634502602336694,
+      "loss": 2.072,
+      "step": 11938
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011640201140391567,
+      "loss": 0.9554,
+      "step": 11939
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011645900189011021,
+      "loss": 2.3134,
+      "step": 11940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011651599747329051,
+      "loss": 2.9106,
+      "step": 11941
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011657299814479514,
+      "loss": 1.2729,
+      "step": 11942
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011663000389596246,
+      "loss": 2.7619,
+      "step": 11943
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011668701471812924,
+      "loss": 2.968,
+      "step": 11944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011674403060263229,
+      "loss": 2.2651,
+      "step": 11945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001168010515408071,
+      "loss": 2.7994,
+      "step": 11946
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011685807752398896,
+      "loss": 2.1795,
+      "step": 11947
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011691510854351156,
+      "loss": 2.6039,
+      "step": 11948
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011697214459070863,
+      "loss": 3.0341,
+      "step": 11949
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011702918565691261,
+      "loss": 2.1898,
+      "step": 11950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011708623173345574,
+      "loss": 2.3907,
+      "step": 11951
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011714328281166863,
+      "loss": 1.3076,
+      "step": 11952
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011720033888288203,
+      "loss": 2.2589,
+      "step": 11953
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011725739993842524,
+      "loss": 2.5828,
+      "step": 11954
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011731446596962738,
+      "loss": 2.5494,
+      "step": 11955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011737153696781641,
+      "loss": 2.4891,
+      "step": 11956
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011742861292431967,
+      "loss": 2.1291,
+      "step": 11957
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011748569383046359,
+      "loss": 1.7921,
+      "step": 11958
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011754277967757423,
+      "loss": 2.2478,
+      "step": 11959
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011759987045697658,
+      "loss": 1.3671,
+      "step": 11960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011765696615999478,
+      "loss": 2.0355,
+      "step": 11961
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001177140667779527,
+      "loss": 2.4742,
+      "step": 11962
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011777117230217304,
+      "loss": 1.9978,
+      "step": 11963
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011782828272397788,
+      "loss": 2.7567,
+      "step": 11964
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011788539803468841,
+      "loss": 2.8819,
+      "step": 11965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011794251822562549,
+      "loss": 2.5634,
+      "step": 11966
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011799964328810884,
+      "loss": 2.1885,
+      "step": 11967
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011805677321345757,
+      "loss": 2.4188,
+      "step": 11968
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011811390799298996,
+      "loss": 2.183,
+      "step": 11969
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011817104761802387,
+      "loss": 1.849,
+      "step": 11970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011822819207987606,
+      "loss": 2.4089,
+      "step": 11971
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011828534136986273,
+      "loss": 1.5835,
+      "step": 11972
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011834249547929916,
+      "loss": 2.4169,
+      "step": 11973
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011839965439950032,
+      "loss": 2.8687,
+      "step": 11974
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001184568181217799,
+      "loss": 2.5127,
+      "step": 11975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011851398663745153,
+      "loss": 2.7468,
+      "step": 11976
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011857115993782722,
+      "loss": 2.2893,
+      "step": 11977
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011862833801421913,
+      "loss": 2.2825,
+      "step": 11978
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011868552085793806,
+      "loss": 2.6285,
+      "step": 11979
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011874270846029472,
+      "loss": 2.8969,
+      "step": 11980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011879990081259822,
+      "loss": 2.6556,
+      "step": 11981
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011885709790615782,
+      "loss": 2.1795,
+      "step": 11982
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011891429973228144,
+      "loss": 2.2192,
+      "step": 11983
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011897150628227688,
+      "loss": 2.4959,
+      "step": 11984
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011902871754745041,
+      "loss": 2.5681,
+      "step": 11985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011908593351910839,
+      "loss": 1.2737,
+      "step": 11986
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011914315418855587,
+      "loss": 2.7112,
+      "step": 11987
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011920037954709769,
+      "loss": 2.6527,
+      "step": 11988
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001192576095860376,
+      "loss": 2.4096,
+      "step": 11989
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011931484429667876,
+      "loss": 1.6393,
+      "step": 11990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011937208367032349,
+      "loss": 2.6737,
+      "step": 11991
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011942932769827377,
+      "loss": 2.4373,
+      "step": 11992
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011948657637183053,
+      "loss": 2.7192,
+      "step": 11993
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011954382968229395,
+      "loss": 2.4938,
+      "step": 11994
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011960108762096393,
+      "loss": 2.4043,
+      "step": 11995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011965835017913928,
+      "loss": 2.007,
+      "step": 11996
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011971561734811816,
+      "loss": 2.6066,
+      "step": 11997
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000119772889119198,
+      "loss": 2.5241,
+      "step": 11998
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011983016548367587,
+      "loss": 2.5049,
+      "step": 11999
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00011988744643284776,
+      "loss": 2.4621,
+      "step": 12000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001199447319580091,
+      "loss": 2.094,
+      "step": 12001
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012000202205045449,
+      "loss": 3.1398,
+      "step": 12002
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012005931670147819,
+      "loss": 2.0134,
+      "step": 12003
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012011661590237345,
+      "loss": 2.5826,
+      "step": 12004
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012017391964443292,
+      "loss": 2.6062,
+      "step": 12005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012023122791894841,
+      "loss": 2.3133,
+      "step": 12006
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012028854071721145,
+      "loss": 2.5067,
+      "step": 12007
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012034585803051253,
+      "loss": 2.4979,
+      "step": 12008
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001204031798501415,
+      "loss": 2.4709,
+      "step": 12009
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012046050616738747,
+      "loss": 2.8123,
+      "step": 12010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012051783697353921,
+      "loss": 2.3573,
+      "step": 12011
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012057517225988447,
+      "loss": 2.3548,
+      "step": 12012
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012063251201771038,
+      "loss": 2.275,
+      "step": 12013
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012068985623830335,
+      "loss": 2.5012,
+      "step": 12014
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012074720491294938,
+      "loss": 1.8623,
+      "step": 12015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012080455803293349,
+      "loss": 2.3765,
+      "step": 12016
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012086191558954019,
+      "loss": 2.6951,
+      "step": 12017
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012091927757405307,
+      "loss": 2.9668,
+      "step": 12018
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012097664397775554,
+      "loss": 2.6183,
+      "step": 12019
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012103401479192987,
+      "loss": 2.2929,
+      "step": 12020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012109139000785774,
+      "loss": 2.1849,
+      "step": 12021
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012114876961682048,
+      "loss": 2.7818,
+      "step": 12022
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001212061536100984,
+      "loss": 1.0892,
+      "step": 12023
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012126354197897125,
+      "loss": 2.6581,
+      "step": 12024
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012132093471471803,
+      "loss": 2.009,
+      "step": 12025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012137833180861741,
+      "loss": 1.7516,
+      "step": 12026
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012143573325194692,
+      "loss": 2.4023,
+      "step": 12027
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012149313903598404,
+      "loss": 2.2676,
+      "step": 12028
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012155054915200474,
+      "loss": 2.6052,
+      "step": 12029
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012160796359128519,
+      "loss": 2.8348,
+      "step": 12030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012166538234510026,
+      "loss": 2.6127,
+      "step": 12031
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012172280540472482,
+      "loss": 2.3319,
+      "step": 12032
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012178023276143217,
+      "loss": 2.2659,
+      "step": 12033
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001218376644064959,
+      "loss": 2.18,
+      "step": 12034
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012189510033118827,
+      "loss": 2.05,
+      "step": 12035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012195254052678155,
+      "loss": 2.3852,
+      "step": 12036
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012200998498454642,
+      "loss": 2.4245,
+      "step": 12037
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012206743369575385,
+      "loss": 2.7638,
+      "step": 12038
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001221248866516735,
+      "loss": 1.9689,
+      "step": 12039
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012218234384357512,
+      "loss": 2.7119,
+      "step": 12040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012223980526272673,
+      "loss": 1.9793,
+      "step": 12041
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012229727090039676,
+      "loss": 2.5072,
+      "step": 12042
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012235474074785233,
+      "loss": 2.7758,
+      "step": 12043
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012241221479636051,
+      "loss": 2.4177,
+      "step": 12044
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012246969303718688,
+      "loss": 2.3973,
+      "step": 12045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001225271754615973,
+      "loss": 2.3127,
+      "step": 12046
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012258466206085627,
+      "loss": 2.212,
+      "step": 12047
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012264215282622824,
+      "loss": 2.4159,
+      "step": 12048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012269964774897662,
+      "loss": 2.6609,
+      "step": 12049
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012275714682036433,
+      "loss": 2.3723,
+      "step": 12050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012281465003165348,
+      "loss": 2.7919,
+      "step": 12051
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012287215737410604,
+      "loss": 2.6405,
+      "step": 12052
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012292966883898284,
+      "loss": 2.4425,
+      "step": 12053
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012298718441754416,
+      "loss": 2.6249,
+      "step": 12054
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012304470410105006,
+      "loss": 2.6565,
+      "step": 12055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012310222788075952,
+      "loss": 1.9463,
+      "step": 12056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012315975574793108,
+      "loss": 2.9332,
+      "step": 12057
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001232172876938225,
+      "loss": 2.8944,
+      "step": 12058
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012327482370969134,
+      "loss": 1.4977,
+      "step": 12059
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000123332363786794,
+      "loss": 2.6488,
+      "step": 12060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012338990791638685,
+      "loss": 1.8743,
+      "step": 12061
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001234474560897249,
+      "loss": 2.0251,
+      "step": 12062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012350500829806332,
+      "loss": 2.5186,
+      "step": 12063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012356256453265602,
+      "loss": 2.646,
+      "step": 12064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012362012478475703,
+      "loss": 2.3114,
+      "step": 12065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012367768904561876,
+      "loss": 2.3364,
+      "step": 12066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012373525730649402,
+      "loss": 2.3076,
+      "step": 12067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012379282955863428,
+      "loss": 2.551,
+      "step": 12068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001238504057932911,
+      "loss": 2.8703,
+      "step": 12069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001239079860017145,
+      "loss": 2.1574,
+      "step": 12070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001239655701751548,
+      "loss": 2.5517,
+      "step": 12071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012402315830486115,
+      "loss": 2.4026,
+      "step": 12072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001240807503820826,
+      "loss": 2.5265,
+      "step": 12073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012413834639806687,
+      "loss": 2.1214,
+      "step": 12074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001241959463440618,
+      "loss": 1.6316,
+      "step": 12075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012425355021131415,
+      "loss": 2.2743,
+      "step": 12076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001243111579910705,
+      "loss": 2.6152,
+      "step": 12077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012436876967457646,
+      "loss": 2.0019,
+      "step": 12078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012442638525307723,
+      "loss": 2.0761,
+      "step": 12079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012448400471781724,
+      "loss": 2.4362,
+      "step": 12080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001245416280600407,
+      "loss": 2.1046,
+      "step": 12081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012459925527099093,
+      "loss": 2.7081,
+      "step": 12082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012465688634191057,
+      "loss": 2.0467,
+      "step": 12083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001247145212640421,
+      "loss": 2.6356,
+      "step": 12084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012477216002862703,
+      "loss": 2.8978,
+      "step": 12085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012482980262690642,
+      "loss": 2.379,
+      "step": 12086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001248874490501206,
+      "loss": 2.1411,
+      "step": 12087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012494509928950967,
+      "loss": 2.4196,
+      "step": 12088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012500275333631288,
+      "loss": 2.3515,
+      "step": 12089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012506041118176892,
+      "loss": 2.209,
+      "step": 12090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012511807281711577,
+      "loss": 2.854,
+      "step": 12091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012517573823359133,
+      "loss": 2.5228,
+      "step": 12092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012523340742243247,
+      "loss": 2.7461,
+      "step": 12093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012529108037487554,
+      "loss": 2.3442,
+      "step": 12094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012534875708215633,
+      "loss": 1.959,
+      "step": 12095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012540643753551037,
+      "loss": 2.584,
+      "step": 12096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012546412172617226,
+      "loss": 2.7521,
+      "step": 12097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001255218096453761,
+      "loss": 2.2855,
+      "step": 12098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012557950128435537,
+      "loss": 2.3339,
+      "step": 12099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012563719663434333,
+      "loss": 2.4106,
+      "step": 12100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012569489568657233,
+      "loss": 2.6432,
+      "step": 12101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012575259843227425,
+      "loss": 1.9332,
+      "step": 12102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012581030486268023,
+      "loss": 2.687,
+      "step": 12103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012586801496902137,
+      "loss": 2.5341,
+      "step": 12104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001259257287425277,
+      "loss": 2.278,
+      "step": 12105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012598344617442892,
+      "loss": 2.7536,
+      "step": 12106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012604116725595393,
+      "loss": 2.1279,
+      "step": 12107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012609889197833158,
+      "loss": 2.7342,
+      "step": 12108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012615662033278968,
+      "loss": 2.574,
+      "step": 12109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001262143523105556,
+      "loss": 2.7049,
+      "step": 12110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012627208790285643,
+      "loss": 2.5702,
+      "step": 12111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001263298271009184,
+      "loss": 1.8643,
+      "step": 12112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001263875698959673,
+      "loss": 2.2414,
+      "step": 12113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012644531627922822,
+      "loss": 1.7114,
+      "step": 12114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012650306624192612,
+      "loss": 2.6867,
+      "step": 12115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001265608197752849,
+      "loss": 2.5312,
+      "step": 12116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012661857687052856,
+      "loss": 1.9172,
+      "step": 12117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012667633751887964,
+      "loss": 2.3394,
+      "step": 12118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012673410171156106,
+      "loss": 2.2338,
+      "step": 12119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012679186943979456,
+      "loss": 2.2533,
+      "step": 12120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012684964069480182,
+      "loss": 2.4778,
+      "step": 12121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012690741546780366,
+      "loss": 2.7639,
+      "step": 12122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012696519375002044,
+      "loss": 2.6183,
+      "step": 12123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001270229755326719,
+      "loss": 2.3303,
+      "step": 12124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012708076080697757,
+      "loss": 2.4006,
+      "step": 12125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012713854956415615,
+      "loss": 2.6406,
+      "step": 12126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012719634179542586,
+      "loss": 2.9329,
+      "step": 12127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012725413749200437,
+      "loss": 2.5281,
+      "step": 12128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012731193664510904,
+      "loss": 2.7875,
+      "step": 12129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012736973924595648,
+      "loss": 2.3327,
+      "step": 12130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012742754528576283,
+      "loss": 1.9948,
+      "step": 12131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001274853547557436,
+      "loss": 2.6533,
+      "step": 12132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001275431676471141,
+      "loss": 2.5926,
+      "step": 12133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012760098395108889,
+      "loss": 2.8825,
+      "step": 12134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012765880365888196,
+      "loss": 2.879,
+      "step": 12135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012771662676170674,
+      "loss": 2.5286,
+      "step": 12136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012777445325077655,
+      "loss": 1.9976,
+      "step": 12137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012783228311730378,
+      "loss": 2.4434,
+      "step": 12138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012789011635250044,
+      "loss": 2.2149,
+      "step": 12139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001279479529475779,
+      "loss": 2.4318,
+      "step": 12140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012800579289374737,
+      "loss": 2.083,
+      "step": 12141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012806363618221922,
+      "loss": 2.1283,
+      "step": 12142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001281214828042033,
+      "loss": 2.3644,
+      "step": 12143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012817933275090934,
+      "loss": 2.5431,
+      "step": 12144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012823718601354613,
+      "loss": 2.2513,
+      "step": 12145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012829504258332215,
+      "loss": 2.6014,
+      "step": 12146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012835290245144523,
+      "loss": 2.5109,
+      "step": 12147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012841076560912305,
+      "loss": 2.2812,
+      "step": 12148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001284686320475623,
+      "loss": 1.9506,
+      "step": 12149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012852650175796987,
+      "loss": 2.2734,
+      "step": 12150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012858437473155116,
+      "loss": 2.3718,
+      "step": 12151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000128642250959512,
+      "loss": 2.4341,
+      "step": 12152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012870013043305713,
+      "loss": 2.1423,
+      "step": 12153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012875801314339138,
+      "loss": 2.751,
+      "step": 12154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001288158990817182,
+      "loss": 2.3386,
+      "step": 12155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001288737882392415,
+      "loss": 2.624,
+      "step": 12156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012893168060716404,
+      "loss": 2.272,
+      "step": 12157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012898957617668867,
+      "loss": 2.4378,
+      "step": 12158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012904747493901688,
+      "loss": 2.6141,
+      "step": 12159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012910537688535066,
+      "loss": 2.0535,
+      "step": 12160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012916328200689075,
+      "loss": 2.2281,
+      "step": 12161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012922119029483818,
+      "loss": 2.3116,
+      "step": 12162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012927910174039247,
+      "loss": 2.4461,
+      "step": 12163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001293370163347536,
+      "loss": 2.3725,
+      "step": 12164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012939493406912051,
+      "loss": 2.1563,
+      "step": 12165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012945285493469222,
+      "loss": 2.4467,
+      "step": 12166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012951077892266636,
+      "loss": 2.1037,
+      "step": 12167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012956870602424104,
+      "loss": 2.7176,
+      "step": 12168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012962663623061323,
+      "loss": 2.8282,
+      "step": 12169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012968456953298,
+      "loss": 1.7071,
+      "step": 12170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012974250592253744,
+      "loss": 2.7465,
+      "step": 12171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012980044539048142,
+      "loss": 2.2387,
+      "step": 12172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012985838792800713,
+      "loss": 2.4135,
+      "step": 12173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012991633352630975,
+      "loss": 2.3008,
+      "step": 12174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00012997428217658354,
+      "loss": 2.3227,
+      "step": 12175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013003223387002234,
+      "loss": 2.7107,
+      "step": 12176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013009018859781992,
+      "loss": 2.4571,
+      "step": 12177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013014814635116918,
+      "loss": 2.7188,
+      "step": 12178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013020610712126266,
+      "loss": 2.5586,
+      "step": 12179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001302640708992924,
+      "loss": 2.7344,
+      "step": 12180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013032203767645027,
+      "loss": 2.3624,
+      "step": 12181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001303800074439274,
+      "loss": 2.2941,
+      "step": 12182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013043798019291445,
+      "loss": 2.2165,
+      "step": 12183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013049595591460168,
+      "loss": 2.4223,
+      "step": 12184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013055393460017917,
+      "loss": 2.3326,
+      "step": 12185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013061191624083615,
+      "loss": 2.8329,
+      "step": 12186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013066990082776157,
+      "loss": 1.9451,
+      "step": 12187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001307278883521438,
+      "loss": 3.032,
+      "step": 12188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001307858788051712,
+      "loss": 2.6493,
+      "step": 12189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013084387217803117,
+      "loss": 2.3407,
+      "step": 12190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013090186846191092,
+      "loss": 2.5568,
+      "step": 12191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013095986764799703,
+      "loss": 2.6034,
+      "step": 12192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013101786972747603,
+      "loss": 3.0109,
+      "step": 12193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001310758746915335,
+      "loss": 2.2658,
+      "step": 12194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013113388253135528,
+      "loss": 2.7733,
+      "step": 12195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013119189323812573,
+      "loss": 2.4402,
+      "step": 12196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001312499068030298,
+      "loss": 2.404,
+      "step": 12197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001313079232172513,
+      "loss": 2.5587,
+      "step": 12198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013136594247197426,
+      "loss": 2.2574,
+      "step": 12199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013142396455838166,
+      "loss": 2.3564,
+      "step": 12200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013148198946765633,
+      "loss": 2.7946,
+      "step": 12201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013154001719098052,
+      "loss": 2.098,
+      "step": 12202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013159804771953643,
+      "loss": 2.478,
+      "step": 12203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013165608104450543,
+      "loss": 2.1924,
+      "step": 12204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001317141171570685,
+      "loss": 2.0364,
+      "step": 12205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013177215604840656,
+      "loss": 1.8917,
+      "step": 12206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013183019770969973,
+      "loss": 2.1579,
+      "step": 12207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013188824213212786,
+      "loss": 3.0081,
+      "step": 12208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013194628930687015,
+      "loss": 2.4011,
+      "step": 12209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001320043392251059,
+      "loss": 2.328,
+      "step": 12210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013206239187801356,
+      "loss": 2.1608,
+      "step": 12211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013212044725677125,
+      "loss": 2.4955,
+      "step": 12212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013217850535255656,
+      "loss": 2.2548,
+      "step": 12213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013223656615654717,
+      "loss": 2.7939,
+      "step": 12214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013229462965991975,
+      "loss": 2.6063,
+      "step": 12215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013235269585385088,
+      "loss": 2.2066,
+      "step": 12216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001324107647295165,
+      "loss": 2.8535,
+      "step": 12217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013246883627809254,
+      "loss": 2.4152,
+      "step": 12218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013252691049075421,
+      "loss": 2.456,
+      "step": 12219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013258498735867634,
+      "loss": 2.5952,
+      "step": 12220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013264306687303326,
+      "loss": 2.5882,
+      "step": 12221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013270114902499933,
+      "loss": 2.6217,
+      "step": 12222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001327592338057481,
+      "loss": 2.4762,
+      "step": 12223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013281732120645285,
+      "loss": 2.6084,
+      "step": 12224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013287541121828627,
+      "loss": 2.3777,
+      "step": 12225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013293350383242113,
+      "loss": 1.8549,
+      "step": 12226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001329915990400294,
+      "loss": 2.4608,
+      "step": 12227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001330496968322827,
+      "loss": 2.1804,
+      "step": 12228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013310779720035227,
+      "loss": 2.8856,
+      "step": 12229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013316590013540922,
+      "loss": 2.7248,
+      "step": 12230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013322400562862396,
+      "loss": 2.1003,
+      "step": 12231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001332821136711665,
+      "loss": 2.2816,
+      "step": 12232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013334022425420678,
+      "loss": 2.8597,
+      "step": 12233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013339833736891407,
+      "loss": 2.3894,
+      "step": 12234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013345645300645733,
+      "loss": 2.3294,
+      "step": 12235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013351457115800498,
+      "loss": 2.254,
+      "step": 12236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001335726918147255,
+      "loss": 1.8911,
+      "step": 12237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013363081496778646,
+      "loss": 2.3313,
+      "step": 12238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001336889406083557,
+      "loss": 1.698,
+      "step": 12239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001337470687275997,
+      "loss": 2.4592,
+      "step": 12240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001338051993166856,
+      "loss": 2.3739,
+      "step": 12241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013386333236677943,
+      "loss": 2.2588,
+      "step": 12242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013392146786904752,
+      "loss": 2.2024,
+      "step": 12243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001339796058146549,
+      "loss": 2.7211,
+      "step": 12244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013403774619476715,
+      "loss": 1.5064,
+      "step": 12245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013409588900054886,
+      "loss": 2.4949,
+      "step": 12246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013415403422316482,
+      "loss": 2.1533,
+      "step": 12247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013421218185377868,
+      "loss": 2.4869,
+      "step": 12248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013427033188355447,
+      "loss": 2.5173,
+      "step": 12249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001343284843036553,
+      "loss": 2.4096,
+      "step": 12250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013438663910524462,
+      "loss": 2.5754,
+      "step": 12251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013444479627948446,
+      "loss": 2.4028,
+      "step": 12252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013450295581753753,
+      "loss": 1.6815,
+      "step": 12253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013456111771056544,
+      "loss": 2.0894,
+      "step": 12254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013461928194973016,
+      "loss": 2.0661,
+      "step": 12255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013467744852619235,
+      "loss": 2.6692,
+      "step": 12256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001347356174311133,
+      "loss": 2.3694,
+      "step": 12257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013479378865565312,
+      "loss": 2.4631,
+      "step": 12258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013485196219097234,
+      "loss": 1.8944,
+      "step": 12259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013491013802823052,
+      "loss": 2.2164,
+      "step": 12260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013496831615858712,
+      "loss": 2.9231,
+      "step": 12261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001350264965732011,
+      "loss": 2.2858,
+      "step": 12262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013508467926323146,
+      "loss": 3.1387,
+      "step": 12263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013514286421983644,
+      "loss": 2.0192,
+      "step": 12264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000135201051434174,
+      "loss": 2.2572,
+      "step": 12265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013525924089740208,
+      "loss": 2.4529,
+      "step": 12266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013531743260067791,
+      "loss": 2.1205,
+      "step": 12267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013537562653515855,
+      "loss": 2.2221,
+      "step": 12268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013543382269200048,
+      "loss": 2.0833,
+      "step": 12269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013549202106236035,
+      "loss": 2.2821,
+      "step": 12270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013555022163739395,
+      "loss": 2.3818,
+      "step": 12271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013560842440825727,
+      "loss": 2.3445,
+      "step": 12272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013566662936610518,
+      "loss": 2.109,
+      "step": 12273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013572483650209306,
+      "loss": 2.1599,
+      "step": 12274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001357830458073753,
+      "loss": 2.2447,
+      "step": 12275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001358412572731067,
+      "loss": 2.7457,
+      "step": 12276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013589947089044072,
+      "loss": 2.4233,
+      "step": 12277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013595768665053146,
+      "loss": 2.644,
+      "step": 12278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013601590454453201,
+      "loss": 2.1366,
+      "step": 12279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013607412456359585,
+      "loss": 2.4039,
+      "step": 12280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013613234669887508,
+      "loss": 2.4332,
+      "step": 12281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013619057094152258,
+      "loss": 2.7458,
+      "step": 12282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013624879728269015,
+      "loss": 2.3876,
+      "step": 12283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013630702571352992,
+      "loss": 2.1596,
+      "step": 12284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001363652562251928,
+      "loss": 2.0731,
+      "step": 12285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013642348880883037,
+      "loss": 2.2216,
+      "step": 12286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001364817234555931,
+      "loss": 2.8126,
+      "step": 12287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013653996015663178,
+      "loss": 2.6764,
+      "step": 12288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001365981989030965,
+      "loss": 2.9051,
+      "step": 12289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013665643968613708,
+      "loss": 2.3878,
+      "step": 12290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013671468249690298,
+      "loss": 2.2447,
+      "step": 12291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013677292732654372,
+      "loss": 2.6679,
+      "step": 12292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013683117416620809,
+      "loss": 2.1697,
+      "step": 12293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013688942300704477,
+      "loss": 2.382,
+      "step": 12294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013694767384020195,
+      "loss": 2.6217,
+      "step": 12295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013700592665682795,
+      "loss": 1.8987,
+      "step": 12296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013706418144807036,
+      "loss": 2.7108,
+      "step": 12297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013712243820507647,
+      "loss": 2.3123,
+      "step": 12298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013718069691899372,
+      "loss": 2.7027,
+      "step": 12299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013723895758096879,
+      "loss": 2.372,
+      "step": 12300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001372972201821482,
+      "loss": 2.2305,
+      "step": 12301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013735548471367813,
+      "loss": 2.4667,
+      "step": 12302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013741375116670476,
+      "loss": 2.1014,
+      "step": 12303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001374720195323736,
+      "loss": 2.758,
+      "step": 12304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013753028980183007,
+      "loss": 2.5503,
+      "step": 12305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001375885619662191,
+      "loss": 2.4056,
+      "step": 12306
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013764683601668573,
+      "loss": 2.5204,
+      "step": 12307
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013770511194437438,
+      "loss": 2.4747,
+      "step": 12308
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013776338974042925,
+      "loss": 2.3766,
+      "step": 12309
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013782166939599417,
+      "loss": 2.6535,
+      "step": 12310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000137879950902213,
+      "loss": 1.6599,
+      "step": 12311
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013793823425022904,
+      "loss": 3.0532,
+      "step": 12312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013799651943118539,
+      "loss": 2.3996,
+      "step": 12313
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001380548064362247,
+      "loss": 1.9023,
+      "step": 12314
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001381130952564898,
+      "loss": 2.7641,
+      "step": 12315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013817138588312282,
+      "loss": 2.3595,
+      "step": 12316
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013822967830726574,
+      "loss": 2.1304,
+      "step": 12317
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013828797252006016,
+      "loss": 2.8257,
+      "step": 12318
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013834626851264776,
+      "loss": 2.5925,
+      "step": 12319
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001384045662761696,
+      "loss": 2.2505,
+      "step": 12320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001384628658017665,
+      "loss": 2.5072,
+      "step": 12321
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013852116708057928,
+      "loss": 2.4195,
+      "step": 12322
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013857947010374825,
+      "loss": 2.1599,
+      "step": 12323
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013863777486241346,
+      "loss": 1.6486,
+      "step": 12324
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013869608134771464,
+      "loss": 2.3454,
+      "step": 12325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013875438955079166,
+      "loss": 2.7743,
+      "step": 12326
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013881269946278355,
+      "loss": 1.9712,
+      "step": 12327
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013887101107482977,
+      "loss": 1.9521,
+      "step": 12328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013892932437806858,
+      "loss": 2.2967,
+      "step": 12329
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013898763936363897,
+      "loss": 2.5321,
+      "step": 12330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001390459560226789,
+      "loss": 2.5258,
+      "step": 12331
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013910427434632683,
+      "loss": 2.3797,
+      "step": 12332
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013916259432572,
+      "loss": 2.2571,
+      "step": 12333
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001392209159519964,
+      "loss": 2.5326,
+      "step": 12334
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013927923921629295,
+      "loss": 2.391,
+      "step": 12335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013933756410974715,
+      "loss": 2.228,
+      "step": 12336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001393958906234952,
+      "loss": 2.5506,
+      "step": 12337
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013945421874867405,
+      "loss": 2.3637,
+      "step": 12338
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013951254847641977,
+      "loss": 2.7243,
+      "step": 12339
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001395708797978686,
+      "loss": 2.185,
+      "step": 12340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001396292127041563,
+      "loss": 2.8964,
+      "step": 12341
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013968754718641835,
+      "loss": 2.4733,
+      "step": 12342
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013974588323579002,
+      "loss": 2.3323,
+      "step": 12343
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001398042208434066,
+      "loss": 2.4479,
+      "step": 12344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013986256000040283,
+      "loss": 2.2189,
+      "step": 12345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001399209006979133,
+      "loss": 2.6753,
+      "step": 12346
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00013997924292707232,
+      "loss": 2.5348,
+      "step": 12347
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001400375866790143,
+      "loss": 2.5832,
+      "step": 12348
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014009593194487296,
+      "loss": 2.6076,
+      "step": 12349
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000140154278715782,
+      "loss": 2.2924,
+      "step": 12350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014021262698287482,
+      "loss": 2.2231,
+      "step": 12351
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014027097673728484,
+      "loss": 2.2591,
+      "step": 12352
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014032932797014496,
+      "loss": 1.1683,
+      "step": 12353
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001403876806725878,
+      "loss": 2.5686,
+      "step": 12354
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014044603483574623,
+      "loss": 2.6765,
+      "step": 12355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014050439045075244,
+      "loss": 2.4574,
+      "step": 12356
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014056274750873853,
+      "loss": 2.598,
+      "step": 12357
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014062110600083626,
+      "loss": 2.3284,
+      "step": 12358
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014067946591817757,
+      "loss": 2.1394,
+      "step": 12359
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001407378272518937,
+      "loss": 2.1966,
+      "step": 12360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001407961899931162,
+      "loss": 2.5244,
+      "step": 12361
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001408545541329757,
+      "loss": 2.6045,
+      "step": 12362
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014091291966260327,
+      "loss": 2.2226,
+      "step": 12363
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014097128657312937,
+      "loss": 1.683,
+      "step": 12364
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014102965485568474,
+      "loss": 2.5888,
+      "step": 12365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014108802450139908,
+      "loss": 2.2859,
+      "step": 12366
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014114639550140274,
+      "loss": 2.1676,
+      "step": 12367
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001412047678468252,
+      "loss": 1.8,
+      "step": 12368
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014126314152879647,
+      "loss": 2.0544,
+      "step": 12369
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014132151653844542,
+      "loss": 2.6122,
+      "step": 12370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014137989286690155,
+      "loss": 2.305,
+      "step": 12371
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001414382705052936,
+      "loss": 2.3503,
+      "step": 12372
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014149664944475073,
+      "loss": 1.8664,
+      "step": 12373
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014155502967640095,
+      "loss": 2.5606,
+      "step": 12374
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001416134111913731,
+      "loss": 1.9962,
+      "step": 12375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014167179398079502,
+      "loss": 3.1054,
+      "step": 12376
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014173017803579513,
+      "loss": 2.6195,
+      "step": 12377
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014178856334750065,
+      "loss": 2.2429,
+      "step": 12378
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014184694990703966,
+      "loss": 2.354,
+      "step": 12379
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014190533770553924,
+      "loss": 2.4496,
+      "step": 12380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014196372673412692,
+      "loss": 2.9513,
+      "step": 12381
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014202211698392955,
+      "loss": 3.3188,
+      "step": 12382
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014208050844607406,
+      "loss": 2.2744,
+      "step": 12383
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001421389011116869,
+      "loss": 2.0284,
+      "step": 12384
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014219729497189486,
+      "loss": 2.1949,
+      "step": 12385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014225569001782414,
+      "loss": 1.8905,
+      "step": 12386
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001423140862406007,
+      "loss": 2.5897,
+      "step": 12387
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014237248363135072,
+      "loss": 2.0835,
+      "step": 12388
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014243088218119987,
+      "loss": 2.0979,
+      "step": 12389
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014248928188127373,
+      "loss": 2.7708,
+      "step": 12390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001425476827226976,
+      "loss": 2.2744,
+      "step": 12391
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014260608469659694,
+      "loss": 2.4428,
+      "step": 12392
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014266448779409676,
+      "loss": 2.8775,
+      "step": 12393
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014272289200632188,
+      "loss": 2.1877,
+      "step": 12394
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014278129732439696,
+      "loss": 2.5886,
+      "step": 12395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014283970373944678,
+      "loss": 2.7377,
+      "step": 12396
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014289811124259562,
+      "loss": 2.4101,
+      "step": 12397
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014295651982496773,
+      "loss": 2.1603,
+      "step": 12398
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014301492947768702,
+      "loss": 2.1053,
+      "step": 12399
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014307334019187764,
+      "loss": 2.335,
+      "step": 12400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014313175195866326,
+      "loss": 1.6756,
+      "step": 12401
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001431901647691674,
+      "loss": 2.1416,
+      "step": 12402
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014324857861451338,
+      "loss": 2.3564,
+      "step": 12403
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014330699348582472,
+      "loss": 2.192,
+      "step": 12404
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014336540937422441,
+      "loss": 1.6654,
+      "step": 12405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001434238262708354,
+      "loss": 2.0802,
+      "step": 12406
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014348224416678033,
+      "loss": 2.2583,
+      "step": 12407
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014354066305318215,
+      "loss": 2.4092,
+      "step": 12408
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014359908292116324,
+      "loss": 2.2091,
+      "step": 12409
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014365750376184577,
+      "loss": 2.3903,
+      "step": 12410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014371592556635223,
+      "loss": 1.9138,
+      "step": 12411
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014377434832580456,
+      "loss": 2.1433,
+      "step": 12412
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014383277203132455,
+      "loss": 2.5305,
+      "step": 12413
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014389119667403418,
+      "loss": 2.5529,
+      "step": 12414
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014394962224505497,
+      "loss": 2.8122,
+      "step": 12415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014400804873550827,
+      "loss": 2.4739,
+      "step": 12416
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001440664761365157,
+      "loss": 2.7068,
+      "step": 12417
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014412490443919832,
+      "loss": 2.3981,
+      "step": 12418
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001441833336346772,
+      "loss": 2.2942,
+      "step": 12419
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014424176371407313,
+      "loss": 2.2894,
+      "step": 12420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001443001946685072,
+      "loss": 2.5612,
+      "step": 12421
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014435862648909994,
+      "loss": 2.8295,
+      "step": 12422
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014441705916697184,
+      "loss": 2.2494,
+      "step": 12423
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014447549269324322,
+      "loss": 2.1585,
+      "step": 12424
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014453392705903462,
+      "loss": 2.743,
+      "step": 12425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014459236225546605,
+      "loss": 2.5331,
+      "step": 12426
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014465079827365748,
+      "loss": 2.9236,
+      "step": 12427
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014470923510472876,
+      "loss": 2.5581,
+      "step": 12428
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014476767273979984,
+      "loss": 2.2296,
+      "step": 12429
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014482611116999029,
+      "loss": 2.4946,
+      "step": 12430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014488455038641964,
+      "loss": 2.5601,
+      "step": 12431
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014494299038020716,
+      "loss": 1.7951,
+      "step": 12432
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014500143114247238,
+      "loss": 2.3884,
+      "step": 12433
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014505987266433438,
+      "loss": 2.5252,
+      "step": 12434
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014511831493691223,
+      "loss": 2.3494,
+      "step": 12435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001451767579513247,
+      "loss": 2.4231,
+      "step": 12436
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014523520169869089,
+      "loss": 2.5067,
+      "step": 12437
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014529364617012943,
+      "loss": 2.5256,
+      "step": 12438
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001453520913567589,
+      "loss": 2.5057,
+      "step": 12439
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014541053724969767,
+      "loss": 1.7259,
+      "step": 12440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014546898384006443,
+      "loss": 1.4506,
+      "step": 12441
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001455274311189773,
+      "loss": 2.5988,
+      "step": 12442
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014558587907755434,
+      "loss": 2.432,
+      "step": 12443
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014564432770691394,
+      "loss": 2.3276,
+      "step": 12444
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014570277699817389,
+      "loss": 2.188,
+      "step": 12445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001457612269424521,
+      "loss": 2.5245,
+      "step": 12446
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001458196775308662,
+      "loss": 2.5726,
+      "step": 12447
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014587812875453416,
+      "loss": 2.4768,
+      "step": 12448
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014593658060457328,
+      "loss": 2.4278,
+      "step": 12449
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014599503307210145,
+      "loss": 2.1246,
+      "step": 12450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014605348614823554,
+      "loss": 2.6672,
+      "step": 12451
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014611193982409326,
+      "loss": 2.2615,
+      "step": 12452
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014617039409079154,
+      "loss": 2.4266,
+      "step": 12453
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014622884893944794,
+      "loss": 2.1967,
+      "step": 12454
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001462873043611789,
+      "loss": 2.4676,
+      "step": 12455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014634576034710183,
+      "loss": 2.7196,
+      "step": 12456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014640421688833326,
+      "loss": 2.2447,
+      "step": 12457
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014646267397599045,
+      "loss": 2.1163,
+      "step": 12458
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014652113160118946,
+      "loss": 2.0869,
+      "step": 12459
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001465795897550474,
+      "loss": 2.5036,
+      "step": 12460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001466380484286805,
+      "loss": 2.761,
+      "step": 12461
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014669650761320557,
+      "loss": 2.4203,
+      "step": 12462
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014675496729973854,
+      "loss": 2.5334,
+      "step": 12463
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014681342747939602,
+      "loss": 2.6411,
+      "step": 12464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000146871888143294,
+      "loss": 2.2901,
+      "step": 12465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014693034928254907,
+      "loss": 2.097,
+      "step": 12466
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014698881088827671,
+      "loss": 2.1726,
+      "step": 12467
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014704727295159338,
+      "loss": 2.5049,
+      "step": 12468
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014710573546361473,
+      "loss": 2.185,
+      "step": 12469
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014716419841545692,
+      "loss": 2.5948,
+      "step": 12470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014722266179823556,
+      "loss": 2.4782,
+      "step": 12471
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014728112560306647,
+      "loss": 2.4976,
+      "step": 12472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001473395898210651,
+      "loss": 2.3433,
+      "step": 12473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001473980544433474,
+      "loss": 2.4665,
+      "step": 12474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014745651946102876,
+      "loss": 2.4884,
+      "step": 12475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014751498486522454,
+      "loss": 2.5665,
+      "step": 12476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014757345064705043,
+      "loss": 2.2816,
+      "step": 12477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014763191679762166,
+      "loss": 2.2665,
+      "step": 12478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014769038330805358,
+      "loss": 2.3873,
+      "step": 12479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014774885016946132,
+      "loss": 2.6943,
+      "step": 12480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014780731737296029,
+      "loss": 2.2423,
+      "step": 12481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001478657849096656,
+      "loss": 2.5789,
+      "step": 12482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014792425277069228,
+      "loss": 2.0593,
+      "step": 12483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001479827209471553,
+      "loss": 2.385,
+      "step": 12484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014804118943016994,
+      "loss": 2.6309,
+      "step": 12485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014809965821085087,
+      "loss": 2.486,
+      "step": 12486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014815812728031345,
+      "loss": 2.4001,
+      "step": 12487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014821659662967197,
+      "loss": 2.1613,
+      "step": 12488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014827506625004165,
+      "loss": 2.3632,
+      "step": 12489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014833353613253708,
+      "loss": 2.466,
+      "step": 12490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014839200626827337,
+      "loss": 1.6693,
+      "step": 12491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014845047664836472,
+      "loss": 2.9992,
+      "step": 12492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014850894726392615,
+      "loss": 2.1862,
+      "step": 12493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001485674181060721,
+      "loss": 2.1749,
+      "step": 12494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001486258891659175,
+      "loss": 2.1686,
+      "step": 12495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014868436043457647,
+      "loss": 2.262,
+      "step": 12496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001487428319031639,
+      "loss": 2.7662,
+      "step": 12497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014880130356279404,
+      "loss": 2.8216,
+      "step": 12498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001488597754045818,
+      "loss": 2.4197,
+      "step": 12499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014891824741964104,
+      "loss": 2.2666,
+      "step": 12500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001489767195990866,
+      "loss": 2.1249,
+      "step": 12501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014903519193403268,
+      "loss": 2.8363,
+      "step": 12502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014909366441559385,
+      "loss": 2.0026,
+      "step": 12503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014915213703488436,
+      "loss": 2.2547,
+      "step": 12504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014921060978301853,
+      "loss": 2.4574,
+      "step": 12505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001492690826511106,
+      "loss": 2.254,
+      "step": 12506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014932755563027504,
+      "loss": 2.7218,
+      "step": 12507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014938602871162606,
+      "loss": 2.0708,
+      "step": 12508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014944450188627775,
+      "loss": 2.6567,
+      "step": 12509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001495029751453447,
+      "loss": 3.1392,
+      "step": 12510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014956144847994091,
+      "loss": 2.5511,
+      "step": 12511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001496199218811807,
+      "loss": 2.5049,
+      "step": 12512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014967839534017806,
+      "loss": 2.9964,
+      "step": 12513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014973686884804758,
+      "loss": 2.6004,
+      "step": 12514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001497953423959032,
+      "loss": 2.7124,
+      "step": 12515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014985381597485922,
+      "loss": 2.7239,
+      "step": 12516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014991228957602962,
+      "loss": 1.8981,
+      "step": 12517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00014997076319052887,
+      "loss": 2.6073,
+      "step": 12518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015002923680947105,
+      "loss": 2.0937,
+      "step": 12519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001500877104239703,
+      "loss": 2.422,
+      "step": 12520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001501461840251407,
+      "loss": 2.1141,
+      "step": 12521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001502046576040967,
+      "loss": 1.2986,
+      "step": 12522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015026313115195234,
+      "loss": 2.1261,
+      "step": 12523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015032160465982183,
+      "loss": 2.2538,
+      "step": 12524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015038007811881923,
+      "loss": 1.5143,
+      "step": 12525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000150438551520059,
+      "loss": 2.709,
+      "step": 12526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015049702485465526,
+      "loss": 2.8171,
+      "step": 12527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015055549811372217,
+      "loss": 2.2104,
+      "step": 12528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015061397128837386,
+      "loss": 2.3315,
+      "step": 12529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015067244436972488,
+      "loss": 1.7314,
+      "step": 12530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015073091734888933,
+      "loss": 2.2286,
+      "step": 12531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001507893902169814,
+      "loss": 2.4751,
+      "step": 12532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001508478629651156,
+      "loss": 2.6829,
+      "step": 12533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015090633558440607,
+      "loss": 2.2799,
+      "step": 12534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015096480806596723,
+      "loss": 2.5636,
+      "step": 12535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001510232804009133,
+      "loss": 2.426,
+      "step": 12536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015108175258035888,
+      "loss": 2.4079,
+      "step": 12537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015114022459541813,
+      "loss": 2.3403,
+      "step": 12538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015119869643720588,
+      "loss": 2.6893,
+      "step": 12539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000151257168096836,
+      "loss": 2.3144,
+      "step": 12540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015131563956542342,
+      "loss": 2.6659,
+      "step": 12541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001513741108340824,
+      "loss": 2.0941,
+      "step": 12542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015143258189392785,
+      "loss": 1.3355,
+      "step": 12543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015149105273607377,
+      "loss": 1.3664,
+      "step": 12544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001515495233516352,
+      "loss": 2.3542,
+      "step": 12545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015160799373172657,
+      "loss": 2.0696,
+      "step": 12546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015166646386746281,
+      "loss": 2.2525,
+      "step": 12547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015172493374995824,
+      "loss": 1.7645,
+      "step": 12548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015178340337032795,
+      "loss": 2.5881,
+      "step": 12549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015184187271968647,
+      "loss": 2.4691,
+      "step": 12550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015190034178914902,
+      "loss": 1.4534,
+      "step": 12551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015195881056982998,
+      "loss": 2.4967,
+      "step": 12552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001520172790528446,
+      "loss": 2.4089,
+      "step": 12553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015207574722930761,
+      "loss": 1.7595,
+      "step": 12554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00015213421509033433,
+      "loss": 2.4385,
+      "step": 12555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001521926826270396,
+      "loss": 2.313,
+      "step": 12556
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015225114983053863,
+      "loss": 2.6956,
+      "step": 12557
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015230961669194634,
+      "loss": 2.715,
+      "step": 12558
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015236808320237823,
+      "loss": 2.5044,
+      "step": 12559
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001524265493529495,
+      "loss": 2.0567,
+      "step": 12560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015248501513477538,
+      "loss": 2.4293,
+      "step": 12561
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015254348053897113,
+      "loss": 2.5932,
+      "step": 12562
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015260194555665252,
+      "loss": 2.6282,
+      "step": 12563
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001526604101789348,
+      "loss": 2.6913,
+      "step": 12564
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015271887439693348,
+      "loss": 2.1094,
+      "step": 12565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015277733820176435,
+      "loss": 2.2555,
+      "step": 12566
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015283580158454303,
+      "loss": 2.5032,
+      "step": 12567
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001528942645363852,
+      "loss": 2.7319,
+      "step": 12568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015295272704840651,
+      "loss": 2.693,
+      "step": 12569
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001530111891117232,
+      "loss": 2.5598,
+      "step": 12570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015306965071745088,
+      "loss": 2.7082,
+      "step": 12571
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001531281118567059,
+      "loss": 2.6398,
+      "step": 12572
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001531865725206039,
+      "loss": 2.6561,
+      "step": 12573
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001532450327002614,
+      "loss": 2.6577,
+      "step": 12574
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015330349238679435,
+      "loss": 2.7277,
+      "step": 12575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015336195157131943,
+      "loss": 2.8319,
+      "step": 12576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015342041024495252,
+      "loss": 2.3681,
+      "step": 12577
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015347886839881046,
+      "loss": 2.5095,
+      "step": 12578
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015353732602400947,
+      "loss": 2.2472,
+      "step": 12579
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015359578311166663,
+      "loss": 2.3339,
+      "step": 12580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015365423965289812,
+      "loss": 2.6004,
+      "step": 12581
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000153712695638821,
+      "loss": 2.4739,
+      "step": 12582
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015377115106055198,
+      "loss": 2.6556,
+      "step": 12583
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015382960590920835,
+      "loss": 2.1027,
+      "step": 12584
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015388806017590666,
+      "loss": 2.495,
+      "step": 12585
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015394651385176435,
+      "loss": 2.9163,
+      "step": 12586
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015400496692789846,
+      "loss": 2.0669,
+      "step": 12587
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015406341939542664,
+      "loss": 3.0184,
+      "step": 12588
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015412187124546578,
+      "loss": 2.3745,
+      "step": 12589
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015418032246913372,
+      "loss": 2.2223,
+      "step": 12590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001542387730575478,
+      "loss": 2.0872,
+      "step": 12591
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015429722300182603,
+      "loss": 2.4496,
+      "step": 12592
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015435567229308598,
+      "loss": 2.2619,
+      "step": 12593
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015441412092244555,
+      "loss": 2.1217,
+      "step": 12594
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015447256888102262,
+      "loss": 3.181,
+      "step": 12595
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001545310161599355,
+      "loss": 2.4302,
+      "step": 12596
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015458946275030225,
+      "loss": 2.5027,
+      "step": 12597
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015464790864324102,
+      "loss": 2.9209,
+      "step": 12598
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015470635382987047,
+      "loss": 1.9597,
+      "step": 12599
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000154764798301309,
+      "loss": 2.3483,
+      "step": 12600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015482324204867524,
+      "loss": 2.8899,
+      "step": 12601
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015488168506308772,
+      "loss": 2.348,
+      "step": 12602
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001549401273356655,
+      "loss": 2.8566,
+      "step": 12603
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015499856885752754,
+      "loss": 2.7873,
+      "step": 12604
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015505700961979276,
+      "loss": 2.7105,
+      "step": 12605
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001551154496135803,
+      "loss": 2.4577,
+      "step": 12606
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015517388883000963,
+      "loss": 2.3619,
+      "step": 12607
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001552323272602001,
+      "loss": 2.3366,
+      "step": 12608
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015529076489527118,
+      "loss": 2.5347,
+      "step": 12609
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001553492017263424,
+      "loss": 2.581,
+      "step": 12610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001554076377445339,
+      "loss": 2.3935,
+      "step": 12611
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001554076377445339,
+      "loss": 2.8865,
+      "step": 12612
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001554660729409653,
+      "loss": 2.4971,
+      "step": 12613
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015552450730675667,
+      "loss": 1.9825,
+      "step": 12614
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015558294083302808,
+      "loss": 2.1477,
+      "step": 12615
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015564137351089998,
+      "loss": 2.5209,
+      "step": 12616
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001556998053314927,
+      "loss": 2.4229,
+      "step": 12617
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001557582362859268,
+      "loss": 2.3012,
+      "step": 12618
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015581666636532273,
+      "loss": 2.765,
+      "step": 12619
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001558750955608016,
+      "loss": 2.3594,
+      "step": 12620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015593352386348422,
+      "loss": 2.3778,
+      "step": 12621
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015599195126449165,
+      "loss": 2.5168,
+      "step": 12622
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015605037775494494,
+      "loss": 2.4119,
+      "step": 12623
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015610880332596574,
+      "loss": 2.2943,
+      "step": 12624
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001561672279686754,
+      "loss": 2.1954,
+      "step": 12625
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015622565167419534,
+      "loss": 2.7229,
+      "step": 12626
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015628407443364766,
+      "loss": 2.4171,
+      "step": 12627
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015634249623815413,
+      "loss": 2.0514,
+      "step": 12628
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015640091707883668,
+      "loss": 2.3586,
+      "step": 12629
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015645933694681774,
+      "loss": 2.3383,
+      "step": 12630
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015651775583321959,
+      "loss": 1.8578,
+      "step": 12631
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015657617372916453,
+      "loss": 2.3831,
+      "step": 12632
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001566345906257755,
+      "loss": 2.8756,
+      "step": 12633
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001566930065141752,
+      "loss": 2.2057,
+      "step": 12634
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001567514213854865,
+      "loss": 2.7794,
+      "step": 12635
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015680983523083252,
+      "loss": 1.8601,
+      "step": 12636
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015686824804133666,
+      "loss": 2.7426,
+      "step": 12637
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015692665980812228,
+      "loss": 2.1788,
+      "step": 12638
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001569850705223129,
+      "loss": 2.5228,
+      "step": 12639
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001570434801750322,
+      "loss": 2.6406,
+      "step": 12640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015710188875740427,
+      "loss": 2.2985,
+      "step": 12641
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015716029626055316,
+      "loss": 2.4963,
+      "step": 12642
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000157218702675603,
+      "loss": 2.8967,
+      "step": 12643
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015727710799367804,
+      "loss": 2.8247,
+      "step": 12644
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015733551220590316,
+      "loss": 2.7893,
+      "step": 12645
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015739391530340295,
+      "loss": 2.5181,
+      "step": 12646
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015745231727730231,
+      "loss": 2.6557,
+      "step": 12647
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001575107181187262,
+      "loss": 1.8181,
+      "step": 12648
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015756911781880005,
+      "loss": 2.5256,
+      "step": 12649
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001576275163686492,
+      "loss": 2.6195,
+      "step": 12650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015768591375939922,
+      "loss": 2.4799,
+      "step": 12651
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015774430998217578,
+      "loss": 2.8261,
+      "step": 12652
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015780270502810503,
+      "loss": 2.2662,
+      "step": 12653
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000157861098888313,
+      "loss": 2.8424,
+      "step": 12654
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015791949155392586,
+      "loss": 1.76,
+      "step": 12655
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015797788301607035,
+      "loss": 2.4914,
+      "step": 12656
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000158036273265873,
+      "loss": 2.5558,
+      "step": 12657
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015809466229446066,
+      "loss": 2.0547,
+      "step": 12658
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015815305009296026,
+      "loss": 2.6631,
+      "step": 12659
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015821143665249924,
+      "loss": 2.2838,
+      "step": 12660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001582698219642048,
+      "loss": 2.5115,
+      "step": 12661
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015832820601920493,
+      "loss": 2.1995,
+      "step": 12662
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015838658880862683,
+      "loss": 2.4584,
+      "step": 12663
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015844497032359894,
+      "loss": 2.8822,
+      "step": 12664
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001585033505552492,
+      "loss": 2.0524,
+      "step": 12665
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015856172949470633,
+      "loss": 1.9503,
+      "step": 12666
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015862010713309837,
+      "loss": 2.3786,
+      "step": 12667
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001586784834615545,
+      "loss": 2.1382,
+      "step": 12668
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015873685847120342,
+      "loss": 1.9779,
+      "step": 12669
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001587952321531747,
+      "loss": 2.8177,
+      "step": 12670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001588536044985972,
+      "loss": 2.524,
+      "step": 12671
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015891197549860084,
+      "loss": 2.8553,
+      "step": 12672
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015897034514431515,
+      "loss": 2.3368,
+      "step": 12673
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015902871342687055,
+      "loss": 2.3662,
+      "step": 12674
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015908708033739662,
+      "loss": 2.7201,
+      "step": 12675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015914544586702426,
+      "loss": 2.8577,
+      "step": 12676
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015920381000688368,
+      "loss": 2.5762,
+      "step": 12677
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015926217274810623,
+      "loss": 1.9159,
+      "step": 12678
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015932053408182237,
+      "loss": 2.4311,
+      "step": 12679
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015937889399916368,
+      "loss": 2.7759,
+      "step": 12680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001594372524912614,
+      "loss": 2.3046,
+      "step": 12681
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015949560954924748,
+      "loss": 2.9464,
+      "step": 12682
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015955396516425366,
+      "loss": 2.4254,
+      "step": 12683
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015961231932741212,
+      "loss": 2.6519,
+      "step": 12684
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000159670672029855,
+      "loss": 2.234,
+      "step": 12685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001597290232627151,
+      "loss": 2.5602,
+      "step": 12686
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001597873730171251,
+      "loss": 2.9213,
+      "step": 12687
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015984572128421788,
+      "loss": 2.6288,
+      "step": 12688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015990406805512698,
+      "loss": 2.9458,
+      "step": 12689
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00015996241332098563,
+      "loss": 1.7911,
+      "step": 12690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001600207570729276,
+      "loss": 2.3983,
+      "step": 12691
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016007909930208658,
+      "loss": 2.2267,
+      "step": 12692
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016013743999959712,
+      "loss": 2.909,
+      "step": 12693
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016019577915659335,
+      "loss": 2.102,
+      "step": 12694
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016025411676420993,
+      "loss": 2.6451,
+      "step": 12695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016031245281358157,
+      "loss": 2.4422,
+      "step": 12696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016037078729584363,
+      "loss": 2.8219,
+      "step": 12697
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001604291202021313,
+      "loss": 2.1568,
+      "step": 12698
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016048745152358017,
+      "loss": 2.492,
+      "step": 12699
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016054578125132587,
+      "loss": 2.5384,
+      "step": 12700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016060410937650474,
+      "loss": 2.4094,
+      "step": 12701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001606624358902528,
+      "loss": 2.6208,
+      "step": 12702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016072076078370697,
+      "loss": 2.7097,
+      "step": 12703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016077908404800356,
+      "loss": 2.5399,
+      "step": 12704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001608374056742799,
+      "loss": 1.6791,
+      "step": 12705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001608957256536731,
+      "loss": 2.1933,
+      "step": 12706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016095404397732102,
+      "loss": 2.5136,
+      "step": 12707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016101236063636095,
+      "loss": 2.125,
+      "step": 12708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016107067562193134,
+      "loss": 2.5894,
+      "step": 12709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016112898892517015,
+      "loss": 2.0729,
+      "step": 12710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016118730053721637,
+      "loss": 2.564,
+      "step": 12711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016124561044920826,
+      "loss": 2.4286,
+      "step": 12712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016130391865228527,
+      "loss": 2.5404,
+      "step": 12713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016136222513758646,
+      "loss": 1.9664,
+      "step": 12714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016142052989625167,
+      "loss": 1.3051,
+      "step": 12715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016147883291942064,
+      "loss": 1.861,
+      "step": 12716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016153713419823343,
+      "loss": 1.6357,
+      "step": 12717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001615954337238303,
+      "loss": 2.2731,
+      "step": 12718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016165373148735215,
+      "loss": 1.5147,
+      "step": 12719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016171202747993976,
+      "loss": 2.0622,
+      "step": 12720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016177032169273418,
+      "loss": 2.4831,
+      "step": 12721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001618286141168771,
+      "loss": 1.9572,
+      "step": 12722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016188690474351012,
+      "loss": 2.4386,
+      "step": 12723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001619451935637752,
+      "loss": 1.9157,
+      "step": 12724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016200348056881453,
+      "loss": 2.1844,
+      "step": 12725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016206176574977085,
+      "loss": 2.3492,
+      "step": 12726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016212004909778693,
+      "loss": 2.2833,
+      "step": 12727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016217833060400578,
+      "loss": 2.4478,
+      "step": 12728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016223661025957064,
+      "loss": 2.4522,
+      "step": 12729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016229488805562554,
+      "loss": 2.4043,
+      "step": 12730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016235316398331416,
+      "loss": 2.9657,
+      "step": 12731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016241143803378085,
+      "loss": 2.4181,
+      "step": 12732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016246971019816985,
+      "loss": 2.1217,
+      "step": 12733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016252798046762632,
+      "loss": 2.0197,
+      "step": 12734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001625862488332952,
+      "loss": 2.5324,
+      "step": 12735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001626445152863218,
+      "loss": 2.5926,
+      "step": 12736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016270277981785168,
+      "loss": 2.2008,
+      "step": 12737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016276104241903113,
+      "loss": 2.646,
+      "step": 12738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001628193030810062,
+      "loss": 2.6284,
+      "step": 12739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016287756179492343,
+      "loss": 2.5217,
+      "step": 12740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016293581855192956,
+      "loss": 2.3248,
+      "step": 12741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016299407334317194,
+      "loss": 3.0344,
+      "step": 12742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016305232615979796,
+      "loss": 2.7947,
+      "step": 12743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016311057699295515,
+      "loss": 2.198,
+      "step": 12744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016316882583379183,
+      "loss": 1.2373,
+      "step": 12745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016322707267345623,
+      "loss": 2.8092,
+      "step": 12746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016328531750309694,
+      "loss": 2.6043,
+      "step": 12747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016334356031386284,
+      "loss": 2.7951,
+      "step": 12748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016340180109690342,
+      "loss": 1.1779,
+      "step": 12749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016346003984336814,
+      "loss": 2.1841,
+      "step": 12750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016351827654440682,
+      "loss": 2.3671,
+      "step": 12751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016357651119116955,
+      "loss": 2.7336,
+      "step": 12752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016363474377480708,
+      "loss": 2.379,
+      "step": 12753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016369297428647,
+      "loss": 2.2751,
+      "step": 12754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016375120271730977,
+      "loss": 2.185,
+      "step": 12755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016380942905847734,
+      "loss": 3.0588,
+      "step": 12756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016386765330112484,
+      "loss": 2.2526,
+      "step": 12757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016392587543640407,
+      "loss": 2.6439,
+      "step": 12758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001639840954554679,
+      "loss": 2.4517,
+      "step": 12759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016404231334946846,
+      "loss": 2.6331,
+      "step": 12760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001641005291095592,
+      "loss": 2.6258,
+      "step": 12761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001641587427268932,
+      "loss": 2.7461,
+      "step": 12762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016421695419262461,
+      "loss": 2.4934,
+      "step": 12763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016427516349790689,
+      "loss": 1.1102,
+      "step": 12764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016433337063389476,
+      "loss": 2.8735,
+      "step": 12765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016439157559174265,
+      "loss": 2.7817,
+      "step": 12766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000164449778362606,
+      "loss": 2.2161,
+      "step": 12767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016450797893763954,
+      "loss": 2.6128,
+      "step": 12768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016456617730799944,
+      "loss": 2.3802,
+      "step": 12769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001646243734648414,
+      "loss": 2.2551,
+      "step": 12770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000164682567399322,
+      "loss": 2.0381,
+      "step": 12771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016474075910259784,
+      "loss": 2.2428,
+      "step": 12772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001647989485658259,
+      "loss": 2.3294,
+      "step": 12773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016485713578016348,
+      "loss": 2.5054,
+      "step": 12774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016491532073676846,
+      "loss": 2.241,
+      "step": 12775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016497350342679885,
+      "loss": 2.5051,
+      "step": 12776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001650316838414128,
+      "loss": 1.8774,
+      "step": 12777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001650898619717694,
+      "loss": 2.0962,
+      "step": 12778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016514803780902758,
+      "loss": 2.2789,
+      "step": 12779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016520621134434677,
+      "loss": 2.3065,
+      "step": 12780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016526438256888663,
+      "loss": 2.4739,
+      "step": 12781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016532255147380757,
+      "loss": 2.4203,
+      "step": 12782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016538071805026976,
+      "loss": 2.497,
+      "step": 12783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016543888228943448,
+      "loss": 2.4757,
+      "step": 12784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016549704418246242,
+      "loss": 2.6816,
+      "step": 12785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016555520372051546,
+      "loss": 2.7662,
+      "step": 12786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001656133608947553,
+      "loss": 2.6091,
+      "step": 12787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001656715156963446,
+      "loss": 2.6787,
+      "step": 12788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016572966811644545,
+      "loss": 2.5544,
+      "step": 12789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016578781814622127,
+      "loss": 2.7375,
+      "step": 12790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016584596577683507,
+      "loss": 2.2554,
+      "step": 12791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016590411099945106,
+      "loss": 2.6232,
+      "step": 12792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016596225380523277,
+      "loss": 2.715,
+      "step": 12793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016602039418534502,
+      "loss": 2.5521,
+      "step": 12794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001660785321309524,
+      "loss": 2.4308,
+      "step": 12795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001661366676332205,
+      "loss": 1.7097,
+      "step": 12796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016619480068331432,
+      "loss": 2.7693,
+      "step": 12797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001662529312724002,
+      "loss": 2.6956,
+      "step": 12798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016631105939164423,
+      "loss": 2.4697,
+      "step": 12799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016636918503221346,
+      "loss": 3.1979,
+      "step": 12800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016642730818527442,
+      "loss": 2.6747,
+      "step": 12801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016648542884199494,
+      "loss": 2.1703,
+      "step": 12802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016654354699354262,
+      "loss": 2.6984,
+      "step": 12803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016660166263108582,
+      "loss": 2.9912,
+      "step": 12804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016665977574579314,
+      "loss": 2.0786,
+      "step": 12805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016671788632883342,
+      "loss": 2.3014,
+      "step": 12806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016677599437137593,
+      "loss": 2.2517,
+      "step": 12807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001668340998645907,
+      "loss": 2.2994,
+      "step": 12808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016689220279964765,
+      "loss": 2.8923,
+      "step": 12809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016695030316771721,
+      "loss": 2.5596,
+      "step": 12810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016700840095997053,
+      "loss": 1.884,
+      "step": 12811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016706649616757881,
+      "loss": 2.109,
+      "step": 12812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016712458878171365,
+      "loss": 2.6213,
+      "step": 12813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016718267879354707,
+      "loss": 2.3461,
+      "step": 12814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001672407661942518,
+      "loss": 2.5934,
+      "step": 12815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016729885097500056,
+      "loss": 2.5426,
+      "step": 12816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016735693312696663,
+      "loss": 2.0467,
+      "step": 12817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016741501264132358,
+      "loss": 1.6582,
+      "step": 12818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001674730895092457,
+      "loss": 2.4708,
+      "step": 12819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016753116372190735,
+      "loss": 2.5716,
+      "step": 12820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016758923527048343,
+      "loss": 2.7214,
+      "step": 12821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016764730414614906,
+      "loss": 1.929,
+      "step": 12822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016770537034008017,
+      "loss": 1.9442,
+      "step": 12823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016776343384345278,
+      "loss": 2.5396,
+      "step": 12824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016782149464744333,
+      "loss": 2.0469,
+      "step": 12825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016787955274322867,
+      "loss": 2.5808,
+      "step": 12826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016793760812198636,
+      "loss": 2.5381,
+      "step": 12827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000167995660774894,
+      "loss": 2.376,
+      "step": 12828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016805371069312977,
+      "loss": 2.6137,
+      "step": 12829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016811175786787206,
+      "loss": 2.605,
+      "step": 12830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016816980229030016,
+      "loss": 1.8365,
+      "step": 12831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016822784395159333,
+      "loss": 1.7416,
+      "step": 12832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016828588284293144,
+      "loss": 1.086,
+      "step": 12833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001683439189554945,
+      "loss": 2.3664,
+      "step": 12834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001684019522804635,
+      "loss": 2.1752,
+      "step": 12835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016845998280901942,
+      "loss": 2.3078,
+      "step": 12836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001685180105323436,
+      "loss": 2.8043,
+      "step": 12837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001685760354416183,
+      "loss": 2.6081,
+      "step": 12838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016863405752802566,
+      "loss": 2.1207,
+      "step": 12839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001686920767827486,
+      "loss": 2.4202,
+      "step": 12840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016875009319697012,
+      "loss": 3.107,
+      "step": 12841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016880810676187422,
+      "loss": 2.1956,
+      "step": 12842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016886611746864467,
+      "loss": 2.2269,
+      "step": 12843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001689241253084664,
+      "loss": 2.4952,
+      "step": 12844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016898213027252386,
+      "loss": 1.9119,
+      "step": 12845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001690401323520029,
+      "loss": 2.356,
+      "step": 12846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000169098131538089,
+      "loss": 2.1661,
+      "step": 12847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016915612782196875,
+      "loss": 2.2215,
+      "step": 12848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016921412119482875,
+      "loss": 2.8955,
+      "step": 12849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001692721116478561,
+      "loss": 2.3379,
+      "step": 12850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016933009917223835,
+      "loss": 2.8128,
+      "step": 12851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016938808375916377,
+      "loss": 2.6479,
+      "step": 12852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016944606539982078,
+      "loss": 2.4613,
+      "step": 12853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016950404408539824,
+      "loss": 2.2957,
+      "step": 12854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016956201980708544,
+      "loss": 2.7808,
+      "step": 12855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016961999255607255,
+      "loss": 2.1315,
+      "step": 12856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016967796232354965,
+      "loss": 2.3199,
+      "step": 12857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016973592910070752,
+      "loss": 3.0387,
+      "step": 12858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016979389287873723,
+      "loss": 2.1244,
+      "step": 12859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016985185364883077,
+      "loss": 2.8288,
+      "step": 12860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016990981140218,
+      "loss": 2.324,
+      "step": 12861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00016996776612997755,
+      "loss": 2.282,
+      "step": 12862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017002571782341638,
+      "loss": 2.2768,
+      "step": 12863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017008366647369016,
+      "loss": 2.5798,
+      "step": 12864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001701416120719928,
+      "loss": 1.8359,
+      "step": 12865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001701995546095185,
+      "loss": 2.1815,
+      "step": 12866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017025749407746248,
+      "loss": 2.6321,
+      "step": 12867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017031543046701995,
+      "loss": 2.5879,
+      "step": 12868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017037336376938666,
+      "loss": 2.5217,
+      "step": 12869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017043129397575888,
+      "loss": 2.6578,
+      "step": 12870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017048922107733356,
+      "loss": 2.5542,
+      "step": 12871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017054714506530773,
+      "loss": 2.2944,
+      "step": 12872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001706050659308794,
+      "loss": 2.4794,
+      "step": 12873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017066298366524632,
+      "loss": 2.1274,
+      "step": 12874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017072089825960748,
+      "loss": 2.2714,
+      "step": 12875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017077880970516174,
+      "loss": 2.355,
+      "step": 12876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017083671799310917,
+      "loss": 2.2125,
+      "step": 12877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017089462311464925,
+      "loss": 2.6292,
+      "step": 12878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017095252506098304,
+      "loss": 2.5063,
+      "step": 12879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017101042382331128,
+      "loss": 2.8393,
+      "step": 12880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001710683193928359,
+      "loss": 2.7516,
+      "step": 12881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017112621176075838,
+      "loss": 2.7082,
+      "step": 12882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001711841009182817,
+      "loss": 1.5283,
+      "step": 12883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017124198685660854,
+      "loss": 2.3935,
+      "step": 12884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001712998695669428,
+      "loss": 2.5842,
+      "step": 12885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001713577490404879,
+      "loss": 2.5537,
+      "step": 12886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017141562526844876,
+      "loss": 2.4129,
+      "step": 12887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017147349824203005,
+      "loss": 2.5552,
+      "step": 12888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001715313679524376,
+      "loss": 2.2943,
+      "step": 12889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017158923439087687,
+      "loss": 2.7077,
+      "step": 12890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001716470975485547,
+      "loss": 2.676,
+      "step": 12891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017170495741667777,
+      "loss": 2.6577,
+      "step": 12892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017176281398645376,
+      "loss": 3.1624,
+      "step": 12893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017182066724909058,
+      "loss": 2.8241,
+      "step": 12894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017187851719579662,
+      "loss": 2.8978,
+      "step": 12895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001719363638177807,
+      "loss": 2.1902,
+      "step": 12896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017199420710625258,
+      "loss": 2.6336,
+      "step": 12897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017205204705242203,
+      "loss": 2.5363,
+      "step": 12898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017210988364749948,
+      "loss": 2.8103,
+      "step": 12899
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017216771688269612,
+      "loss": 2.4558,
+      "step": 12900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017222554674922337,
+      "loss": 2.7415,
+      "step": 12901
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017228337323829318,
+      "loss": 2.378,
+      "step": 12902
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017234119634111796,
+      "loss": 2.7132,
+      "step": 12903
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017239901604891103,
+      "loss": 2.673,
+      "step": 12904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001724568323528858,
+      "loss": 2.5089,
+      "step": 12905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017251464524425632,
+      "loss": 2.1464,
+      "step": 12906
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017257245471423712,
+      "loss": 2.6123,
+      "step": 12907
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017263026075404344,
+      "loss": 2.2981,
+      "step": 12908
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017268806335489088,
+      "loss": 2.3792,
+      "step": 12909
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017274586250799558,
+      "loss": 2.4588,
+      "step": 12910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017280365820457403,
+      "loss": 2.3139,
+      "step": 12911
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017286145043584377,
+      "loss": 2.8397,
+      "step": 12912
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017291923919302235,
+      "loss": 2.2491,
+      "step": 12913
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017297702446732803,
+      "loss": 2.7189,
+      "step": 12914
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017303480624997948,
+      "loss": 2.811,
+      "step": 12915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017309258453219626,
+      "loss": 2.4028,
+      "step": 12916
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001731503593051981,
+      "loss": 2.3149,
+      "step": 12917
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017320813056020536,
+      "loss": 2.4188,
+      "step": 12918
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017326589828843886,
+      "loss": 2.6254,
+      "step": 12919
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017332366248112028,
+      "loss": 2.5262,
+      "step": 12920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017338142312947133,
+      "loss": 2.0773,
+      "step": 12921
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000173439180224715,
+      "loss": 2.5341,
+      "step": 12922
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001734969337580738,
+      "loss": 1.2478,
+      "step": 12923
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017355468372077168,
+      "loss": 2.3196,
+      "step": 12924
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017361243010403262,
+      "loss": 2.7163,
+      "step": 12925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017367017289908148,
+      "loss": 2.6563,
+      "step": 12926
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017372791209714346,
+      "loss": 2.4694,
+      "step": 12927
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017378564768944432,
+      "loss": 2.2319,
+      "step": 12928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001738433796672102,
+      "loss": 2.4437,
+      "step": 12929
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017390110802166834,
+      "loss": 2.0645,
+      "step": 12930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000173958832744046,
+      "loss": 2.6414,
+      "step": 12931
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017401655382557102,
+      "loss": 2.3911,
+      "step": 12932
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017407427125747221,
+      "loss": 2.099,
+      "step": 12933
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017413198503097855,
+      "loss": 2.2699,
+      "step": 12934
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001741896951373197,
+      "loss": 3.2874,
+      "step": 12935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001742474015677257,
+      "loss": 2.685,
+      "step": 12936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017430510431342756,
+      "loss": 2.8713,
+      "step": 12937
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001743628033656566,
+      "loss": 2.2351,
+      "step": 12938
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017442049871564455,
+      "loss": 2.9443,
+      "step": 12939
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001744781903546238,
+      "loss": 2.3855,
+      "step": 12940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017453587827382766,
+      "loss": 2.2717,
+      "step": 12941
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017459356246448955,
+      "loss": 1.98,
+      "step": 12942
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001746512429178436,
+      "loss": 2.5029,
+      "step": 12943
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017470891962512438,
+      "loss": 2.4237,
+      "step": 12944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017476659257756747,
+      "loss": 2.8496,
+      "step": 12945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001748242617664086,
+      "loss": 2.5801,
+      "step": 12946
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017488192718288415,
+      "loss": 2.8573,
+      "step": 12947
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000174939588818231,
+      "loss": 2.5498,
+      "step": 12948
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017499724666368704,
+      "loss": 2.4781,
+      "step": 12949
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017505490071049025,
+      "loss": 2.6272,
+      "step": 12950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017511255094987936,
+      "loss": 2.684,
+      "step": 12951
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001751701973730935,
+      "loss": 2.2379,
+      "step": 12952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001752278399713729,
+      "loss": 2.7222,
+      "step": 12953
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017528547873595782,
+      "loss": 2.5692,
+      "step": 12954
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017534311365808935,
+      "loss": 2.1182,
+      "step": 12955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000175400744729009,
+      "loss": 2.2975,
+      "step": 12956
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001754583719399592,
+      "loss": 2.9307,
+      "step": 12957
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017551599528218268,
+      "loss": 2.2559,
+      "step": 12958
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017557361474692269,
+      "loss": 2.5989,
+      "step": 12959
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017563123032542346,
+      "loss": 2.5516,
+      "step": 12960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017568884200892942,
+      "loss": 1.5606,
+      "step": 12961
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017574644978868574,
+      "loss": 2.9309,
+      "step": 12962
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001758040536559381,
+      "loss": 2.19,
+      "step": 12963
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017586165360193305,
+      "loss": 2.1841,
+      "step": 12964
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017591924961791729,
+      "loss": 2.2643,
+      "step": 12965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017597684169513877,
+      "loss": 2.9524,
+      "step": 12966
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001760344298248451,
+      "loss": 2.912,
+      "step": 12967
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017609201399828545,
+      "loss": 1.9323,
+      "step": 12968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017614959420670883,
+      "loss": 2.5612,
+      "step": 12969
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017620717044136564,
+      "loss": 2.2253,
+      "step": 12970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001762647426935059,
+      "loss": 2.8818,
+      "step": 12971
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017632231095438114,
+      "loss": 2.5762,
+      "step": 12972
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001763798752152429,
+      "loss": 2.4349,
+      "step": 12973
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001764374354673439,
+      "loss": 2.1464,
+      "step": 12974
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001764949917019366,
+      "loss": 2.7882,
+      "step": 12975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017655254391027501,
+      "loss": 2.5148,
+      "step": 12976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017661009208361306,
+      "loss": 2.475,
+      "step": 12977
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017666763621320593,
+      "loss": 2.6245,
+      "step": 12978
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017672517629030858,
+      "loss": 2.6218,
+      "step": 12979
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017678271230617742,
+      "loss": 2.6156,
+      "step": 12980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017684024425206884,
+      "loss": 2.2603,
+      "step": 12981
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017689777211924042,
+      "loss": 2.4636,
+      "step": 12982
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017695529589894986,
+      "loss": 1.9507,
+      "step": 12983
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017701281558245576,
+      "loss": 2.0593,
+      "step": 12984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017707033116101708,
+      "loss": 2.0252,
+      "step": 12985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001771278426258939,
+      "loss": 2.7165,
+      "step": 12986
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017718534996834644,
+      "loss": 2.8518,
+      "step": 12987
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001772428531796356,
+      "loss": 2.5596,
+      "step": 12988
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001773003522510233,
+      "loss": 2.6374,
+      "step": 12989
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017735784717377168,
+      "loss": 2.1059,
+      "step": 12990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017741533793914368,
+      "loss": 2.9525,
+      "step": 12991
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001774728245384026,
+      "loss": 2.0106,
+      "step": 12992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000177530306962813,
+      "loss": 2.4012,
+      "step": 12993
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001775877852036394,
+      "loss": 2.2864,
+      "step": 12994
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001776452592521476,
+      "loss": 0.9684,
+      "step": 12995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017770272909960316,
+      "loss": 2.5663,
+      "step": 12996
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001777601947372732,
+      "loss": 3.1762,
+      "step": 12997
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017781765615642483,
+      "loss": 2.3754,
+      "step": 12998
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001778751133483264,
+      "loss": 2.5563,
+      "step": 12999
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001779325663042461,
+      "loss": 2.7433,
+      "step": 13000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001779900150154535,
+      "loss": 1.7734,
+      "step": 13001
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017804745947321837,
+      "loss": 2.4636,
+      "step": 13002
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017810489966881162,
+      "loss": 2.5031,
+      "step": 13003
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017810489966881162,
+      "loss": 2.6474,
+      "step": 13004
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017816233559350402,
+      "loss": 1.9895,
+      "step": 13005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017821976723856776,
+      "loss": 2.1241,
+      "step": 13006
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001782771945952751,
+      "loss": 2.6193,
+      "step": 13007
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017833461765489966,
+      "loss": 2.7067,
+      "step": 13008
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001783920364087147,
+      "loss": 2.5203,
+      "step": 13009
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017844945084799518,
+      "loss": 2.1007,
+      "step": 13010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017850686096401586,
+      "loss": 2.5621,
+      "step": 13011
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017856426674805301,
+      "loss": 2.8987,
+      "step": 13012
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001786216681913825,
+      "loss": 2.2109,
+      "step": 13013
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001786790652852819,
+      "loss": 2.6743,
+      "step": 13014
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017873645802102866,
+      "loss": 1.323,
+      "step": 13015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017879384638990153,
+      "loss": 1.9895,
+      "step": 13016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017885123038317944,
+      "loss": 1.9103,
+      "step": 13017
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001789086099921422,
+      "loss": 2.08,
+      "step": 13018
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017896598520807005,
+      "loss": 3.0467,
+      "step": 13019
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001790233560222444,
+      "loss": 1.9751,
+      "step": 13020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017908072242594684,
+      "loss": 2.7231,
+      "step": 13021
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017913808441045976,
+      "loss": 2.261,
+      "step": 13022
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001791954419670664,
+      "loss": 2.8447,
+      "step": 13023
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017925279508705053,
+      "loss": 2.6745,
+      "step": 13024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001793101437616966,
+      "loss": 2.58,
+      "step": 13025
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017936748798228952,
+      "loss": 2.6745,
+      "step": 13026
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017942482774011546,
+      "loss": 2.2599,
+      "step": 13027
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001794821630264607,
+      "loss": 2.6796,
+      "step": 13028
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017953949383261246,
+      "loss": 2.5581,
+      "step": 13029
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017959682014985842,
+      "loss": 2.4785,
+      "step": 13030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017965414196948742,
+      "loss": 1.7041,
+      "step": 13031
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017971145928278844,
+      "loss": 2.5334,
+      "step": 13032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001797687720810515,
+      "loss": 2.373,
+      "step": 13033
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000179826080355567,
+      "loss": 2.4818,
+      "step": 13034
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017988338409762644,
+      "loss": 2.5779,
+      "step": 13035
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001799406832985217,
+      "loss": 2.4698,
+      "step": 13036
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00017999797794954545,
+      "loss": 2.6958,
+      "step": 13037
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018005526804199082,
+      "loss": 2.5148,
+      "step": 13038
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018011255356715217,
+      "loss": 2.5344,
+      "step": 13039
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018016983451632403,
+      "loss": 1.5977,
+      "step": 13040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018022711088080192,
+      "loss": 3.1004,
+      "step": 13041
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018028438265188176,
+      "loss": 2.7081,
+      "step": 13042
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018034164982086065,
+      "loss": 1.9492,
+      "step": 13043
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018039891237903597,
+      "loss": 1.9967,
+      "step": 13044
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018045617031770596,
+      "loss": 2.3143,
+      "step": 13045
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001805134236281694,
+      "loss": 2.4712,
+      "step": 13046
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018057067230172616,
+      "loss": 2.4695,
+      "step": 13047
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018062791632967646,
+      "loss": 2.5619,
+      "step": 13048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018068515570332118,
+      "loss": 2.5819,
+      "step": 13049
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018074239041396235,
+      "loss": 2.4734,
+      "step": 13050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018079962045290223,
+      "loss": 2.0568,
+      "step": 13051
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018085684581144404,
+      "loss": 2.4737,
+      "step": 13052
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018091406648089152,
+      "loss": 2.4769,
+      "step": 13053
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018097128245254953,
+      "loss": 2.6477,
+      "step": 13054
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018102849371772302,
+      "loss": 2.2936,
+      "step": 13055
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001810857002677185,
+      "loss": 2.6873,
+      "step": 13056
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018114290209384208,
+      "loss": 2.5224,
+      "step": 13057
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018120009918740171,
+      "loss": 2.7665,
+      "step": 13058
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001812572915397052,
+      "loss": 2.4779,
+      "step": 13059
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018131447914206185,
+      "loss": 2.221,
+      "step": 13060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001813716619857808,
+      "loss": 2.4522,
+      "step": 13061
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001814288400621727,
+      "loss": 2.3076,
+      "step": 13062
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001814860133625484,
+      "loss": 2.1598,
+      "step": 13063
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018154318187822002,
+      "loss": 2.6,
+      "step": 13064
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018160034560049963,
+      "loss": 2.8165,
+      "step": 13065
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018165750452070077,
+      "loss": 2.0371,
+      "step": 13066
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018171465863013717,
+      "loss": 2.5777,
+      "step": 13067
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018177180792012385,
+      "loss": 2.5994,
+      "step": 13068
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018182895238197605,
+      "loss": 1.9333,
+      "step": 13069
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018188609200700995,
+      "loss": 2.5242,
+      "step": 13070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018194322678654232,
+      "loss": 2.2013,
+      "step": 13071
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018200035671189107,
+      "loss": 2.4497,
+      "step": 13072
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018205748177437442,
+      "loss": 2.1172,
+      "step": 13073
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001821146019653115,
+      "loss": 2.5875,
+      "step": 13074
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000182171717276022,
+      "loss": 2.1408,
+      "step": 13075
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018222882769782685,
+      "loss": 2.5031,
+      "step": 13076
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018228593322204719,
+      "loss": 2.5439,
+      "step": 13077
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018234303384000514,
+      "loss": 2.5342,
+      "step": 13078
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018240012954302335,
+      "loss": 2.4248,
+      "step": 13079
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001824572203224257,
+      "loss": 2.5176,
+      "step": 13080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018251430616953636,
+      "loss": 2.5166,
+      "step": 13081
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018257138707568023,
+      "loss": 2.7581,
+      "step": 13082
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001826284630321835,
+      "loss": 2.5641,
+      "step": 13083
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018268553403037253,
+      "loss": 3.8828,
+      "step": 13084
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018274260006157468,
+      "loss": 2.2618,
+      "step": 13085
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001827996611171179,
+      "loss": 2.8324,
+      "step": 13086
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001828567171883313,
+      "loss": 2.6552,
+      "step": 13087
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018291376826654415,
+      "loss": 2.2352,
+      "step": 13088
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001829708143430873,
+      "loss": 2.6278,
+      "step": 13089
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001830278554092913,
+      "loss": 2.2381,
+      "step": 13090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001830848914564884,
+      "loss": 1.9915,
+      "step": 13091
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018314192247601094,
+      "loss": 2.6527,
+      "step": 13092
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018319894845919283,
+      "loss": 3.0024,
+      "step": 13093
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018325596939736764,
+      "loss": 2.4446,
+      "step": 13094
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001833129852818707,
+      "loss": 2.4071,
+      "step": 13095
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018336999610403746,
+      "loss": 2.4795,
+      "step": 13096
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018342700185520477,
+      "loss": 2.7565,
+      "step": 13097
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001834840025267094,
+      "loss": 2.0998,
+      "step": 13098
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018354099810988974,
+      "loss": 2.4446,
+      "step": 13099
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018359798859608425,
+      "loss": 2.8808,
+      "step": 13100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000183654973976633,
+      "loss": 2.1703,
+      "step": 13101
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001837119542428757,
+      "loss": 2.6189,
+      "step": 13102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018376892938615393,
+      "loss": 2.5099,
+      "step": 13103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018382589939780925,
+      "loss": 2.47,
+      "step": 13104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001838828642691847,
+      "loss": 2.59,
+      "step": 13105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018393982399162354,
+      "loss": 2.3431,
+      "step": 13106
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018399677855647,
+      "loss": 2.7028,
+      "step": 13107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018405372795506897,
+      "loss": 2.3996,
+      "step": 13108
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018411067217876658,
+      "loss": 2.0422,
+      "step": 13109
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018416761121890923,
+      "loss": 1.7114,
+      "step": 13110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001842245450668442,
+      "loss": 2.5966,
+      "step": 13111
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001842814737139199,
+      "loss": 2.0708,
+      "step": 13112
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018433839715148516,
+      "loss": 2.2476,
+      "step": 13113
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018439531537088976,
+      "loss": 2.2366,
+      "step": 13114
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018445222836348412,
+      "loss": 2.6515,
+      "step": 13115
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018450913612061979,
+      "loss": 2.6017,
+      "step": 13116
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001845660386336488,
+      "loss": 2.8196,
+      "step": 13117
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001846229358939241,
+      "loss": 2.4406,
+      "step": 13118
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018467982789279928,
+      "loss": 2.4676,
+      "step": 13119
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001847367146216291,
+      "loss": 2.5028,
+      "step": 13120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001847935960717688,
+      "loss": 2.9352,
+      "step": 13121
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018485047223457452,
+      "loss": 2.7182,
+      "step": 13122
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000184907343101403,
+      "loss": 2.317,
+      "step": 13123
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018496420866361235,
+      "loss": 2.2776,
+      "step": 13124
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018502106891256088,
+      "loss": 2.2072,
+      "step": 13125
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018507792383960797,
+      "loss": 2.3208,
+      "step": 13126
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018513477343611373,
+      "loss": 2.2782,
+      "step": 13127
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018519161769343928,
+      "loss": 2.3507,
+      "step": 13128
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018524845660294637,
+      "loss": 1.5872,
+      "step": 13129
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001853052901559975,
+      "loss": 1.9575,
+      "step": 13130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018536211834395606,
+      "loss": 2.9265,
+      "step": 13131
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018541894115818648,
+      "loss": 2.2359,
+      "step": 13132
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018547575859005367,
+      "loss": 2.2279,
+      "step": 13133
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001855325706309235,
+      "loss": 2.5242,
+      "step": 13134
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001855893772721625,
+      "loss": 2.4027,
+      "step": 13135
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018564617850513846,
+      "loss": 2.8615,
+      "step": 13136
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001857029743212196,
+      "loss": 2.442,
+      "step": 13137
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018575976471177488,
+      "loss": 2.5582,
+      "step": 13138
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018581654966817453,
+      "loss": 2.5362,
+      "step": 13139
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018587332918178922,
+      "loss": 2.5098,
+      "step": 13140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018593010324399048,
+      "loss": 2.4711,
+      "step": 13141
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000185986871846151,
+      "loss": 1.5594,
+      "step": 13142
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018604363497964397,
+      "loss": 2.7768,
+      "step": 13143
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018610039263584327,
+      "loss": 2.4955,
+      "step": 13144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001861571448061242,
+      "loss": 2.6493,
+      "step": 13145
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621389148186238,
+      "loss": 2.2038,
+      "step": 13146
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018627063265443442,
+      "loss": 2.5516,
+      "step": 13147
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018632736831521765,
+      "loss": 2.7361,
+      "step": 13148
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018638409845559057,
+      "loss": 2.3231,
+      "step": 13149
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018644082306693225,
+      "loss": 2.7949,
+      "step": 13150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018649754214062264,
+      "loss": 2.6859,
+      "step": 13151
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018655425566804236,
+      "loss": 2.7235,
+      "step": 13152
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018661096364057335,
+      "loss": 2.4139,
+      "step": 13153
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018666766604959802,
+      "loss": 3.2359,
+      "step": 13154
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018672436288649967,
+      "loss": 1.7952,
+      "step": 13155
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018678105414266234,
+      "loss": 2.5663,
+      "step": 13156
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868377398094714,
+      "loss": 2.4461,
+      "step": 13157
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868944198783125,
+      "loss": 2.6722,
+      "step": 13158
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018695109434057244,
+      "loss": 2.8483,
+      "step": 13159
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018700776318763868,
+      "loss": 2.3634,
+      "step": 13160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018706442641089989,
+      "loss": 2.4335,
+      "step": 13161
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018712108400174523,
+      "loss": 2.4791,
+      "step": 13162
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001871777359515649,
+      "loss": 2.4551,
+      "step": 13163
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018723438225174972,
+      "loss": 2.6521,
+      "step": 13164
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001872910228936919,
+      "loss": 2.1913,
+      "step": 13165
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018734765786878392,
+      "loss": 2.7414,
+      "step": 13166
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018740428716841945,
+      "loss": 2.5919,
+      "step": 13167
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018746091078399285,
+      "loss": 2.2573,
+      "step": 13168
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018751752870689955,
+      "loss": 2.3153,
+      "step": 13169
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018757414092853574,
+      "loss": 2.4245,
+      "step": 13170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018763074744029826,
+      "loss": 2.4178,
+      "step": 13171
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001876873482335853,
+      "loss": 2.5703,
+      "step": 13172
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018774394329979552,
+      "loss": 2.5747,
+      "step": 13173
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018780053263032855,
+      "loss": 2.0406,
+      "step": 13174
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018785711621658485,
+      "loss": 2.6678,
+      "step": 13175
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000187913694049966,
+      "loss": 2.0815,
+      "step": 13176
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000187970266121874,
+      "loss": 2.5305,
+      "step": 13177
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001880268324237125,
+      "loss": 2.8541,
+      "step": 13178
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018808339294688488,
+      "loss": 2.5543,
+      "step": 13179
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018813994768279648,
+      "loss": 2.4747,
+      "step": 13180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018819649662285285,
+      "loss": 2.8064,
+      "step": 13181
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000188253039758461,
+      "loss": 2.4201,
+      "step": 13182
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018830957708102799,
+      "loss": 2.3445,
+      "step": 13183
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018836610858196257,
+      "loss": 2.2373,
+      "step": 13184
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018842263425267387,
+      "loss": 2.7183,
+      "step": 13185
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018847915408457244,
+      "loss": 2.6239,
+      "step": 13186
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018853566806906888,
+      "loss": 2.5493,
+      "step": 13187
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018859217619757548,
+      "loss": 2.5839,
+      "step": 13188
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018864867846150491,
+      "loss": 2.6283,
+      "step": 13189
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018870517485227128,
+      "loss": 2.5673,
+      "step": 13190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018876166536128876,
+      "loss": 2.4389,
+      "step": 13191
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018881814997997323,
+      "loss": 0.8018,
+      "step": 13192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001888746286997409,
+      "loss": 2.64,
+      "step": 13193
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018893110151200937,
+      "loss": 2.034,
+      "step": 13194
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018898756840819668,
+      "loss": 2.3422,
+      "step": 13195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018904402937972202,
+      "loss": 2.4202,
+      "step": 13196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001891004844180053,
+      "loss": 2.3802,
+      "step": 13197
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018915693351446763,
+      "loss": 1.6365,
+      "step": 13198
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018921337666053081,
+      "loss": 2.0418,
+      "step": 13199
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018926981384761755,
+      "loss": 3.1487,
+      "step": 13200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001893262450671513,
+      "loss": 2.7802,
+      "step": 13201
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018938267031055696,
+      "loss": 2.5454,
+      "step": 13202
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018943908956925982,
+      "loss": 2.66,
+      "step": 13203
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018949550283468608,
+      "loss": 2.8174,
+      "step": 13204
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001895519100982634,
+      "loss": 2.725,
+      "step": 13205
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018960831135141968,
+      "loss": 2.7693,
+      "step": 13206
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018966470658558416,
+      "loss": 2.7939,
+      "step": 13207
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018972109579218664,
+      "loss": 2.3921,
+      "step": 13208
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018977747896265835,
+      "loss": 2.714,
+      "step": 13209
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018983385608843103,
+      "loss": 2.2591,
+      "step": 13210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018989022716093743,
+      "loss": 2.816,
+      "step": 13211
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001899465921716111,
+      "loss": 2.4112,
+      "step": 13212
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019000295111188696,
+      "loss": 2.3423,
+      "step": 13213
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019005930397320023,
+      "loss": 2.6313,
+      "step": 13214
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001901156507469878,
+      "loss": 2.0494,
+      "step": 13215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001901719914246865,
+      "loss": 2.4008,
+      "step": 13216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019022832599773507,
+      "loss": 2.4167,
+      "step": 13217
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019028465445757243,
+      "loss": 1.6413,
+      "step": 13218
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001903409767956392,
+      "loss": 2.1969,
+      "step": 13219
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001903972930033759,
+      "loss": 2.481,
+      "step": 13220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000190453603072225,
+      "loss": 2.4251,
+      "step": 13221
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019050990699362918,
+      "loss": 2.5314,
+      "step": 13222
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019056620475903268,
+      "loss": 2.4039,
+      "step": 13223
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019062249635987992,
+      "loss": 1.7721,
+      "step": 13224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019067878178761694,
+      "loss": 2.4991,
+      "step": 13225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019073506103369026,
+      "loss": 2.6426,
+      "step": 13226
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019079133408954774,
+      "loss": 2.7208,
+      "step": 13227
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019084760094663792,
+      "loss": 2.5879,
+      "step": 13228
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001909038615964102,
+      "loss": 2.7051,
+      "step": 13229
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019096011603031502,
+      "loss": 2.8754,
+      "step": 13230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019101636423980402,
+      "loss": 2.7071,
+      "step": 13231
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019107260621632944,
+      "loss": 1.825,
+      "step": 13232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019112884195134444,
+      "loss": 2.493,
+      "step": 13233
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001911850714363036,
+      "loss": 2.2413,
+      "step": 13234
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019124129466266187,
+      "loss": 2.243,
+      "step": 13235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019129751162187556,
+      "loss": 1.1782,
+      "step": 13236
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019135372230540153,
+      "loss": 2.0094,
+      "step": 13237
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001914099267046982,
+      "loss": 2.2151,
+      "step": 13238
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001914661248112244,
+      "loss": 2.7257,
+      "step": 13239
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001915223166164401,
+      "loss": 2.5346,
+      "step": 13240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019157850211180615,
+      "loss": 2.3772,
+      "step": 13241
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019163468128878465,
+      "loss": 1.9244,
+      "step": 13242
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019169085413883836,
+      "loss": 2.1961,
+      "step": 13243
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019174702065343107,
+      "loss": 2.5022,
+      "step": 13244
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019180318082402746,
+      "loss": 2.5811,
+      "step": 13245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019185933464209348,
+      "loss": 2.8101,
+      "step": 13246
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019191548209909572,
+      "loss": 2.4742,
+      "step": 13247
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001919716231865019,
+      "loss": 2.5315,
+      "step": 13248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019202775789578047,
+      "loss": 2.5214,
+      "step": 13249
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019208388621840133,
+      "loss": 2.824,
+      "step": 13250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019214000814583493,
+      "loss": 2.4932,
+      "step": 13251
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019219612366955282,
+      "loss": 2.5757,
+      "step": 13252
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001922522327810274,
+      "loss": 2.7737,
+      "step": 13253
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001923083354717324,
+      "loss": 2.2251,
+      "step": 13254
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019236443173314217,
+      "loss": 2.3792,
+      "step": 13255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019242052155673224,
+      "loss": 2.635,
+      "step": 13256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019247660493397882,
+      "loss": 2.8566,
+      "step": 13257
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019253268185635957,
+      "loss": 2.9569,
+      "step": 13258
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019258875231535286,
+      "loss": 2.4273,
+      "step": 13259
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019264481630243782,
+      "loss": 2.4844,
+      "step": 13260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019270087380909512,
+      "loss": 2.3352,
+      "step": 13261
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000192756924826806,
+      "loss": 2.3793,
+      "step": 13262
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001928129693470527,
+      "loss": 2.6299,
+      "step": 13263
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019286900736131848,
+      "loss": 1.8929,
+      "step": 13264
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019292503886108786,
+      "loss": 2.5902,
+      "step": 13265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001929810638378459,
+      "loss": 2.2972,
+      "step": 13266
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019303708228307926,
+      "loss": 2.2984,
+      "step": 13267
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019309309418827472,
+      "loss": 2.6892,
+      "step": 13268
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019314909954492093,
+      "loss": 2.285,
+      "step": 13269
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019320509834450689,
+      "loss": 2.7386,
+      "step": 13270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019326109057852324,
+      "loss": 2.6176,
+      "step": 13271
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019331707623846075,
+      "loss": 2.5171,
+      "step": 13272
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019337305531581206,
+      "loss": 2.2687,
+      "step": 13273
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019342902780207017,
+      "loss": 2.0221,
+      "step": 13274
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019348499368872972,
+      "loss": 3.0901,
+      "step": 13275
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019354095296728548,
+      "loss": 2.1681,
+      "step": 13276
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019359690562923402,
+      "loss": 3.1099,
+      "step": 13277
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019365285166607249,
+      "loss": 2.7483,
+      "step": 13278
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019370879106929949,
+      "loss": 2.771,
+      "step": 13279
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019376472383041376,
+      "loss": 2.4287,
+      "step": 13280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019382064994091604,
+      "loss": 2.7144,
+      "step": 13281
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001938765693923074,
+      "loss": 2.7709,
+      "step": 13282
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019393248217609035,
+      "loss": 2.2261,
+      "step": 13283
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001939883882837681,
+      "loss": 3.027,
+      "step": 13284
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019404428770684504,
+      "loss": 2.5239,
+      "step": 13285
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001941001804368264,
+      "loss": 1.786,
+      "step": 13286
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019415606646521882,
+      "loss": 2.398,
+      "step": 13287
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019421194578352955,
+      "loss": 2.4531,
+      "step": 13288
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000194267818383267,
+      "loss": 2.8176,
+      "step": 13289
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019432368425594053,
+      "loss": 2.5909,
+      "step": 13290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001943795433930608,
+      "loss": 2.4637,
+      "step": 13291
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019443539578613926,
+      "loss": 2.3694,
+      "step": 13292
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019449124142668816,
+      "loss": 2.4183,
+      "step": 13293
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019454708030622138,
+      "loss": 2.5392,
+      "step": 13294
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019460291241625336,
+      "loss": 2.7541,
+      "step": 13295
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019465873774829968,
+      "loss": 1.9105,
+      "step": 13296
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019471455629387684,
+      "loss": 2.8559,
+      "step": 13297
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019477036804450273,
+      "loss": 2.3275,
+      "step": 13298
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001948261729916958,
+      "loss": 2.1804,
+      "step": 13299
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019488197112697614,
+      "loss": 2.4706,
+      "step": 13300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019493776244186402,
+      "loss": 1.748,
+      "step": 13301
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001949935469278816,
+      "loss": 2.9283,
+      "step": 13302
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019504932457655144,
+      "loss": 2.6197,
+      "step": 13303
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001951050953793978,
+      "loss": 2.1342,
+      "step": 13304
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019516085932794511,
+      "loss": 3.1443,
+      "step": 13305
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019521661641371966,
+      "loss": 2.3926,
+      "step": 13306
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019527236662824824,
+      "loss": 2.3679,
+      "step": 13307
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001953281099630592,
+      "loss": 1.4199,
+      "step": 13308
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019538384640968118,
+      "loss": 2.6808,
+      "step": 13309
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019543957595964467,
+      "loss": 2.9086,
+      "step": 13310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019549529860448062,
+      "loss": 2.1902,
+      "step": 13311
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019555101433572155,
+      "loss": 2.6965,
+      "step": 13312
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019560672314490032,
+      "loss": 2.9948,
+      "step": 13313
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019566242502355163,
+      "loss": 2.1778,
+      "step": 13314
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019571811996321056,
+      "loss": 2.413,
+      "step": 13315
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019577380795541386,
+      "loss": 2.7323,
+      "step": 13316
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001958294889916989,
+      "loss": 3.2047,
+      "step": 13317
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019588516306360418,
+      "loss": 1.7786,
+      "step": 13318
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019594083016266918,
+      "loss": 2.2008,
+      "step": 13319
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019599649028043491,
+      "loss": 1.3362,
+      "step": 13320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019605214340844286,
+      "loss": 1.0191,
+      "step": 13321
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019610778953823581,
+      "loss": 2.3501,
+      "step": 13322
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019616342866135772,
+      "loss": 2.7073,
+      "step": 13323
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019621906076935352,
+      "loss": 2.7912,
+      "step": 13324
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019627468585376915,
+      "loss": 2.2501,
+      "step": 13325
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019633030390615156,
+      "loss": 2.3232,
+      "step": 13326
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019638591491804905,
+      "loss": 2.8574,
+      "step": 13327
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019644151888101077,
+      "loss": 2.6629,
+      "step": 13328
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019649711578658693,
+      "loss": 2.7366,
+      "step": 13329
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019655270562632875,
+      "loss": 2.6878,
+      "step": 13330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001966082883917889,
+      "loss": 2.39,
+      "step": 13331
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019666386407452078,
+      "loss": 2.6818,
+      "step": 13332
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019671943266607887,
+      "loss": 2.7588,
+      "step": 13333
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019677499415801872,
+      "loss": 2.3816,
+      "step": 13334
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019683054854189731,
+      "loss": 1.8518,
+      "step": 13335
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019688609580927234,
+      "loss": 1.8746,
+      "step": 13336
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001969416359517026,
+      "loss": 2.3167,
+      "step": 13337
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000196997168960748,
+      "loss": 2.9518,
+      "step": 13338
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019705269482796984,
+      "loss": 1.6343,
+      "step": 13339
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019710821354493004,
+      "loss": 2.3373,
+      "step": 13340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019716372510319196,
+      "loss": 2.9365,
+      "step": 13341
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019721922949431962,
+      "loss": 2.7181,
+      "step": 13342
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001972747267098788,
+      "loss": 2.7468,
+      "step": 13343
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019733021674143576,
+      "loss": 1.4595,
+      "step": 13344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019738569958055815,
+      "loss": 2.6112,
+      "step": 13345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001974411752188145,
+      "loss": 2.8604,
+      "step": 13346
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019749664364777472,
+      "loss": 2.709,
+      "step": 13347
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019755210485900972,
+      "loss": 2.3734,
+      "step": 13348
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001976075588440912,
+      "loss": 1.421,
+      "step": 13349
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019766300559459255,
+      "loss": 2.475,
+      "step": 13350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019771844510208772,
+      "loss": 2.2234,
+      "step": 13351
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019777387735815202,
+      "loss": 2.6571,
+      "step": 13352
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019782930235436166,
+      "loss": 2.4298,
+      "step": 13353
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019788472008229436,
+      "loss": 1.9637,
+      "step": 13354
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019794013053352842,
+      "loss": 2.3732,
+      "step": 13355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019799553369964374,
+      "loss": 2.2203,
+      "step": 13356
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019805092957222105,
+      "loss": 2.4035,
+      "step": 13357
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019810631814284215,
+      "loss": 3.1108,
+      "step": 13358
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019816169940308997,
+      "loss": 2.431,
+      "step": 13359
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019821707334454882,
+      "loss": 2.4578,
+      "step": 13360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001982724399588038,
+      "loss": 3.2075,
+      "step": 13361
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019832779923744127,
+      "loss": 2.941,
+      "step": 13362
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019838315117204854,
+      "loss": 2.8394,
+      "step": 13363
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019843849575421445,
+      "loss": 1.9687,
+      "step": 13364
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019849383297552848,
+      "loss": 2.1598,
+      "step": 13365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019854916282758148,
+      "loss": 2.014,
+      "step": 13366
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019860448530196524,
+      "loss": 2.0435,
+      "step": 13367
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000198659800390273,
+      "loss": 2.8854,
+      "step": 13368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001987151080840989,
+      "loss": 1.7133,
+      "step": 13369
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019877040837503813,
+      "loss": 2.3433,
+      "step": 13370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019882570125468702,
+      "loss": 2.3672,
+      "step": 13371
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988809867146434,
+      "loss": 2.6502,
+      "step": 13372
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019893626474650568,
+      "loss": 2.3862,
+      "step": 13373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019899153534187376,
+      "loss": 1.954,
+      "step": 13374
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001990467984923484,
+      "loss": 2.2294,
+      "step": 13375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019910205418953192,
+      "loss": 1.6498,
+      "step": 13376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019915730242502737,
+      "loss": 2.2762,
+      "step": 13377
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019921254319043907,
+      "loss": 1.7163,
+      "step": 13378
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019926777647737232,
+      "loss": 1.1904,
+      "step": 13379
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019932300227743405,
+      "loss": 2.6826,
+      "step": 13380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019937822058223182,
+      "loss": 2.9227,
+      "step": 13381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001994334313833743,
+      "loss": 2.2977,
+      "step": 13382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019948863467247186,
+      "loss": 3.2017,
+      "step": 13383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019954383044113548,
+      "loss": 2.1031,
+      "step": 13384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019959901868097743,
+      "loss": 2.3986,
+      "step": 13385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019965419938361106,
+      "loss": 2.8424,
+      "step": 13386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001997093725406512,
+      "loss": 2.7037,
+      "step": 13387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019976453814371333,
+      "loss": 2.7946,
+      "step": 13388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019981969618441467,
+      "loss": 2.7603,
+      "step": 13389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019987484665437278,
+      "loss": 3.102,
+      "step": 13390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019992998954520717,
+      "loss": 2.1368,
+      "step": 13391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019998512484853802,
+      "loss": 2.4204,
+      "step": 13392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020004025255598704,
+      "loss": 2.6372,
+      "step": 13393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002000953726591765,
+      "loss": 2.4803,
+      "step": 13394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020015048514973052,
+      "loss": 2.0951,
+      "step": 13395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020020559001927377,
+      "loss": 1.3621,
+      "step": 13396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020026068725943275,
+      "loss": 3.1247,
+      "step": 13397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020031577686183423,
+      "loss": 2.1332,
+      "step": 13398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000200370858818107,
+      "loss": 2.8038,
+      "step": 13399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020042593311988042,
+      "loss": 2.6658,
+      "step": 13400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020048099975878564,
+      "loss": 2.6605,
+      "step": 13401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020053605872645403,
+      "loss": 1.2499,
+      "step": 13402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000200591110014519,
+      "loss": 2.7926,
+      "step": 13403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020064615361461467,
+      "loss": 2.6276,
+      "step": 13404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020070118951837678,
+      "loss": 2.425,
+      "step": 13405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020075621771744137,
+      "loss": 2.3535,
+      "step": 13406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020081123820344664,
+      "loss": 2.4923,
+      "step": 13407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020086625096803128,
+      "loss": 2.9301,
+      "step": 13408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020092125600283558,
+      "loss": 2.6579,
+      "step": 13409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002009762532995007,
+      "loss": 1.7203,
+      "step": 13410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020103124284966915,
+      "loss": 1.4168,
+      "step": 13411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020108622464498434,
+      "loss": 2.7371,
+      "step": 13412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020114119867709147,
+      "loss": 2.3032,
+      "step": 13413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020119616493763625,
+      "loss": 2.4604,
+      "step": 13414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020125112341826584,
+      "loss": 1.8968,
+      "step": 13415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002013060741106288,
+      "loss": 2.3625,
+      "step": 13416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020136101700637458,
+      "loss": 2.7849,
+      "step": 13417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002014159520971538,
+      "loss": 2.0666,
+      "step": 13418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020147087937461838,
+      "loss": 2.3896,
+      "step": 13419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020152579883042155,
+      "loss": 2.2399,
+      "step": 13420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002015807104562175,
+      "loss": 2.6765,
+      "step": 13421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020163561424366172,
+      "loss": 2.617,
+      "step": 13422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020169051018441078,
+      "loss": 2.6548,
+      "step": 13423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002017453982701227,
+      "loss": 2.581,
+      "step": 13424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020180027849245647,
+      "loss": 2.4122,
+      "step": 13425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020185515084307232,
+      "loss": 2.458,
+      "step": 13426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020191001531363154,
+      "loss": 1.9981,
+      "step": 13427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002019648718957971,
+      "loss": 2.2833,
+      "step": 13428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020201972058123252,
+      "loss": 2.595,
+      "step": 13429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002020745613616032,
+      "loss": 2.9714,
+      "step": 13430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020212939422857492,
+      "loss": 2.3959,
+      "step": 13431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020218421917381546,
+      "loss": 2.4461,
+      "step": 13432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002022390361889932,
+      "loss": 2.2683,
+      "step": 13433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020229384526577847,
+      "loss": 2.7604,
+      "step": 13434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020234864639584173,
+      "loss": 2.4937,
+      "step": 13435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020240343957085563,
+      "loss": 1.5511,
+      "step": 13436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020245822478249343,
+      "loss": 2.6958,
+      "step": 13437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020251300202243002,
+      "loss": 2.6892,
+      "step": 13438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020256777128234123,
+      "loss": 2.4561,
+      "step": 13439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002026225325539041,
+      "loss": 2.4721,
+      "step": 13440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020267728582879687,
+      "loss": 2.8842,
+      "step": 13441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020273203109869925,
+      "loss": 1.1654,
+      "step": 13442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000202786768355292,
+      "loss": 2.3931,
+      "step": 13443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020284149759025686,
+      "loss": 2.4337,
+      "step": 13444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020289621879527732,
+      "loss": 2.6797,
+      "step": 13445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020295093196203763,
+      "loss": 2.2824,
+      "step": 13446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020300563708222345,
+      "loss": 2.1924,
+      "step": 13447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002030603341475215,
+      "loss": 2.7968,
+      "step": 13448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020311502314962007,
+      "loss": 2.7031,
+      "step": 13449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020316970408020842,
+      "loss": 2.2475,
+      "step": 13450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000203224376930977,
+      "loss": 2.4145,
+      "step": 13451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020327904169361753,
+      "loss": 2.3524,
+      "step": 13452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020333369835982316,
+      "loss": 1.8795,
+      "step": 13453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000203388346921288,
+      "loss": 2.416,
+      "step": 13454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020344298736970753,
+      "loss": 2.3579,
+      "step": 13455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002034976196967783,
+      "loss": 2.8685,
+      "step": 13456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020355224389419848,
+      "loss": 2.4346,
+      "step": 13457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020360685995366706,
+      "loss": 2.1786,
+      "step": 13458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020366146786688445,
+      "loss": 2.5357,
+      "step": 13459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020371606762555216,
+      "loss": 2.3671,
+      "step": 13460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020377065922137327,
+      "loss": 2.537,
+      "step": 13461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020382524264605186,
+      "loss": 2.3625,
+      "step": 13462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020387981789129318,
+      "loss": 2.4546,
+      "step": 13463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002039343849488037,
+      "loss": 1.9606,
+      "step": 13464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002039889438102915,
+      "loss": 2.8082,
+      "step": 13465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020404349446746561,
+      "loss": 1.8671,
+      "step": 13466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002040980369120363,
+      "loss": 2.1146,
+      "step": 13467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000204152571135715,
+      "loss": 2.3811,
+      "step": 13468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020420709713021485,
+      "loss": 2.5017,
+      "step": 13469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020426161488724972,
+      "loss": 2.5983,
+      "step": 13470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020431612439853486,
+      "loss": 2.391,
+      "step": 13471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020437062565578713,
+      "loss": 2.325,
+      "step": 13472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002044251186507242,
+      "loss": 2.2711,
+      "step": 13473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002044796033750651,
+      "loss": 2.6996,
+      "step": 13474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020453407982053015,
+      "loss": 2.2343,
+      "step": 13475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020458854797884116,
+      "loss": 2.9155,
+      "step": 13476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020464300784172075,
+      "loss": 2.4246,
+      "step": 13477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020469745940089338,
+      "loss": 2.0547,
+      "step": 13478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020475190264808402,
+      "loss": 2.1054,
+      "step": 13479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020480633757501965,
+      "loss": 2.8696,
+      "step": 13480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002048607641734279,
+      "loss": 2.7113,
+      "step": 13481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020491518243503836,
+      "loss": 2.5038,
+      "step": 13482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000204969592351581,
+      "loss": 2.5019,
+      "step": 13483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020502399391478788,
+      "loss": 2.9641,
+      "step": 13484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002050783871163917,
+      "loss": 1.9822,
+      "step": 13485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002051327719481271,
+      "loss": 2.424,
+      "step": 13486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020518714840172918,
+      "loss": 2.651,
+      "step": 13487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020524151646893505,
+      "loss": 2.237,
+      "step": 13488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020529587614148253,
+      "loss": 2.8894,
+      "step": 13489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020535022741111135,
+      "loss": 2.458,
+      "step": 13490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020540457026956168,
+      "loss": 2.3049,
+      "step": 13491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020545890470857574,
+      "loss": 2.8174,
+      "step": 13492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002055132307198965,
+      "loss": 2.6333,
+      "step": 13493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020556754829526879,
+      "loss": 2.6372,
+      "step": 13494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020562185742643784,
+      "loss": 2.2829,
+      "step": 13495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020567615810515108,
+      "loss": 2.406,
+      "step": 13496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020573045032315658,
+      "loss": 2.4039,
+      "step": 13497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020578473407220415,
+      "loss": 2.7296,
+      "step": 13498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020583900934404456,
+      "loss": 2.6708,
+      "step": 13499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020589327613043003,
+      "loss": 2.4936,
+      "step": 13500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002059475344231139,
+      "loss": 2.3664,
+      "step": 13501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020600178421385107,
+      "loss": 2.0142,
+      "step": 13502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020605602549439762,
+      "loss": 1.7075,
+      "step": 13503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002061102582565106,
+      "loss": 1.8818,
+      "step": 13504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020616448249194904,
+      "loss": 2.5923,
+      "step": 13505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020621869819247268,
+      "loss": 2.5527,
+      "step": 13506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020627290534984272,
+      "loss": 2.0953,
+      "step": 13507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020632710395582162,
+      "loss": 2.3891,
+      "step": 13508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020638129400217338,
+      "loss": 2.3998,
+      "step": 13509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020643547548066296,
+      "loss": 2.3446,
+      "step": 13510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020648964838305712,
+      "loss": 2.5307,
+      "step": 13511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002065438127011231,
+      "loss": 1.5849,
+      "step": 13512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020659796842663035,
+      "loss": 2.4092,
+      "step": 13513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020665211555134888,
+      "loss": 2.3423,
+      "step": 13514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020670625406705073,
+      "loss": 1.9139,
+      "step": 13515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002067603839655084,
+      "loss": 3.0815,
+      "step": 13516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020681450523849656,
+      "loss": 3.2488,
+      "step": 13517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020686861787779043,
+      "loss": 1.8836,
+      "step": 13518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020692272187516737,
+      "loss": 1.2366,
+      "step": 13519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020697681722240503,
+      "loss": 2.6693,
+      "step": 13520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002070309039112833,
+      "loss": 2.0962,
+      "step": 13521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020708498193358282,
+      "loss": 3.1233,
+      "step": 13522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020713905128108605,
+      "loss": 2.7951,
+      "step": 13523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000207193111945576,
+      "loss": 1.8663,
+      "step": 13524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020724716391883782,
+      "loss": 2.4334,
+      "step": 13525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020730120719265733,
+      "loss": 2.5897,
+      "step": 13526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020735524175882238,
+      "loss": 2.7919,
+      "step": 13527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020740926760912126,
+      "loss": 2.0338,
+      "step": 13528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020746328473534431,
+      "loss": 3.2278,
+      "step": 13529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002075172931292828,
+      "loss": 2.3964,
+      "step": 13530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020757129278272963,
+      "loss": 2.8763,
+      "step": 13531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020762528368747877,
+      "loss": 2.8291,
+      "step": 13532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002076792658353256,
+      "loss": 2.6072,
+      "step": 13533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020773323921806673,
+      "loss": 2.3505,
+      "step": 13534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002077872038275004,
+      "loss": 3.1532,
+      "step": 13535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020784115965542596,
+      "loss": 1.4194,
+      "step": 13536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020789510669364395,
+      "loss": 2.6158,
+      "step": 13537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020794904493395667,
+      "loss": 2.416,
+      "step": 13538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020800297436816744,
+      "loss": 2.9469,
+      "step": 13539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020805689498808094,
+      "loss": 2.5621,
+      "step": 13540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020811080678550308,
+      "loss": 2.3531,
+      "step": 13541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020816470975224162,
+      "loss": 2.1897,
+      "step": 13542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020821860388010508,
+      "loss": 2.5288,
+      "step": 13543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002082724891609036,
+      "loss": 2.47,
+      "step": 13544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020832636558644854,
+      "loss": 2.4661,
+      "step": 13545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020838023314855288,
+      "loss": 2.5678,
+      "step": 13546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020843409183903062,
+      "loss": 2.0452,
+      "step": 13547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002084879416496973,
+      "loss": 1.8398,
+      "step": 13548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020854178257236953,
+      "loss": 2.7061,
+      "step": 13549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020859561459886586,
+      "loss": 2.7557,
+      "step": 13550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002086494377210056,
+      "loss": 2.4354,
+      "step": 13551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020870325193060964,
+      "loss": 2.3698,
+      "step": 13552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020875705721950017,
+      "loss": 2.6619,
+      "step": 13553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020881085357950095,
+      "loss": 2.6775,
+      "step": 13554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020886464100243685,
+      "loss": 2.1846,
+      "step": 13555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020891841948013416,
+      "loss": 1.8449,
+      "step": 13556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020897218900442046,
+      "loss": 2.5157,
+      "step": 13557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020902594956712502,
+      "loss": 2.6949,
+      "step": 13558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020907970116007806,
+      "loss": 2.6986,
+      "step": 13559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020913344377511124,
+      "loss": 2.6796,
+      "step": 13560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020918717740405794,
+      "loss": 2.5096,
+      "step": 13561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020924090203875255,
+      "loss": 2.8254,
+      "step": 13562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020929461767103084,
+      "loss": 2.6141,
+      "step": 13563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020934832429272997,
+      "loss": 2.7104,
+      "step": 13564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020940202189568874,
+      "loss": 2.6099,
+      "step": 13565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020945571047174688,
+      "loss": 2.4937,
+      "step": 13566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020950939001274611,
+      "loss": 2.6383,
+      "step": 13567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002095630605105286,
+      "loss": 2.4846,
+      "step": 13568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020961672195693882,
+      "loss": 2.3449,
+      "step": 13569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020967037434382194,
+      "loss": 1.8847,
+      "step": 13570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020972401766302523,
+      "loss": 2.5493,
+      "step": 13571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020977765190639637,
+      "loss": 2.7193,
+      "step": 13572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020983127706578533,
+      "loss": 2.7025,
+      "step": 13573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002098848931330428,
+      "loss": 2.4777,
+      "step": 13574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020993850010002143,
+      "loss": 2.7991,
+      "step": 13575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00020999209795857473,
+      "loss": 2.3986,
+      "step": 13576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021004568670055792,
+      "loss": 3.276,
+      "step": 13577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021009926631782735,
+      "loss": 2.7956,
+      "step": 13578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021015283680224112,
+      "loss": 2.359,
+      "step": 13579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021020639814565842,
+      "loss": 2.5822,
+      "step": 13580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021025995033993987,
+      "loss": 2.0699,
+      "step": 13581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021031349337694748,
+      "loss": 2.1126,
+      "step": 13582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002103670272485449,
+      "loss": 2.399,
+      "step": 13583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002104205519465968,
+      "loss": 2.5931,
+      "step": 13584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021047406746296953,
+      "loss": 2.7486,
+      "step": 13585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021052757378953052,
+      "loss": 2.9723,
+      "step": 13586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000210581070918149,
+      "loss": 2.4144,
+      "step": 13587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021063455884069533,
+      "loss": 2.467,
+      "step": 13588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021068803754904135,
+      "loss": 2.6212,
+      "step": 13589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021074150703506017,
+      "loss": 2.6228,
+      "step": 13590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021079496729062657,
+      "loss": 2.1012,
+      "step": 13591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021084841830761652,
+      "loss": 2.3764,
+      "step": 13592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002109018600779073,
+      "loss": 2.6962,
+      "step": 13593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021095529259337802,
+      "loss": 2.7058,
+      "step": 13594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021100871584590883,
+      "loss": 2.8718,
+      "step": 13595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002110621298273813,
+      "loss": 2.5086,
+      "step": 13596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002111155345296784,
+      "loss": 2.1363,
+      "step": 13597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002111689299446849,
+      "loss": 1.1852,
+      "step": 13598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021122231606428632,
+      "loss": 2.1396,
+      "step": 13599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021127569288037043,
+      "loss": 3.1114,
+      "step": 13600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021132906038482543,
+      "loss": 2.5834,
+      "step": 13601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021138241856954171,
+      "loss": 2.7764,
+      "step": 13602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002114357674264107,
+      "loss": 2.6584,
+      "step": 13603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021148910694732562,
+      "loss": 2.4305,
+      "step": 13604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002115424371241804,
+      "loss": 2.8614,
+      "step": 13605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021159575794887122,
+      "loss": 1.7773,
+      "step": 13606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021164906941329502,
+      "loss": 3.3384,
+      "step": 13607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021170237150935082,
+      "loss": 2.6521,
+      "step": 13608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021175566422893812,
+      "loss": 2.4237,
+      "step": 13609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021180894756395893,
+      "loss": 2.5249,
+      "step": 13610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021186222150631576,
+      "loss": 3.1136,
+      "step": 13611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021191548604791342,
+      "loss": 2.4215,
+      "step": 13612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021196874118065712,
+      "loss": 2.5387,
+      "step": 13613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021202198689645449,
+      "loss": 1.8057,
+      "step": 13614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021207522318721385,
+      "loss": 2.4176,
+      "step": 13615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021212845004484567,
+      "loss": 2.3586,
+      "step": 13616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021218166746126096,
+      "loss": 2.859,
+      "step": 13617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021223487542837297,
+      "loss": 2.76,
+      "step": 13618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002122880739380959,
+      "loss": 2.3519,
+      "step": 13619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002123412629823457,
+      "loss": 2.4788,
+      "step": 13620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021239444255303957,
+      "loss": 2.6023,
+      "step": 13621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021244761264209617,
+      "loss": 2.5573,
+      "step": 13622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021250077324143547,
+      "loss": 2.0328,
+      "step": 13623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021255392434297934,
+      "loss": 2.7438,
+      "step": 13624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021260706593865056,
+      "loss": 2.6403,
+      "step": 13625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021266019802037358,
+      "loss": 2.6662,
+      "step": 13626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021271332058007443,
+      "loss": 2.7594,
+      "step": 13627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021276643360968044,
+      "loss": 2.8413,
+      "step": 13628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021281953710112033,
+      "loss": 2.1737,
+      "step": 13629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021287263104632424,
+      "loss": 2.833,
+      "step": 13630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021292571543722407,
+      "loss": 2.4079,
+      "step": 13631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021297879026575294,
+      "loss": 2.3371,
+      "step": 13632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021303185552384534,
+      "loss": 2.1961,
+      "step": 13633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002130849112034373,
+      "loss": 2.7694,
+      "step": 13634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002131379572964665,
+      "loss": 2.2502,
+      "step": 13635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002131909937948718,
+      "loss": 2.5558,
+      "step": 13636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021324402069059357,
+      "loss": 2.7952,
+      "step": 13637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021329703797557368,
+      "loss": 1.8184,
+      "step": 13638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021335004564175563,
+      "loss": 2.9215,
+      "step": 13639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021340304368108417,
+      "loss": 2.6184,
+      "step": 13640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002134560320855055,
+      "loss": 2.9148,
+      "step": 13641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002135090108469673,
+      "loss": 2.3173,
+      "step": 13642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021356197995741892,
+      "loss": 2.4631,
+      "step": 13643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000213614939408811,
+      "loss": 2.7041,
+      "step": 13644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021366788919309558,
+      "loss": 2.3729,
+      "step": 13645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021372082930222618,
+      "loss": 2.8274,
+      "step": 13646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021377375972815812,
+      "loss": 2.2214,
+      "step": 13647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002138266804628477,
+      "loss": 2.2085,
+      "step": 13648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002138795914982533,
+      "loss": 2.2337,
+      "step": 13649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002139324928263339,
+      "loss": 2.2687,
+      "step": 13650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021398538443905083,
+      "loss": 2.1826,
+      "step": 13651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021403826632836632,
+      "loss": 2.2964,
+      "step": 13652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021409113848624447,
+      "loss": 2.9309,
+      "step": 13653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021414400090465064,
+      "loss": 2.1802,
+      "step": 13654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021419685357555162,
+      "loss": 2.626,
+      "step": 13655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002142496964909157,
+      "loss": 2.2509,
+      "step": 13656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021430252964271292,
+      "loss": 2.218,
+      "step": 13657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021435535302291456,
+      "loss": 2.8639,
+      "step": 13658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021440816662349324,
+      "loss": 2.8741,
+      "step": 13659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002144609704364235,
+      "loss": 2.4276,
+      "step": 13660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021451376445368105,
+      "loss": 2.5094,
+      "step": 13661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002145665486672431,
+      "loss": 2.6382,
+      "step": 13662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002146193230690883,
+      "loss": 2.1567,
+      "step": 13663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021467208765119718,
+      "loss": 2.9138,
+      "step": 13664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021472484240555133,
+      "loss": 2.3778,
+      "step": 13665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021477758732413402,
+      "loss": 2.2775,
+      "step": 13666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002148303223989298,
+      "loss": 2.2822,
+      "step": 13667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021488304762192522,
+      "loss": 2.1159,
+      "step": 13668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021493576298510782,
+      "loss": 2.5096,
+      "step": 13669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021498846848046684,
+      "loss": 2.4997,
+      "step": 13670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021504116409999292,
+      "loss": 2.6972,
+      "step": 13671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002150938498356785,
+      "loss": 2.0155,
+      "step": 13672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021514652567951716,
+      "loss": 2.4499,
+      "step": 13673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021519919162350413,
+      "loss": 1.3588,
+      "step": 13674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021525184765963605,
+      "loss": 2.6536,
+      "step": 13675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021530449377991137,
+      "loss": 2.8107,
+      "step": 13676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021535712997632978,
+      "loss": 2.7299,
+      "step": 13677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021540975624089248,
+      "loss": 2.2859,
+      "step": 13678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021546237256560207,
+      "loss": 2.1074,
+      "step": 13679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021551497894246318,
+      "loss": 2.4787,
+      "step": 13680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021556757536348136,
+      "loss": 2.432,
+      "step": 13681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021562016182066388,
+      "loss": 2.3698,
+      "step": 13682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002156727383060197,
+      "loss": 2.5914,
+      "step": 13683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021572530481155913,
+      "loss": 2.3956,
+      "step": 13684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021577786132929394,
+      "loss": 2.2599,
+      "step": 13685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021583040785123737,
+      "loss": 2.2398,
+      "step": 13686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021588294436940462,
+      "loss": 2.9779,
+      "step": 13687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021593547087581173,
+      "loss": 2.5222,
+      "step": 13688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021598798736247707,
+      "loss": 1.9514,
+      "step": 13689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002160404938214196,
+      "loss": 2.4178,
+      "step": 13690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002160929902446606,
+      "loss": 2.4528,
+      "step": 13691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021614547662422237,
+      "loss": 2.5981,
+      "step": 13692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021619795295212927,
+      "loss": 2.5423,
+      "step": 13693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002162504192204064,
+      "loss": 2.2219,
+      "step": 13694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021630287542108114,
+      "loss": 2.8269,
+      "step": 13695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002163553215461819,
+      "loss": 1.9856,
+      "step": 13696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021640775758773918,
+      "loss": 2.7525,
+      "step": 13697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021646018353778412,
+      "loss": 2.8754,
+      "step": 13698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021651259938835034,
+      "loss": 2.3319,
+      "step": 13699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021656500513147228,
+      "loss": 2.312,
+      "step": 13700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021661740075918666,
+      "loss": 2.8755,
+      "step": 13701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021666978626353072,
+      "loss": 2.1012,
+      "step": 13702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021672216163654419,
+      "loss": 2.6114,
+      "step": 13703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002167745268702678,
+      "loss": 2.5765,
+      "step": 13704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021682688195674422,
+      "loss": 2.8143,
+      "step": 13705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021687922688801704,
+      "loss": 2.5891,
+      "step": 13706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021693156165613207,
+      "loss": 1.7242,
+      "step": 13707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021698388625313617,
+      "loss": 2.4039,
+      "step": 13708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021703620067107813,
+      "loss": 2.267,
+      "step": 13709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021708850490200805,
+      "loss": 2.6552,
+      "step": 13710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021714079893797762,
+      "loss": 2.5989,
+      "step": 13711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002171930827710399,
+      "loss": 2.1951,
+      "step": 13712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021724535639325006,
+      "loss": 2.2532,
+      "step": 13713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021729761979666416,
+      "loss": 2.6664,
+      "step": 13714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021734987297334014,
+      "loss": 1.8226,
+      "step": 13715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021740211591533758,
+      "loss": 2.6432,
+      "step": 13716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021745434861471748,
+      "loss": 2.3004,
+      "step": 13717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021750657106354233,
+      "loss": 2.4746,
+      "step": 13718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021755878325387617,
+      "loss": 2.7408,
+      "step": 13719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021761098517778494,
+      "loss": 2.9206,
+      "step": 13720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021766317682733562,
+      "loss": 2.1147,
+      "step": 13721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002177153581945974,
+      "loss": 2.4095,
+      "step": 13722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021776752927164014,
+      "loss": 2.5014,
+      "step": 13723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021781969005053614,
+      "loss": 2.1554,
+      "step": 13724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002178718405233587,
+      "loss": 1.8494,
+      "step": 13725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002179239806821832,
+      "loss": 2.4856,
+      "step": 13726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021797611051908581,
+      "loss": 2.7532,
+      "step": 13727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002180282300261451,
+      "loss": 2.5629,
+      "step": 13728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021808033919544061,
+      "loss": 2.4677,
+      "step": 13729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021813243801905398,
+      "loss": 2.0532,
+      "step": 13730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021818452648906774,
+      "loss": 2.357,
+      "step": 13731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021823660459756663,
+      "loss": 2.5709,
+      "step": 13732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002182886723366366,
+      "loss": 2.3516,
+      "step": 13733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021834072969836553,
+      "loss": 3.2255,
+      "step": 13734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021839277667484223,
+      "loss": 2.5602,
+      "step": 13735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021844481325815782,
+      "loss": 2.2443,
+      "step": 13736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021849683944040444,
+      "loss": 1.9984,
+      "step": 13737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021854885521367635,
+      "loss": 2.6394,
+      "step": 13738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021860086057006874,
+      "loss": 2.2673,
+      "step": 13739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021865285550167897,
+      "loss": 2.4659,
+      "step": 13740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021870484000060552,
+      "loss": 3.0514,
+      "step": 13741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021875681405894893,
+      "loss": 2.922,
+      "step": 13742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021880877766881096,
+      "loss": 2.6044,
+      "step": 13743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002188607308222951,
+      "loss": 2.4943,
+      "step": 13744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021891267351150623,
+      "loss": 2.1879,
+      "step": 13745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021896460572855126,
+      "loss": 2.3911,
+      "step": 13746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021901652746553825,
+      "loss": 2.4023,
+      "step": 13747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021906843871457703,
+      "loss": 2.4204,
+      "step": 13748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002191203394677791,
+      "loss": 2.8603,
+      "step": 13749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021917222971725749,
+      "loss": 2.3334,
+      "step": 13750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002192241094551267,
+      "loss": 2.5446,
+      "step": 13751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021927597867350292,
+      "loss": 2.4517,
+      "step": 13752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021932783736450413,
+      "loss": 2.3915,
+      "step": 13753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002193796855202496,
+      "loss": 2.1068,
+      "step": 13754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021943152313286037,
+      "loss": 2.577,
+      "step": 13755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021948335019445893,
+      "loss": 3.002,
+      "step": 13756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021953516669716974,
+      "loss": 2.7258,
+      "step": 13757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021958697263311846,
+      "loss": 2.3551,
+      "step": 13758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021963876799443253,
+      "loss": 2.4961,
+      "step": 13759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021969055277324092,
+      "loss": 2.7995,
+      "step": 13760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002197423269616744,
+      "loss": 2.3926,
+      "step": 13761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021979409055186518,
+      "loss": 2.2726,
+      "step": 13762
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021984584353594707,
+      "loss": 2.6644,
+      "step": 13763
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00021989758590605543,
+      "loss": 2.3858,
+      "step": 13764
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002199493176543276,
+      "loss": 2.6258,
+      "step": 13765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022000103877290212,
+      "loss": 2.3477,
+      "step": 13766
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002200527492539194,
+      "loss": 2.3233,
+      "step": 13767
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002201044490895211,
+      "loss": 1.7993,
+      "step": 13768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022015613827185105,
+      "loss": 2.7157,
+      "step": 13769
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022020781679305438,
+      "loss": 1.8904,
+      "step": 13770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002202594846452777,
+      "loss": 2.2963,
+      "step": 13771
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022031114182066965,
+      "loss": 2.929,
+      "step": 13772
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002203627883113801,
+      "loss": 2.2291,
+      "step": 13773
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022041442410956078,
+      "loss": 2.5202,
+      "step": 13774
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022046604920736483,
+      "loss": 2.8458,
+      "step": 13775
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022051766359694735,
+      "loss": 2.9614,
+      "step": 13776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022056926727046465,
+      "loss": 2.5705,
+      "step": 13777
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022062086022007528,
+      "loss": 2.4182,
+      "step": 13778
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002206724424379385,
+      "loss": 2.6327,
+      "step": 13779
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002207240139162161,
+      "loss": 2.4188,
+      "step": 13780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002207755746470709,
+      "loss": 2.6193,
+      "step": 13781
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022082712462266797,
+      "loss": 2.5938,
+      "step": 13782
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022087866383517312,
+      "loss": 2.7502,
+      "step": 13783
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022093019227675467,
+      "loss": 2.4595,
+      "step": 13784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022098170993958195,
+      "loss": 0.8651,
+      "step": 13785
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002210332168158266,
+      "loss": 2.4031,
+      "step": 13786
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000221084712897661,
+      "loss": 2.7679,
+      "step": 13787
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022113619817725993,
+      "loss": 2.6872,
+      "step": 13788
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022118767264679942,
+      "loss": 2.6975,
+      "step": 13789
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022123913629845752,
+      "loss": 2.6355,
+      "step": 13790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022129058912441328,
+      "loss": 2.2648,
+      "step": 13791
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002213420311168481,
+      "loss": 2.4373,
+      "step": 13792
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022139346226794445,
+      "loss": 2.8123,
+      "step": 13793
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022144488256988697,
+      "loss": 2.69,
+      "step": 13794
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022149629201486152,
+      "loss": 1.9161,
+      "step": 13795
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022154769059505588,
+      "loss": 2.475,
+      "step": 13796
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022159907830265915,
+      "loss": 2.8222,
+      "step": 13797
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022165045512986258,
+      "loss": 2.8144,
+      "step": 13798
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002217018210688587,
+      "loss": 3.0301,
+      "step": 13799
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022175317611184176,
+      "loss": 2.8295,
+      "step": 13800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022180452025100764,
+      "loss": 2.5873,
+      "step": 13801
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002218558534785541,
+      "loss": 2.8108,
+      "step": 13802
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022190717578668033,
+      "loss": 3.1611,
+      "step": 13803
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002219584871675871,
+      "loss": 2.3045,
+      "step": 13804
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022200978761347727,
+      "loss": 2.6045,
+      "step": 13805
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002220610771165549,
+      "loss": 2.3983,
+      "step": 13806
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022211235566902595,
+      "loss": 2.4343,
+      "step": 13807
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002221636232630978,
+      "loss": 1.8748,
+      "step": 13808
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022221487989098,
+      "loss": 2.7552,
+      "step": 13809
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002222661255448831,
+      "loss": 2.6991,
+      "step": 13810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022231736021702015,
+      "loss": 2.7195,
+      "step": 13811
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002223685838996048,
+      "loss": 2.7421,
+      "step": 13812
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022241979658485335,
+      "loss": 2.2043,
+      "step": 13813
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022247099826498313,
+      "loss": 2.3175,
+      "step": 13814
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022252218893221374,
+      "loss": 2.8469,
+      "step": 13815
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022257336857876568,
+      "loss": 2.889,
+      "step": 13816
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022262453719686182,
+      "loss": 2.98,
+      "step": 13817
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022267569477872626,
+      "loss": 2.7287,
+      "step": 13818
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022272684131658524,
+      "loss": 1.3939,
+      "step": 13819
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022277797680266595,
+      "loss": 1.4749,
+      "step": 13820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022282910122919808,
+      "loss": 1.6704,
+      "step": 13821
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022288021458841231,
+      "loss": 2.5786,
+      "step": 13822
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022293131687254177,
+      "loss": 2.2729,
+      "step": 13823
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002229824080738202,
+      "loss": 2.2519,
+      "step": 13824
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022303348818448405,
+      "loss": 2.3917,
+      "step": 13825
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022308455719677087,
+      "loss": 2.3145,
+      "step": 13826
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022313561510292032,
+      "loss": 3.1489,
+      "step": 13827
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022318666189517308,
+      "loss": 2.423,
+      "step": 13828
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002232376975657723,
+      "loss": 2.2547,
+      "step": 13829
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022328872210696218,
+      "loss": 2.0532,
+      "step": 13830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002233397355109891,
+      "loss": 2.0366,
+      "step": 13831
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022339073777010086,
+      "loss": 2.6836,
+      "step": 13832
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000223441728876547,
+      "loss": 2.3503,
+      "step": 13833
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022349270882257862,
+      "loss": 2.9763,
+      "step": 13834
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022354367760044893,
+      "loss": 2.5134,
+      "step": 13835
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022359463520241243,
+      "loss": 2.7091,
+      "step": 13836
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022364558162072534,
+      "loss": 2.6379,
+      "step": 13837
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002236965168476459,
+      "loss": 1.7651,
+      "step": 13838
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022374744087543381,
+      "loss": 2.3938,
+      "step": 13839
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022379835369635046,
+      "loss": 2.6849,
+      "step": 13840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002238492553026589,
+      "loss": 2.2808,
+      "step": 13841
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022390014568662416,
+      "loss": 2.6999,
+      "step": 13842
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022395102484051274,
+      "loss": 2.4145,
+      "step": 13843
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002240018927565928,
+      "loss": 2.5682,
+      "step": 13844
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022405274942713434,
+      "loss": 2.7075,
+      "step": 13845
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022410359484440915,
+      "loss": 2.8283,
+      "step": 13846
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022415442900069053,
+      "loss": 2.2687,
+      "step": 13847
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022420525188825358,
+      "loss": 2.5168,
+      "step": 13848
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000224256063499375,
+      "loss": 3.2754,
+      "step": 13849
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022430686382633356,
+      "loss": 2.7179,
+      "step": 13850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022435765286140934,
+      "loss": 2.4552,
+      "step": 13851
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022440843059688432,
+      "loss": 2.2565,
+      "step": 13852
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022445919702504202,
+      "loss": 1.8832,
+      "step": 13853
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002245099521381681,
+      "loss": 2.5317,
+      "step": 13854
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022456069592854955,
+      "loss": 2.5302,
+      "step": 13855
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002246114283884752,
+      "loss": 2.2566,
+      "step": 13856
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002246621495102355,
+      "loss": 2.291,
+      "step": 13857
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022471285928612287,
+      "loss": 2.584,
+      "step": 13858
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002247635577084313,
+      "loss": 2.5117,
+      "step": 13859
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022481424476945646,
+      "loss": 2.5304,
+      "step": 13860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022486492046149567,
+      "loss": 2.6923,
+      "step": 13861
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022491558477684837,
+      "loss": 2.5306,
+      "step": 13862
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022496623770781535,
+      "loss": 2.4313,
+      "step": 13863
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022501687924669913,
+      "loss": 2.5874,
+      "step": 13864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022506750938580428,
+      "loss": 2.8384,
+      "step": 13865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002251181281174368,
+      "loss": 2.3496,
+      "step": 13866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022516873543390446,
+      "loss": 2.2336,
+      "step": 13867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022521933132751694,
+      "loss": 2.8805,
+      "step": 13868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022526991579058554,
+      "loss": 2.3381,
+      "step": 13869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022532048881542316,
+      "loss": 1.1465,
+      "step": 13870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022537105039434474,
+      "loss": 2.1797,
+      "step": 13871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022542160051966677,
+      "loss": 2.1711,
+      "step": 13872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002254721391837075,
+      "loss": 2.6293,
+      "step": 13873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022552266637878676,
+      "loss": 2.4441,
+      "step": 13874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022557318209722656,
+      "loss": 2.4592,
+      "step": 13875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022562368633135028,
+      "loss": 2.3699,
+      "step": 13876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002256741790734831,
+      "loss": 2.8797,
+      "step": 13877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022572466031595196,
+      "loss": 2.3901,
+      "step": 13878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022577513005108576,
+      "loss": 2.9501,
+      "step": 13879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022582558827121492,
+      "loss": 2.5387,
+      "step": 13880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002258760349686716,
+      "loss": 2.1836,
+      "step": 13881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002259264701357897,
+      "loss": 2.4609,
+      "step": 13882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022597689376490518,
+      "loss": 2.3314,
+      "step": 13883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022602730584835536,
+      "loss": 1.8131,
+      "step": 13884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002260777063784795,
+      "loss": 2.0272,
+      "step": 13885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002261280953476185,
+      "loss": 1.101,
+      "step": 13886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022617847274811537,
+      "loss": 2.4081,
+      "step": 13887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022622883857231437,
+      "loss": 1.9492,
+      "step": 13888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022627919281256192,
+      "loss": 2.4557,
+      "step": 13889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022632953546120583,
+      "loss": 2.2012,
+      "step": 13890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022637986651059615,
+      "loss": 2.5581,
+      "step": 13891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022643018595308435,
+      "loss": 2.435,
+      "step": 13892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022648049378102354,
+      "loss": 2.3589,
+      "step": 13893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022653078998676916,
+      "loss": 2.5904,
+      "step": 13894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022658107456267781,
+      "loss": 2.7614,
+      "step": 13895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022663134750110816,
+      "loss": 2.2786,
+      "step": 13896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022668160879442048,
+      "loss": 2.8519,
+      "step": 13897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022673185843497716,
+      "loss": 2.2774,
+      "step": 13898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022678209641514187,
+      "loss": 3.2088,
+      "step": 13899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022683232272728064,
+      "loss": 2.6343,
+      "step": 13900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022688253736376048,
+      "loss": 1.8762,
+      "step": 13901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000226932740316951,
+      "loss": 2.6955,
+      "step": 13902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022698293157922297,
+      "loss": 2.1446,
+      "step": 13903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022703311114294952,
+      "loss": 2.663,
+      "step": 13904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002270832790005048,
+      "loss": 2.549,
+      "step": 13905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022713343514426545,
+      "loss": 2.8409,
+      "step": 13906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002271835795666094,
+      "loss": 2.071,
+      "step": 13907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022723371225991687,
+      "loss": 2.5448,
+      "step": 13908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022728383321656914,
+      "loss": 2.8646,
+      "step": 13909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022733394242895006,
+      "loss": 2.9332,
+      "step": 13910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022738403988944455,
+      "loss": 2.6971,
+      "step": 13911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002274341255904401,
+      "loss": 2.1401,
+      "step": 13912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00022748419952432506,
+      "loss": 2.9698,
+      "step": 13913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002275342616834904,
+      "loss": 2.3001,
+      "step": 13914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002275843120603283,
+      "loss": 2.2534,
+      "step": 13915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022763435064723322,
+      "loss": 2.1889,
+      "step": 13916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022768437743660085,
+      "loss": 2.6741,
+      "step": 13917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002277343924208292,
+      "loss": 2.2807,
+      "step": 13918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002277843955923177,
+      "loss": 2.3175,
+      "step": 13919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022783438694346788,
+      "loss": 2.7658,
+      "step": 13920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022788436646668288,
+      "loss": 2.6051,
+      "step": 13921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002279343341543676,
+      "loss": 2.664,
+      "step": 13922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002279842899989287,
+      "loss": 2.3671,
+      "step": 13923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022803423399277503,
+      "loss": 2.6622,
+      "step": 13924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002280841661283168,
+      "loss": 2.6802,
+      "step": 13925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022813408639796605,
+      "loss": 2.1243,
+      "step": 13926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022818399479413706,
+      "loss": 2.0622,
+      "step": 13927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002282338913092454,
+      "loss": 2.3448,
+      "step": 13928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022828377593570872,
+      "loss": 2.2548,
+      "step": 13929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022833364866594626,
+      "loss": 2.3371,
+      "step": 13930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022838350949237944,
+      "loss": 2.213,
+      "step": 13931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022843335840743115,
+      "loss": 2.3268,
+      "step": 13932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002284831954035263,
+      "loss": 2.2489,
+      "step": 13933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022853302047309127,
+      "loss": 1.9706,
+      "step": 13934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022858283360855478,
+      "loss": 2.582,
+      "step": 13935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022863263480234702,
+      "loss": 2.0572,
+      "step": 13936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002286824240469,
+      "loss": 2.6673,
+      "step": 13937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002287322013346475,
+      "loss": 2.0871,
+      "step": 13938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022878196665802547,
+      "loss": 2.385,
+      "step": 13939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002288317200094712,
+      "loss": 2.2826,
+      "step": 13940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022888146138142434,
+      "loss": 2.78,
+      "step": 13941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022893119076632562,
+      "loss": 2.6653,
+      "step": 13942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022898090815661834,
+      "loss": 2.6164,
+      "step": 13943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022903061354474712,
+      "loss": 2.4389,
+      "step": 13944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002290803069231589,
+      "loss": 2.5823,
+      "step": 13945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022912998828430165,
+      "loss": 2.3877,
+      "step": 13946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022917965762062601,
+      "loss": 1.9855,
+      "step": 13947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022922931492458393,
+      "loss": 2.5608,
+      "step": 13948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022927896018862956,
+      "loss": 2.6771,
+      "step": 13949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002293285934052183,
+      "loss": 2.6829,
+      "step": 13950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022937821456680806,
+      "loss": 2.3491,
+      "step": 13951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022942782366585802,
+      "loss": 2.8398,
+      "step": 13952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002294774206948297,
+      "loss": 2.3961,
+      "step": 13953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022952700564618603,
+      "loss": 1.9359,
+      "step": 13954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022957657851239204,
+      "loss": 2.4744,
+      "step": 13955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022962613928591432,
+      "loss": 2.046,
+      "step": 13956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002296756879592216,
+      "loss": 2.698,
+      "step": 13957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022972522452478439,
+      "loss": 2.3129,
+      "step": 13958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022977474897507478,
+      "loss": 2.3061,
+      "step": 13959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002298242613025671,
+      "loss": 2.6741,
+      "step": 13960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002298737614997372,
+      "loss": 2.5395,
+      "step": 13961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00022992324955906294,
+      "loss": 2.4768,
+      "step": 13962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002299727254730238,
+      "loss": 2.2462,
+      "step": 13963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023002218923410156,
+      "loss": 2.5377,
+      "step": 13964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002300716408347794,
+      "loss": 2.4574,
+      "step": 13965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023012108026754257,
+      "loss": 2.4093,
+      "step": 13966
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023017050752487795,
+      "loss": 2.3601,
+      "step": 13967
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023021992259927463,
+      "loss": 1.8484,
+      "step": 13968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023026932548322337,
+      "loss": 2.7439,
+      "step": 13969
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023031871616921665,
+      "loss": 2.7996,
+      "step": 13970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023036809464974887,
+      "loss": 2.5493,
+      "step": 13971
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002304174609173165,
+      "loss": 2.5555,
+      "step": 13972
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023046681496441767,
+      "loss": 2.6044,
+      "step": 13973
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023051615678355231,
+      "loss": 2.5193,
+      "step": 13974
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023056548636722225,
+      "loss": 2.7853,
+      "step": 13975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023061480370793142,
+      "loss": 2.165,
+      "step": 13976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002306641087981853,
+      "loss": 2.4435,
+      "step": 13977
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002307134016304914,
+      "loss": 2.3124,
+      "step": 13978
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023076268219735885,
+      "loss": 2.3855,
+      "step": 13979
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023081195049129915,
+      "loss": 1.2146,
+      "step": 13980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002308612065048251,
+      "loss": 2.8219,
+      "step": 13981
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002309104502304518,
+      "loss": 3.131,
+      "step": 13982
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023095968166069573,
+      "loss": 3.0148,
+      "step": 13983
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002310089007880759,
+      "loss": 1.7415,
+      "step": 13984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023105810760511272,
+      "loss": 2.8183,
+      "step": 13985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023110730210432838,
+      "loss": 2.0754,
+      "step": 13986
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023115648427824738,
+      "loss": 1.9765,
+      "step": 13987
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023120565411939584,
+      "loss": 2.2141,
+      "step": 13988
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002312548116203017,
+      "loss": 2.6526,
+      "step": 13989
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023130395677349475,
+      "loss": 2.6338,
+      "step": 13990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002313530895715069,
+      "loss": 2.9028,
+      "step": 13991
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023140221000687172,
+      "loss": 3.0247,
+      "step": 13992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023145131807212496,
+      "loss": 2.7641,
+      "step": 13993
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023150041375980356,
+      "loss": 1.9279,
+      "step": 13994
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002315494970624472,
+      "loss": 3.001,
+      "step": 13995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023159856797259678,
+      "loss": 2.839,
+      "step": 13996
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023164762648279568,
+      "loss": 1.5906,
+      "step": 13997
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023169667258558836,
+      "loss": 2.4832,
+      "step": 13998
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023174570627352203,
+      "loss": 2.6421,
+      "step": 13999
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023179472753914512,
+      "loss": 2.453,
+      "step": 14000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023184373637500863,
+      "loss": 2.5137,
+      "step": 14001
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023189273277366446,
+      "loss": 2.6947,
+      "step": 14002
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023194171672766736,
+      "loss": 2.5605,
+      "step": 14003
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023199068822957344,
+      "loss": 2.7305,
+      "step": 14004
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023203964727194112,
+      "loss": 2.7062,
+      "step": 14005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023208859384733004,
+      "loss": 2.587,
+      "step": 14006
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002321375279483024,
+      "loss": 2.7933,
+      "step": 14007
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023218644956742184,
+      "loss": 2.6055,
+      "step": 14008
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023223535869725437,
+      "loss": 1.8091,
+      "step": 14009
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023228425533036745,
+      "loss": 2.1168,
+      "step": 14010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023233313945933065,
+      "loss": 2.6546,
+      "step": 14011
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023238201107671525,
+      "loss": 2.5319,
+      "step": 14012
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023243087017509482,
+      "loss": 2.4103,
+      "step": 14013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023247971674704446,
+      "loss": 2.4192,
+      "step": 14014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023252855078514123,
+      "loss": 2.3806,
+      "step": 14015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023257737228196435,
+      "loss": 2.8493,
+      "step": 14016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002326261812300947,
+      "loss": 2.451,
+      "step": 14017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023267497762211509,
+      "loss": 2.2329,
+      "step": 14018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023272376145061025,
+      "loss": 2.4129,
+      "step": 14019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023277253270816698,
+      "loss": 2.8036,
+      "step": 14020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023282129138737365,
+      "loss": 1.9992,
+      "step": 14021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023287003748082117,
+      "loss": 2.1962,
+      "step": 14022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002329187709811014,
+      "loss": 2.7578,
+      "step": 14023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000232967491880809,
+      "loss": 2.4878,
+      "step": 14024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023301620017254004,
+      "loss": 2.6418,
+      "step": 14025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023306489584889298,
+      "loss": 2.7257,
+      "step": 14026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023311357890246746,
+      "loss": 3.0809,
+      "step": 14027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023316224932586575,
+      "loss": 3.0195,
+      "step": 14028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023321090711169156,
+      "loss": 3.2128,
+      "step": 14029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023325955225255102,
+      "loss": 2.5861,
+      "step": 14030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002333081847410515,
+      "loss": 2.7487,
+      "step": 14031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023335680456980295,
+      "loss": 2.7743,
+      "step": 14032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002334054117314167,
+      "loss": 2.6266,
+      "step": 14033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023345400621850668,
+      "loss": 2.2113,
+      "step": 14034
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023350258802368786,
+      "loss": 1.6317,
+      "step": 14035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023355115713957793,
+      "loss": 2.6346,
+      "step": 14036
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000233599713558796,
+      "loss": 2.5822,
+      "step": 14037
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002336482572739636,
+      "loss": 2.5511,
+      "step": 14038
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023369678827770343,
+      "loss": 1.5561,
+      "step": 14039
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023374530656264093,
+      "loss": 1.858,
+      "step": 14040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023379381212140297,
+      "loss": 3.1379,
+      "step": 14041
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023384230494661858,
+      "loss": 2.2446,
+      "step": 14042
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023389078503091866,
+      "loss": 2.5038,
+      "step": 14043
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023393925236693604,
+      "loss": 2.6367,
+      "step": 14044
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023398770694730534,
+      "loss": 2.7457,
+      "step": 14045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023403614876466345,
+      "loss": 2.2184,
+      "step": 14046
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023408457781164894,
+      "loss": 2.1726,
+      "step": 14047
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023413299408090233,
+      "loss": 2.6588,
+      "step": 14048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023418139756506624,
+      "loss": 2.7037,
+      "step": 14049
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023422978825678517,
+      "loss": 2.7072,
+      "step": 14050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023427816614870543,
+      "loss": 2.6344,
+      "step": 14051
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023432653123347535,
+      "loss": 2.7247,
+      "step": 14052
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023437488350374537,
+      "loss": 1.8366,
+      "step": 14053
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023442322295216768,
+      "loss": 2.4591,
+      "step": 14054
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002344715495713965,
+      "loss": 2.2796,
+      "step": 14055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002345198633540878,
+      "loss": 1.1709,
+      "step": 14056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002345681642929,
+      "loss": 2.5621,
+      "step": 14057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023461645238049292,
+      "loss": 2.6104,
+      "step": 14058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023466472760952866,
+      "loss": 2.7654,
+      "step": 14059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023471298997267102,
+      "loss": 2.8671,
+      "step": 14060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023476123946258618,
+      "loss": 2.2852,
+      "step": 14061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023480947607194182,
+      "loss": 1.0192,
+      "step": 14062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002348576997934079,
+      "loss": 2.7646,
+      "step": 14063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023490591061965592,
+      "loss": 2.3756,
+      "step": 14064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023495410854336,
+      "loss": 2.8342,
+      "step": 14065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023500229355719564,
+      "loss": 2.3675,
+      "step": 14066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023505046565384057,
+      "loss": 2.6291,
+      "step": 14067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002350986248259743,
+      "loss": 2.1289,
+      "step": 14068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023514677106627855,
+      "loss": 2.6044,
+      "step": 14069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023519490436743693,
+      "loss": 2.1115,
+      "step": 14070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023524302472213485,
+      "loss": 2.5609,
+      "step": 14071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023529113212305975,
+      "loss": 2.503,
+      "step": 14072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023533922656290127,
+      "loss": 2.9134,
+      "step": 14073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023538730803435075,
+      "loss": 2.0052,
+      "step": 14074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002354353765301015,
+      "loss": 2.5092,
+      "step": 14075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023548343204284907,
+      "loss": 2.5475,
+      "step": 14076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023553147456529077,
+      "loss": 2.6954,
+      "step": 14077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023557950409012584,
+      "loss": 2.1617,
+      "step": 14078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023562752061005553,
+      "loss": 2.4909,
+      "step": 14079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023567552411778332,
+      "loss": 2.6552,
+      "step": 14080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002357235146060142,
+      "loss": 2.3309,
+      "step": 14081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023577149206745565,
+      "loss": 2.6531,
+      "step": 14082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002358194564948168,
+      "loss": 2.6417,
+      "step": 14083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023586740788080876,
+      "loss": 2.6089,
+      "step": 14084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002359153462181447,
+      "loss": 2.9279,
+      "step": 14085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023596327149953984,
+      "loss": 1.7677,
+      "step": 14086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023601118371771135,
+      "loss": 2.1701,
+      "step": 14087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023605908286537832,
+      "loss": 2.3069,
+      "step": 14088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023610696893526167,
+      "loss": 2.785,
+      "step": 14089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002361548419200848,
+      "loss": 2.828,
+      "step": 14090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023620270181257268,
+      "loss": 2.8173,
+      "step": 14091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023625054860545235,
+      "loss": 2.4218,
+      "step": 14092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023629838229145277,
+      "loss": 2.6183,
+      "step": 14093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023634620286330525,
+      "loss": 2.293,
+      "step": 14094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023639401031374269,
+      "loss": 2.4593,
+      "step": 14095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023644180463550015,
+      "loss": 2.1572,
+      "step": 14096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023648958582131457,
+      "loss": 1.8552,
+      "step": 14097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023653735386392516,
+      "loss": 2.6261,
+      "step": 14098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023658510875607287,
+      "loss": 2.7471,
+      "step": 14099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023663285049050077,
+      "loss": 2.6338,
+      "step": 14100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023668057905995376,
+      "loss": 2.6597,
+      "step": 14101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000236728294457179,
+      "loss": 2.1177,
+      "step": 14102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023677599667492558,
+      "loss": 2.5934,
+      "step": 14103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023682368570594438,
+      "loss": 2.6958,
+      "step": 14104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023687136154298843,
+      "loss": 2.2609,
+      "step": 14105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002369190241788129,
+      "loss": 2.1994,
+      "step": 14106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023696667360617485,
+      "loss": 2.9388,
+      "step": 14107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023701430981783314,
+      "loss": 2.233,
+      "step": 14108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023706193280654907,
+      "loss": 2.1009,
+      "step": 14109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002371095425650856,
+      "loss": 2.2116,
+      "step": 14110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023715713908620785,
+      "loss": 2.8943,
+      "step": 14111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002372047223626828,
+      "loss": 1.7258,
+      "step": 14112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023725229238727972,
+      "loss": 2.2873,
+      "step": 14113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002372998491527696,
+      "loss": 2.4672,
+      "step": 14114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023734739265192587,
+      "loss": 2.1892,
+      "step": 14115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023739492287752325,
+      "loss": 2.964,
+      "step": 14116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023744243982233917,
+      "loss": 1.746,
+      "step": 14117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023748994347915273,
+      "loss": 2.269,
+      "step": 14118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023753743384074538,
+      "loss": 2.4612,
+      "step": 14119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023758491089989994,
+      "loss": 2.3749,
+      "step": 14120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023763237464940196,
+      "loss": 1.3335,
+      "step": 14121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023767982508203852,
+      "loss": 1.9419,
+      "step": 14122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023772726219059925,
+      "loss": 1.1654,
+      "step": 14123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023777468596787504,
+      "loss": 2.3758,
+      "step": 14124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023782209640665954,
+      "loss": 2.3512,
+      "step": 14125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002378694934997479,
+      "loss": 2.567,
+      "step": 14126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023791687723993788,
+      "loss": 2.5208,
+      "step": 14127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002379642476200285,
+      "loss": 2.8135,
+      "step": 14128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002380116046328215,
+      "loss": 2.8877,
+      "step": 14129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023805894827112013,
+      "loss": 2.0849,
+      "step": 14130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023810627852773018,
+      "loss": 2.6203,
+      "step": 14131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023815359539545907,
+      "loss": 2.7309,
+      "step": 14132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002382008988671165,
+      "loss": 1.7916,
+      "step": 14133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023824818893551384,
+      "loss": 2.0502,
+      "step": 14134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023829546559346506,
+      "loss": 2.9846,
+      "step": 14135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023834272883378575,
+      "loss": 1.8648,
+      "step": 14136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002383899786492935,
+      "loss": 2.4102,
+      "step": 14137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023843721503280838,
+      "loss": 2.8034,
+      "step": 14138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023848443797715204,
+      "loss": 2.5956,
+      "step": 14139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002385316474751484,
+      "loss": 2.4577,
+      "step": 14140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002385788435196232,
+      "loss": 3.0901,
+      "step": 14141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023862602610340465,
+      "loss": 2.4839,
+      "step": 14142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023867319521932268,
+      "loss": 2.9849,
+      "step": 14143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023872035086020923,
+      "loss": 2.381,
+      "step": 14144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002387674930188984,
+      "loss": 2.6984,
+      "step": 14145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023881462168822651,
+      "loss": 2.3444,
+      "step": 14146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023886173686103163,
+      "loss": 2.153,
+      "step": 14147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000238908838530154,
+      "loss": 2.9577,
+      "step": 14148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002389559266884358,
+      "loss": 2.8384,
+      "step": 14149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023900300132872167,
+      "loss": 1.1386,
+      "step": 14150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002390500624438578,
+      "loss": 2.4177,
+      "step": 14151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002390971100266927,
+      "loss": 2.234,
+      "step": 14152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002391441440700768,
+      "loss": 2.6729,
+      "step": 14153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023919116456686286,
+      "loss": 2.5246,
+      "step": 14154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002392381715099053,
+      "loss": 2.3421,
+      "step": 14155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023928516489206116,
+      "loss": 2.2273,
+      "step": 14156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023933214470618871,
+      "loss": 2.5793,
+      "step": 14157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023937911094514908,
+      "loss": 2.5344,
+      "step": 14158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000239426063601805,
+      "loss": 2.7146,
+      "step": 14159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023947300266902166,
+      "loss": 2.4524,
+      "step": 14160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023951992813966576,
+      "loss": 2.5709,
+      "step": 14161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023956684000660657,
+      "loss": 2.06,
+      "step": 14162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023961373826271502,
+      "loss": 2.3573,
+      "step": 14163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023966062290086456,
+      "loss": 3.1843,
+      "step": 14164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023970749391393037,
+      "loss": 2.5199,
+      "step": 14165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023975435129478976,
+      "loss": 2.2615,
+      "step": 14166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023980119503632215,
+      "loss": 2.0128,
+      "step": 14167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002398480251314091,
+      "loss": 2.6388,
+      "step": 14168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023989484157293416,
+      "loss": 2.7608,
+      "step": 14169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023994164435378287,
+      "loss": 2.5568,
+      "step": 14170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00023998843346684307,
+      "loss": 2.4966,
+      "step": 14171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024003520890500456,
+      "loss": 2.7264,
+      "step": 14172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002400819706611591,
+      "loss": 2.4868,
+      "step": 14173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024012871872820061,
+      "loss": 2.6797,
+      "step": 14174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024017545309902534,
+      "loss": 2.8833,
+      "step": 14175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024022217376653125,
+      "loss": 2.359,
+      "step": 14176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024026888072361854,
+      "loss": 2.65,
+      "step": 14177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024031557396318943,
+      "loss": 2.3938,
+      "step": 14178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024036225347814842,
+      "loss": 2.2273,
+      "step": 14179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002404089192614019,
+      "loss": 2.0519,
+      "step": 14180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002404555713058584,
+      "loss": 2.4394,
+      "step": 14181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024050220960442843,
+      "loss": 3.214,
+      "step": 14182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002405488341500249,
+      "loss": 2.3867,
+      "step": 14183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024059544493556257,
+      "loss": 2.6505,
+      "step": 14184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024064204195395824,
+      "loss": 2.319,
+      "step": 14185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024068862519813084,
+      "loss": 2.5794,
+      "step": 14186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024073519466100162,
+      "loss": 2.7051,
+      "step": 14187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002407817503354937,
+      "loss": 2.839,
+      "step": 14188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024082829221453234,
+      "loss": 2.5802,
+      "step": 14189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024087482029104475,
+      "loss": 2.813,
+      "step": 14190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024092133455796066,
+      "loss": 2.7021,
+      "step": 14191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002409678350082115,
+      "loss": 2.6234,
+      "step": 14192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002410143216347309,
+      "loss": 2.4427,
+      "step": 14193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024106079443045455,
+      "loss": 2.3715,
+      "step": 14194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002411072533883205,
+      "loss": 2.2678,
+      "step": 14195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024115369850126865,
+      "loss": 2.8538,
+      "step": 14196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024120012976224092,
+      "loss": 3.1079,
+      "step": 14197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024124654716418172,
+      "loss": 2.7849,
+      "step": 14198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002412929507000372,
+      "loss": 2.5321,
+      "step": 14199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024133934036275575,
+      "loss": 2.667,
+      "step": 14200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024138571614528779,
+      "loss": 0.5835,
+      "step": 14201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024143207804058607,
+      "loss": 2.421,
+      "step": 14202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024147842604160517,
+      "loss": 2.5018,
+      "step": 14203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002415247601413022,
+      "loss": 2.7632,
+      "step": 14204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024157108033263567,
+      "loss": 2.2855,
+      "step": 14205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024161738660856695,
+      "loss": 2.4505,
+      "step": 14206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000241663678962059,
+      "loss": 2.8453,
+      "step": 14207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024170995738607742,
+      "loss": 2.5182,
+      "step": 14208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024175622187358917,
+      "loss": 2.3841,
+      "step": 14209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002418024724175641,
+      "loss": 2.0334,
+      "step": 14210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024184870901097365,
+      "loss": 2.4409,
+      "step": 14211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024189493164679184,
+      "loss": 2.281,
+      "step": 14212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002419411403179942,
+      "loss": 2.4115,
+      "step": 14213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024198733501755898,
+      "loss": 2.3903,
+      "step": 14214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024203351573846613,
+      "loss": 3.2357,
+      "step": 14215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024207968247369818,
+      "loss": 2.5619,
+      "step": 14216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002421258352162391,
+      "loss": 2.787,
+      "step": 14217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024217197395907567,
+      "loss": 1.9493,
+      "step": 14218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024221809869519635,
+      "loss": 2.3352,
+      "step": 14219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002422642094175921,
+      "loss": 2.5388,
+      "step": 14220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024231030611925568,
+      "loss": 1.7345,
+      "step": 14221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024235638879318213,
+      "loss": 1.4537,
+      "step": 14222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024240245743236845,
+      "loss": 2.8293,
+      "step": 14223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024244851202981414,
+      "loss": 2.2957,
+      "step": 14224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024249455257852054,
+      "loss": 2.8708,
+      "step": 14225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024254057907149108,
+      "loss": 2.0957,
+      "step": 14226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002425865915017316,
+      "loss": 2.3504,
+      "step": 14227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002426325898622499,
+      "loss": 2.5382,
+      "step": 14228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024267857414605584,
+      "loss": 2.402,
+      "step": 14229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024272454434616153,
+      "loss": 2.8957,
+      "step": 14230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024277050045558134,
+      "loss": 2.5031,
+      "step": 14231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002428164424673315,
+      "loss": 2.7344,
+      "step": 14232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002428623703744306,
+      "loss": 2.6301,
+      "step": 14233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024290828416989927,
+      "loss": 2.5194,
+      "step": 14234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024295418384676038,
+      "loss": 2.9917,
+      "step": 14235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024300006939803875,
+      "loss": 2.535,
+      "step": 14236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002430459408167618,
+      "loss": 2.0276,
+      "step": 14237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024309179809595832,
+      "loss": 2.5119,
+      "step": 14238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024313764122866005,
+      "loss": 2.6043,
+      "step": 14239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024318347020790035,
+      "loss": 2.6794,
+      "step": 14240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002432292850267152,
+      "loss": 3.2494,
+      "step": 14241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024327508567814204,
+      "loss": 2.3852,
+      "step": 14242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002433208721552212,
+      "loss": 2.2608,
+      "step": 14243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024336664445099456,
+      "loss": 2.2584,
+      "step": 14244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024341240255850683,
+      "loss": 2.2257,
+      "step": 14245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024345814647080402,
+      "loss": 2.7254,
+      "step": 14246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002435038761809351,
+      "loss": 3.2734,
+      "step": 14247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024354959168195055,
+      "loss": 2.5161,
+      "step": 14248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024359529296690376,
+      "loss": 2.1176,
+      "step": 14249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024364098002884932,
+      "loss": 2.3723,
+      "step": 14250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024368665286084488,
+      "loss": 2.3811,
+      "step": 14251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024373231145594955,
+      "loss": 2.6537,
+      "step": 14252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024377795580722523,
+      "loss": 2.8804,
+      "step": 14253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024382358590773552,
+      "loss": 2.4728,
+      "step": 14254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024386920175054636,
+      "loss": 2.4533,
+      "step": 14255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002439148033287257,
+      "loss": 2.6177,
+      "step": 14256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024396039063534402,
+      "loss": 2.5854,
+      "step": 14257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024400596366347366,
+      "loss": 2.1037,
+      "step": 14258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024405152240618912,
+      "loss": 2.4136,
+      "step": 14259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024409706685656732,
+      "loss": 2.9774,
+      "step": 14260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024414259700768714,
+      "loss": 2.26,
+      "step": 14261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024418811285262964,
+      "loss": 1.7728,
+      "step": 14262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024423361438447803,
+      "loss": 2.6141,
+      "step": 14263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000244279101596318,
+      "loss": 2.7649,
+      "step": 14264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000244324574481237,
+      "loss": 2.6717,
+      "step": 14265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002443700330323249,
+      "loss": 2.3776,
+      "step": 14266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002444154772426736,
+      "loss": 2.5051,
+      "step": 14267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024446090710537747,
+      "loss": 2.5168,
+      "step": 14268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002445063226135327,
+      "loss": 2.6485,
+      "step": 14269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024455172376023786,
+      "loss": 2.7248,
+      "step": 14270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024459711053859357,
+      "loss": 2.4738,
+      "step": 14271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002446424829417029,
+      "loss": 2.6713,
+      "step": 14272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024468784096267086,
+      "loss": 2.0636,
+      "step": 14273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002447331845946047,
+      "loss": 2.2551,
+      "step": 14274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024477851383061377,
+      "loss": 2.5483,
+      "step": 14275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002448238286638099,
+      "loss": 2.3689,
+      "step": 14276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002448691290873068,
+      "loss": 2.5627,
+      "step": 14277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002449144150942206,
+      "loss": 2.7125,
+      "step": 14278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002449596866776693,
+      "loss": 2.5728,
+      "step": 14279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002450049438307735,
+      "loss": 2.307,
+      "step": 14280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024505018654665575,
+      "loss": 3.0734,
+      "step": 14281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024509541481844085,
+      "loss": 2.3009,
+      "step": 14282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002451406286392556,
+      "loss": 2.6672,
+      "step": 14283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002451858280022294,
+      "loss": 2.5667,
+      "step": 14284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002452310129004936,
+      "loss": 2.4859,
+      "step": 14285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002452761833271816,
+      "loss": 1.2438,
+      "step": 14286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002453213392754294,
+      "loss": 2.4395,
+      "step": 14287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024536648073837495,
+      "loss": 2.811,
+      "step": 14288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002454116077091583,
+      "loss": 2.4484,
+      "step": 14289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024545672018092175,
+      "loss": 3.3285,
+      "step": 14290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002455018181468101,
+      "loss": 2.5806,
+      "step": 14291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002455469015999699,
+      "loss": 2.5604,
+      "step": 14292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002455919705335506,
+      "loss": 2.5184,
+      "step": 14293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024563702494070276,
+      "loss": 1.949,
+      "step": 14294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024568206481458024,
+      "loss": 2.3746,
+      "step": 14295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002457270901483384,
+      "loss": 2.7147,
+      "step": 14296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024577210093513536,
+      "loss": 2.7364,
+      "step": 14297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002458170971681308,
+      "loss": 3.0805,
+      "step": 14298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002458620788404872,
+      "loss": 2.9401,
+      "step": 14299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002459070459453689,
+      "loss": 2.8183,
+      "step": 14300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024595199847594266,
+      "loss": 2.4311,
+      "step": 14301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024599693642537735,
+      "loss": 2.8722,
+      "step": 14302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024604185978684405,
+      "loss": 2.714,
+      "step": 14303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000246086768553516,
+      "loss": 2.5343,
+      "step": 14304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002461316627185689,
+      "loss": 2.3576,
+      "step": 14305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024617654227518035,
+      "loss": 2.6336,
+      "step": 14306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024622140721653043,
+      "loss": 2.7575,
+      "step": 14307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024626625753580115,
+      "loss": 2.3668,
+      "step": 14308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002463110932261772,
+      "loss": 2.6222,
+      "step": 14309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002463559142808451,
+      "loss": 2.5877,
+      "step": 14310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024640072069299375,
+      "loss": 2.4522,
+      "step": 14311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024644551245581403,
+      "loss": 2.6028,
+      "step": 14312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024649028956249955,
+      "loss": 2.128,
+      "step": 14313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002465350520062457,
+      "loss": 1.5918,
+      "step": 14314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002465797997802503,
+      "loss": 2.1794,
+      "step": 14315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024662453287771323,
+      "loss": 2.842,
+      "step": 14316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024666925129183693,
+      "loss": 2.4838,
+      "step": 14317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002467139550158257,
+      "loss": 2.3805,
+      "step": 14318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024675864404288627,
+      "loss": 2.2573,
+      "step": 14319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002468033183662276,
+      "loss": 2.5763,
+      "step": 14320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024684797797906087,
+      "loss": 2.0596,
+      "step": 14321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024689262287459947,
+      "loss": 2.7242,
+      "step": 14322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002469372530460589,
+      "loss": 2.6175,
+      "step": 14323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002469818684866572,
+      "loss": 1.9823,
+      "step": 14324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024702646918961437,
+      "loss": 1.9118,
+      "step": 14325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024707105514815294,
+      "loss": 2.5086,
+      "step": 14326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024711562635549726,
+      "loss": 2.6889,
+      "step": 14327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024716018280487433,
+      "loss": 2.7552,
+      "step": 14328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002472047244895131,
+      "loss": 2.6166,
+      "step": 14329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024724925140264507,
+      "loss": 2.3362,
+      "step": 14330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002472937635375036,
+      "loss": 2.7299,
+      "step": 14331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024733826088732465,
+      "loss": 2.385,
+      "step": 14332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002473827434453461,
+      "loss": 2.809,
+      "step": 14333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002474272112048086,
+      "loss": 2.2011,
+      "step": 14334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024747166415895426,
+      "loss": 3.06,
+      "step": 14335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002475161023010282,
+      "loss": 2.1405,
+      "step": 14336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024756052562427725,
+      "loss": 2.3152,
+      "step": 14337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000247604934121951,
+      "loss": 2.0261,
+      "step": 14338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024764932778730073,
+      "loss": 2.2646,
+      "step": 14339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002476937066135804,
+      "loss": 2.3459,
+      "step": 14340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000247738070594046,
+      "loss": 2.9891,
+      "step": 14341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024778241972195597,
+      "loss": 2.754,
+      "step": 14342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002478267539905708,
+      "loss": 2.1535,
+      "step": 14343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024787107339315336,
+      "loss": 2.6874,
+      "step": 14344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002479153779229686,
+      "loss": 2.7833,
+      "step": 14345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002479596675732841,
+      "loss": 2.5886,
+      "step": 14346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002480039423373695,
+      "loss": 1.7419,
+      "step": 14347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002480482022084963,
+      "loss": 3.0853,
+      "step": 14348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002480924471799391,
+      "loss": 2.526,
+      "step": 14349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002481366772449741,
+      "loss": 2.4893,
+      "step": 14350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024818089239688,
+      "loss": 2.6678,
+      "step": 14351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002482250926289377,
+      "loss": 2.3415,
+      "step": 14352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002482692779344304,
+      "loss": 2.4012,
+      "step": 14353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002483134483066437,
+      "loss": 2.1374,
+      "step": 14354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024835760373886526,
+      "loss": 2.3053,
+      "step": 14355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000248401744224385,
+      "loss": 2.1887,
+      "step": 14356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024844586975649537,
+      "loss": 2.5238,
+      "step": 14357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002484899803284908,
+      "loss": 2.685,
+      "step": 14358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002485340759336683,
+      "loss": 2.3723,
+      "step": 14359
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002485781565653267,
+      "loss": 2.5021,
+      "step": 14360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024862222221676766,
+      "loss": 2.2138,
+      "step": 14361
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024866627288129464,
+      "loss": 1.9975,
+      "step": 14362
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002487103085522137,
+      "loss": 2.4976,
+      "step": 14363
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002487543292228329,
+      "loss": 2.3134,
+      "step": 14364
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024879833488646296,
+      "loss": 2.7552,
+      "step": 14365
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002488423255364165,
+      "loss": 3.0081,
+      "step": 14366
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002488863011660086,
+      "loss": 2.7072,
+      "step": 14367
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024893026176855656,
+      "loss": 2.6608,
+      "step": 14368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024897420733738007,
+      "loss": 2.6788,
+      "step": 14369
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024901813786580107,
+      "loss": 2.3846,
+      "step": 14370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002490620533471437,
+      "loss": 2.9254,
+      "step": 14371
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002491059537747343,
+      "loss": 2.4016,
+      "step": 14372
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024914983914190185,
+      "loss": 2.292,
+      "step": 14373
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002491937094419773,
+      "loss": 2.2179,
+      "step": 14374
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024923756466829403,
+      "loss": 2.0443,
+      "step": 14375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002492814048141877,
+      "loss": 1.4243,
+      "step": 14376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024932522987299624,
+      "loss": 2.9169,
+      "step": 14377
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024936903983805965,
+      "loss": 2.1231,
+      "step": 14378
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024941283470272075,
+      "loss": 2.9173,
+      "step": 14379
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024945661446032426,
+      "loss": 2.7935,
+      "step": 14380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002495003791042171,
+      "loss": 2.1227,
+      "step": 14381
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024954412862774895,
+      "loss": 2.4616,
+      "step": 14382
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024958786302427136,
+      "loss": 2.4621,
+      "step": 14383
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002496315822871383,
+      "loss": 2.656,
+      "step": 14384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024967528640970605,
+      "loss": 2.7444,
+      "step": 14385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024971897538533333,
+      "loss": 2.6935,
+      "step": 14386
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024976264920738094,
+      "loss": 2.5319,
+      "step": 14387
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024980630786921215,
+      "loss": 2.4176,
+      "step": 14388
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002498499513641923,
+      "loss": 2.6151,
+      "step": 14389
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002498935796856894,
+      "loss": 2.5101,
+      "step": 14390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002499371928270736,
+      "loss": 3.1457,
+      "step": 14391
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00024998079078171714,
+      "loss": 2.2239,
+      "step": 14392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002500243735429947,
+      "loss": 2.3565,
+      "step": 14393
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025006794110428356,
+      "loss": 2.4363,
+      "step": 14394
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000250111493458963,
+      "loss": 2.5464,
+      "step": 14395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025015503060041457,
+      "loss": 1.589,
+      "step": 14396
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025019855252202225,
+      "loss": 2.3548,
+      "step": 14397
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002502420592171724,
+      "loss": 2.8277,
+      "step": 14398
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002502855506792537,
+      "loss": 3.1306,
+      "step": 14399
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025032902690165694,
+      "loss": 2.8869,
+      "step": 14400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002503724878777753,
+      "loss": 2.5501,
+      "step": 14401
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025041593360100443,
+      "loss": 1.9835,
+      "step": 14402
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002504593640647422,
+      "loss": 1.212,
+      "step": 14403
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002505027792623887,
+      "loss": 2.6763,
+      "step": 14404
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002505461791873464,
+      "loss": 2.6331,
+      "step": 14405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002505895638330204,
+      "loss": 2.6805,
+      "step": 14406
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002506329331928176,
+      "loss": 1.8724,
+      "step": 14407
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002506762872601474,
+      "loss": 2.0969,
+      "step": 14408
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025071962602842186,
+      "loss": 2.6029,
+      "step": 14409
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002507629494910549,
+      "loss": 2.2127,
+      "step": 14410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002508062576414631,
+      "loss": 2.2628,
+      "step": 14411
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000250849550473065,
+      "loss": 2.7303,
+      "step": 14412
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025089282797928193,
+      "loss": 2.9819,
+      "step": 14413
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025093609015353714,
+      "loss": 2.8301,
+      "step": 14414
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002509793369892567,
+      "loss": 2.7417,
+      "step": 14415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002510225684798682,
+      "loss": 2.766,
+      "step": 14416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025106578461880257,
+      "loss": 2.3084,
+      "step": 14417
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025110898539949214,
+      "loss": 2.3192,
+      "step": 14418
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002511521708153724,
+      "loss": 2.4063,
+      "step": 14419
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025119534085988037,
+      "loss": 2.2591,
+      "step": 14420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002512384955264561,
+      "loss": 2.7704,
+      "step": 14421
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025128163480854146,
+      "loss": 2.2786,
+      "step": 14422
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002513247586995812,
+      "loss": 2.6921,
+      "step": 14423
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002513678671930217,
+      "loss": 2.4163,
+      "step": 14424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002514109602823124,
+      "loss": 2.4231,
+      "step": 14425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025145403796090446,
+      "loss": 2.5773,
+      "step": 14426
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025149710022225204,
+      "loss": 1.1921,
+      "step": 14427
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025154014705981086,
+      "loss": 2.2753,
+      "step": 14428
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002515831784670397,
+      "loss": 2.0065,
+      "step": 14429
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002516261944373992,
+      "loss": 2.7498,
+      "step": 14430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025166919496435276,
+      "loss": 2.3174,
+      "step": 14431
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025171218004136557,
+      "loss": 2.7807,
+      "step": 14432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025175514966190585,
+      "loss": 2.6659,
+      "step": 14433
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002517981038194435,
+      "loss": 2.397,
+      "step": 14434
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002518410425074513,
+      "loss": 2.2393,
+      "step": 14435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025188396571940413,
+      "loss": 2.626,
+      "step": 14436
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002519268734487792,
+      "loss": 2.5332,
+      "step": 14437
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025196976568905605,
+      "loss": 2.6915,
+      "step": 14438
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002520126424337168,
+      "loss": 2.8688,
+      "step": 14439
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002520555036762458,
+      "loss": 2.7797,
+      "step": 14440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002520983494101296,
+      "loss": 2.075,
+      "step": 14441
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002521411796288573,
+      "loss": 2.3952,
+      "step": 14442
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025218399432592033,
+      "loss": 3.0162,
+      "step": 14443
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002522267934948124,
+      "loss": 2.5774,
+      "step": 14444
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002522695771290296,
+      "loss": 2.5937,
+      "step": 14445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025231234522207053,
+      "loss": 2.0043,
+      "step": 14446
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025235509776743586,
+      "loss": 2.4799,
+      "step": 14447
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025239783475862904,
+      "loss": 2.4948,
+      "step": 14448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025244055618915527,
+      "loss": 2.6638,
+      "step": 14449
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002524832620525228,
+      "loss": 2.4353,
+      "step": 14450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025252595234224177,
+      "loss": 2.1641,
+      "step": 14451
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000252568627051825,
+      "loss": 0.0073,
+      "step": 14452
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025261128617478723,
+      "loss": 2.8206,
+      "step": 14453
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002526539297046461,
+      "loss": 2.5629,
+      "step": 14454
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025269655763492127,
+      "loss": 2.6848,
+      "step": 14455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025273916995913507,
+      "loss": 2.5837,
+      "step": 14456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002527817666708117,
+      "loss": 2.0391,
+      "step": 14457
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025282434776347824,
+      "loss": 2.3574,
+      "step": 14458
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002528669132306638,
+      "loss": 2.2607,
+      "step": 14459
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002529094630659003,
+      "loss": 2.9098,
+      "step": 14460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002529519972627214,
+      "loss": 1.8361,
+      "step": 14461
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025299451581466375,
+      "loss": 2.5087,
+      "step": 14462
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002530370187152659,
+      "loss": 2.6218,
+      "step": 14463
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002530795059580692,
+      "loss": 2.6107,
+      "step": 14464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000253121977536617,
+      "loss": 2.5699,
+      "step": 14465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002531644334444553,
+      "loss": 2.2749,
+      "step": 14466
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002532068736751322,
+      "loss": 2.8604,
+      "step": 14467
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025324929822219866,
+      "loss": 2.0306,
+      "step": 14468
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002532917070792075,
+      "loss": 2.8617,
+      "step": 14469
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002533341002397141,
+      "loss": 2.7387,
+      "step": 14470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002533764776972766,
+      "loss": 3.252,
+      "step": 14471
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025341883944545487,
+      "loss": 2.842,
+      "step": 14472
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002534611854778116,
+      "loss": 2.315,
+      "step": 14473
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002535035157879117,
+      "loss": 2.3226,
+      "step": 14474
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025354583036932277,
+      "loss": 2.6113,
+      "step": 14475
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025358812921561435,
+      "loss": 2.3025,
+      "step": 14476
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025363041232035865,
+      "loss": 2.9728,
+      "step": 14477
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025367267967713015,
+      "loss": 1.9216,
+      "step": 14478
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000253714931279506,
+      "loss": 3.1585,
+      "step": 14479
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002537571671210652,
+      "loss": 2.5889,
+      "step": 14480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025379938719538976,
+      "loss": 2.4905,
+      "step": 14481
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002538415914960635,
+      "loss": 2.0675,
+      "step": 14482
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002538837800166732,
+      "loss": 2.6265,
+      "step": 14483
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002539259527508077,
+      "loss": 2.2987,
+      "step": 14484
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025396810969205837,
+      "loss": 2.2467,
+      "step": 14485
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002540102508340187,
+      "loss": 2.0406,
+      "step": 14486
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025405237617028497,
+      "loss": 2.0611,
+      "step": 14487
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002540944856944557,
+      "loss": 2.7367,
+      "step": 14488
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002541365794001318,
+      "loss": 1.8353,
+      "step": 14489
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025417865728091647,
+      "loss": 2.7221,
+      "step": 14490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002542207193304156,
+      "loss": 2.3639,
+      "step": 14491
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002542627655422373,
+      "loss": 2.0761,
+      "step": 14492
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000254304795909992,
+      "loss": 2.5826,
+      "step": 14493
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002543468104272926,
+      "loss": 2.6248,
+      "step": 14494
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002543888090877547,
+      "loss": 2.1775,
+      "step": 14495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025443079188499593,
+      "loss": 2.6069,
+      "step": 14496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002544727588126363,
+      "loss": 2.4169,
+      "step": 14497
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002545147098642987,
+      "loss": 2.5018,
+      "step": 14498
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000254556645033608,
+      "loss": 2.5879,
+      "step": 14499
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025459856431419154,
+      "loss": 2.1959,
+      "step": 14500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002546404676996791,
+      "loss": 1.8251,
+      "step": 14501
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002546823551837032,
+      "loss": 2.8543,
+      "step": 14502
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025472422675989814,
+      "loss": 2.3552,
+      "step": 14503
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002547660824219014,
+      "loss": 2.1588,
+      "step": 14504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002548079221633521,
+      "loss": 2.0899,
+      "step": 14505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025484974597789233,
+      "loss": 2.3204,
+      "step": 14506
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025489155385916636,
+      "loss": 2.2922,
+      "step": 14507
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025493334580082113,
+      "loss": 2.0649,
+      "step": 14508
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002549751217965055,
+      "loss": 2.4833,
+      "step": 14509
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025501688183987136,
+      "loss": 2.2782,
+      "step": 14510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002550586259245725,
+      "loss": 2.6024,
+      "step": 14511
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002551003540442656,
+      "loss": 2.6517,
+      "step": 14512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025514206619260937,
+      "loss": 2.5433,
+      "step": 14513
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002551837623632652,
+      "loss": 2.3471,
+      "step": 14514
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025522544254989667,
+      "loss": 2.2717,
+      "step": 14515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002552671067461703,
+      "loss": 2.7934,
+      "step": 14516
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002553087549457542,
+      "loss": 2.5876,
+      "step": 14517
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002553503871423198,
+      "loss": 2.1818,
+      "step": 14518
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025539200332954016,
+      "loss": 3.0018,
+      "step": 14519
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025543360350109155,
+      "loss": 2.6113,
+      "step": 14520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025547518765065214,
+      "loss": 2.4308,
+      "step": 14521
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002555167557719027,
+      "loss": 2.1547,
+      "step": 14522
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025555830785852624,
+      "loss": 2.2652,
+      "step": 14523
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025559984390420863,
+      "loss": 2.2361,
+      "step": 14524
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025564136390263787,
+      "loss": 1.5016,
+      "step": 14525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002556828678475044,
+      "loss": 2.2611,
+      "step": 14526
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025572435573250116,
+      "loss": 3.0238,
+      "step": 14527
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002557658275513236,
+      "loss": 2.8562,
+      "step": 14528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025580728329766954,
+      "loss": 2.4319,
+      "step": 14529
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002558487229652391,
+      "loss": 2.6903,
+      "step": 14530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025589014654773526,
+      "loss": 2.98,
+      "step": 14531
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025593155403886295,
+      "loss": 2.1637,
+      "step": 14532
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002559729454323299,
+      "loss": 2.3145,
+      "step": 14533
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000256014320721846,
+      "loss": 2.6563,
+      "step": 14534
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025605567990112394,
+      "loss": 2.5572,
+      "step": 14535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002560970229638784,
+      "loss": 2.6665,
+      "step": 14536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002561383499038272,
+      "loss": 2.9833,
+      "step": 14537
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002561796607146897,
+      "loss": 2.104,
+      "step": 14538
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002562209553901885,
+      "loss": 2.5536,
+      "step": 14539
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002562622339240482,
+      "loss": 2.0136,
+      "step": 14540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002563034963099962,
+      "loss": 0.4606,
+      "step": 14541
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002563447425417618,
+      "loss": 2.6373,
+      "step": 14542
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025638597261307735,
+      "loss": 2.2063,
+      "step": 14543
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002564271865176773,
+      "loss": 2.5848,
+      "step": 14544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000256468384249299,
+      "loss": 2.3947,
+      "step": 14545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025650956580168137,
+      "loss": 2.9491,
+      "step": 14546
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025655073116856675,
+      "loss": 2.7091,
+      "step": 14547
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025659188034369933,
+      "loss": 2.0455,
+      "step": 14548
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002566330133208262,
+      "loss": 2.9913,
+      "step": 14549
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025667413009369637,
+      "loss": 2.4804,
+      "step": 14550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002567152306560619,
+      "loss": 2.9265,
+      "step": 14551
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002567563150016767,
+      "loss": 1.9778,
+      "step": 14552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000256797383124298,
+      "loss": 2.5794,
+      "step": 14553
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002568384350176844,
+      "loss": 2.5536,
+      "step": 14554
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025687947067559784,
+      "loss": 2.7738,
+      "step": 14555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002569204900918023,
+      "loss": 1.1586,
+      "step": 14556
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002569614932600645,
+      "loss": 2.6077,
+      "step": 14557
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002570024801741534,
+      "loss": 2.6378,
+      "step": 14558
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002570434508278406,
+      "loss": 0.8995,
+      "step": 14559
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002570844052148998,
+      "loss": 2.4297,
+      "step": 14560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002571253433291079,
+      "loss": 2.4732,
+      "step": 14561
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025716626516424345,
+      "loss": 2.769,
+      "step": 14562
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002572071707140879,
+      "loss": 2.4837,
+      "step": 14563
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002572480599724253,
+      "loss": 2.1533,
+      "step": 14564
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025728893293304196,
+      "loss": 2.468,
+      "step": 14565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025732978958972665,
+      "loss": 1.8898,
+      "step": 14566
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025737062993627056,
+      "loss": 2.2802,
+      "step": 14567
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002574114539664677,
+      "loss": 3.0228,
+      "step": 14568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025745226167411426,
+      "loss": 2.0742,
+      "step": 14569
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025749305305300895,
+      "loss": 2.8127,
+      "step": 14570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002575338280969529,
+      "loss": 2.9503,
+      "step": 14571
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025757458679975,
+      "loss": 2.59,
+      "step": 14572
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002576153291552064,
+      "loss": 2.7404,
+      "step": 14573
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002576560551571307,
+      "loss": 2.5595,
+      "step": 14574
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000257696764799334,
+      "loss": 2.4923,
+      "step": 14575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002577374580756302,
+      "loss": 2.2933,
+      "step": 14576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002577781349798352,
+      "loss": 2.3076,
+      "step": 14577
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025781879550576775,
+      "loss": 2.5649,
+      "step": 14578
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002578594396472488,
+      "loss": 2.4308,
+      "step": 14579
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002579000673981022,
+      "loss": 2.0603,
+      "step": 14580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002579406787521539,
+      "loss": 1.6393,
+      "step": 14581
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025798127370323245,
+      "loss": 2.7901,
+      "step": 14582
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002580218522451689,
+      "loss": 2.4674,
+      "step": 14583
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025806241437179695,
+      "loss": 2.277,
+      "step": 14584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025810296007695263,
+      "loss": 3.0402,
+      "step": 14585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002581434893544743,
+      "loss": 2.1869,
+      "step": 14586
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002581840021982033,
+      "loss": 2.6872,
+      "step": 14587
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002582244986019831,
+      "loss": 2.0622,
+      "step": 14588
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002582649785596597,
+      "loss": 2.5086,
+      "step": 14589
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002583054420650816,
+      "loss": 2.5019,
+      "step": 14590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00025834588911209996,
+      "loss": 2.3226,
+      "step": 14591
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002583863196945682,
+      "loss": 2.4005,
+      "step": 14592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002584267338063426,
+      "loss": 2.3354,
+      "step": 14593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025846713144128154,
+      "loss": 2.8595,
+      "step": 14594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002585075125932461,
+      "loss": 1.7221,
+      "step": 14595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025854787725609977,
+      "loss": 2.1262,
+      "step": 14596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002585882254237088,
+      "loss": 2.2415,
+      "step": 14597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025862855708994173,
+      "loss": 2.3579,
+      "step": 14598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002586688722486696,
+      "loss": 2.4585,
+      "step": 14599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025870917089376586,
+      "loss": 2.2041,
+      "step": 14600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025874945301910683,
+      "loss": 2.5996,
+      "step": 14601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002587897186185711,
+      "loss": 2.5953,
+      "step": 14602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025882996768603967,
+      "loss": 2.5627,
+      "step": 14603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025887020021539626,
+      "loss": 2.7171,
+      "step": 14604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000258910416200527,
+      "loss": 2.2646,
+      "step": 14605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002589506156353206,
+      "loss": 2.4523,
+      "step": 14606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002589907985136682,
+      "loss": 2.7952,
+      "step": 14607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002590309648294634,
+      "loss": 2.4845,
+      "step": 14608
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002590711145766026,
+      "loss": 2.6074,
+      "step": 14609
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025911124774898443,
+      "loss": 2.5822,
+      "step": 14610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002591513643405102,
+      "loss": 2.6692,
+      "step": 14611
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025919146434508347,
+      "loss": 2.3716,
+      "step": 14612
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002592315477566108,
+      "loss": 3.1488,
+      "step": 14613
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002592716145690009,
+      "loss": 3.1563,
+      "step": 14614
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002593116647761651,
+      "loss": 2.6202,
+      "step": 14615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025935169837201707,
+      "loss": 2.8359,
+      "step": 14616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002593917153504735,
+      "loss": 2.8689,
+      "step": 14617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025943171570545314,
+      "loss": 2.3561,
+      "step": 14618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002594716994308774,
+      "loss": 2.3255,
+      "step": 14619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002595116665206704,
+      "loss": 2.5197,
+      "step": 14620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002595516169687585,
+      "loss": 2.915,
+      "step": 14621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025959155076907064,
+      "loss": 3.3354,
+      "step": 14622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002596314679155384,
+      "loss": 2.7824,
+      "step": 14623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025967136840209605,
+      "loss": 2.9039,
+      "step": 14624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002597112522226799,
+      "loss": 2.2264,
+      "step": 14625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025975111937122944,
+      "loss": 2.6877,
+      "step": 14626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000259790969841686,
+      "loss": 2.5634,
+      "step": 14627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000259830803627994,
+      "loss": 2.4914,
+      "step": 14628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002598706207241001,
+      "loss": 1.913,
+      "step": 14629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002599104211239537,
+      "loss": 2.8229,
+      "step": 14630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025995020482150644,
+      "loss": 2.5435,
+      "step": 14631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00025998997181071276,
+      "loss": 2.3116,
+      "step": 14632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002600297220855295,
+      "loss": 0.903,
+      "step": 14633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026006945563991635,
+      "loss": 2.0592,
+      "step": 14634
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026010917246783484,
+      "loss": 2.3177,
+      "step": 14635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002601488725632499,
+      "loss": 2.739,
+      "step": 14636
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002601885559201283,
+      "loss": 2.4718,
+      "step": 14637
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026022822253243995,
+      "loss": 3.0847,
+      "step": 14638
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026026787239415665,
+      "loss": 2.6627,
+      "step": 14639
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026030750549925334,
+      "loss": 2.6134,
+      "step": 14640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026034712184170705,
+      "loss": 2.8159,
+      "step": 14641
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002603867214154979,
+      "loss": 2.6657,
+      "step": 14642
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002604263042146079,
+      "loss": 2.7042,
+      "step": 14643
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002604658702330221,
+      "loss": 2.8157,
+      "step": 14644
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026050541946472783,
+      "loss": 2.1795,
+      "step": 14645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002605449519037152,
+      "loss": 2.9773,
+      "step": 14646
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002605844675439767,
+      "loss": 1.962,
+      "step": 14647
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002606239663795074,
+      "loss": 2.0511,
+      "step": 14648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002606634484043049,
+      "loss": 2.4069,
+      "step": 14649
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026070291361236954,
+      "loss": 1.894,
+      "step": 14650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026074236199770395,
+      "loss": 2.6405,
+      "step": 14651
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002607817935543134,
+      "loss": 2.2906,
+      "step": 14652
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002608212082762059,
+      "loss": 2.5938,
+      "step": 14653
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002608606061573919,
+      "loss": 2.942,
+      "step": 14654
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002608999871918842,
+      "loss": 2.1988,
+      "step": 14655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026093935137369844,
+      "loss": 1.4658,
+      "step": 14656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002609786986968528,
+      "loss": 3.0055,
+      "step": 14657
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002610180291553678,
+      "loss": 2.6187,
+      "step": 14658
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002610573427432668,
+      "loss": 2.2695,
+      "step": 14659
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002610966394545755,
+      "loss": 2.511,
+      "step": 14660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002611359192833224,
+      "loss": 2.7302,
+      "step": 14661
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002611751822235382,
+      "loss": 2.4768,
+      "step": 14662
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026121442826925666,
+      "loss": 2.3628,
+      "step": 14663
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002612536574145136,
+      "loss": 1.3806,
+      "step": 14664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002612928696533478,
+      "loss": 2.3427,
+      "step": 14665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002613320649798003,
+      "loss": 2.1548,
+      "step": 14666
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026137124338791517,
+      "loss": 2.3487,
+      "step": 14667
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002614104048717384,
+      "loss": 2.2626,
+      "step": 14668
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002614495494253191,
+      "loss": 2.2911,
+      "step": 14669
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002614886770427086,
+      "loss": 2.6825,
+      "step": 14670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026152778771796134,
+      "loss": 2.4227,
+      "step": 14671
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002615668814451334,
+      "loss": 2.7214,
+      "step": 14672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002616059582182844,
+      "loss": 2.5306,
+      "step": 14673
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002616450180314758,
+      "loss": 2.1933,
+      "step": 14674
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002616840608787723,
+      "loss": 2.6472,
+      "step": 14675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002617230867542406,
+      "loss": 1.741,
+      "step": 14676
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026176209565195036,
+      "loss": 2.8446,
+      "step": 14677
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026180108756597344,
+      "loss": 2.5719,
+      "step": 14678
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026184006249038483,
+      "loss": 2.5628,
+      "step": 14679
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002618790204192617,
+      "loss": 2.7166,
+      "step": 14680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026191796134668376,
+      "loss": 2.5636,
+      "step": 14681
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002619568852667334,
+      "loss": 2.3489,
+      "step": 14682
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026199579217349586,
+      "loss": 2.5442,
+      "step": 14683
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026203468206105864,
+      "loss": 1.9686,
+      "step": 14684
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002620735549235118,
+      "loss": 2.167,
+      "step": 14685
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002621124107549483,
+      "loss": 2.4413,
+      "step": 14686
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026215124954946336,
+      "loss": 2.4388,
+      "step": 14687
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000262190071301155,
+      "loss": 2.9906,
+      "step": 14688
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002622288760041236,
+      "loss": 2.4897,
+      "step": 14689
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002622676636524726,
+      "loss": 2.4969,
+      "step": 14690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026230643424030747,
+      "loss": 2.5423,
+      "step": 14691
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002623451877617366,
+      "loss": 2.1988,
+      "step": 14692
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002623839242108708,
+      "loss": 2.6817,
+      "step": 14693
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002624226435818237,
+      "loss": 2.3355,
+      "step": 14694
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026246134586871135,
+      "loss": 2.5062,
+      "step": 14695
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002625000310656524,
+      "loss": 1.9909,
+      "step": 14696
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002625386991667682,
+      "loss": 2.4731,
+      "step": 14697
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026257735016618257,
+      "loss": 2.3243,
+      "step": 14698
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002626159840580221,
+      "loss": 1.9315,
+      "step": 14699
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026265460083641575,
+      "loss": 2.5379,
+      "step": 14700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002626932004954952,
+      "loss": 2.5007,
+      "step": 14701
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026273178302939484,
+      "loss": 2.3057,
+      "step": 14702
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026277034843225154,
+      "loss": 2.6259,
+      "step": 14703
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026280889669820474,
+      "loss": 2.5006,
+      "step": 14704
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002628474278213965,
+      "loss": 2.5659,
+      "step": 14705
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026288594179597164,
+      "loss": 2.4803,
+      "step": 14706
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002629244386160774,
+      "loss": 2.5481,
+      "step": 14707
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002629629182758636,
+      "loss": 2.5953,
+      "step": 14708
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000263001380769483,
+      "loss": 2.7615,
+      "step": 14709
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026303982609109057,
+      "loss": 2.3894,
+      "step": 14710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000263078254234844,
+      "loss": 2.5119,
+      "step": 14711
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002631166651949037,
+      "loss": 2.4335,
+      "step": 14712
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002631550589654327,
+      "loss": 2.3798,
+      "step": 14713
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002631934355405964,
+      "loss": 2.3976,
+      "step": 14714
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002632317949145633,
+      "loss": 2.7233,
+      "step": 14715
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026327013708150383,
+      "loss": 2.8666,
+      "step": 14716
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026330846203559166,
+      "loss": 2.4922,
+      "step": 14717
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026334676977100266,
+      "loss": 2.8218,
+      "step": 14718
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026338506028191573,
+      "loss": 2.5898,
+      "step": 14719
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026342333356251175,
+      "loss": 2.5826,
+      "step": 14720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026346158960697494,
+      "loss": 2.898,
+      "step": 14721
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026349982840949155,
+      "loss": 2.7224,
+      "step": 14722
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000263538049964251,
+      "loss": 2.4487,
+      "step": 14723
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026357625426544465,
+      "loss": 2.3931,
+      "step": 14724
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002636144413072672,
+      "loss": 2.0684,
+      "step": 14725
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002636526110839154,
+      "loss": 2.0949,
+      "step": 14726
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002636907635895892,
+      "loss": 2.7025,
+      "step": 14727
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002637288988184904,
+      "loss": 2.6112,
+      "step": 14728
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026376701676482416,
+      "loss": 2.4539,
+      "step": 14729
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002638051174227978,
+      "loss": 2.8039,
+      "step": 14730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002638432007866217,
+      "loss": 2.4614,
+      "step": 14731
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026388126685050826,
+      "loss": 2.7664,
+      "step": 14732
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002639193156086731,
+      "loss": 2.2836,
+      "step": 14733
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000263957347055334,
+      "loss": 2.2807,
+      "step": 14734
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026399536118471185,
+      "loss": 3.0059,
+      "step": 14735
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002640333579910298,
+      "loss": 2.2401,
+      "step": 14736
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002640713374685137,
+      "loss": 2.8183,
+      "step": 14737
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026410929961139207,
+      "loss": 2.1005,
+      "step": 14738
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002641472444138962,
+      "loss": 1.1452,
+      "step": 14739
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002641851718702598,
+      "loss": 2.3595,
+      "step": 14740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026422308197471923,
+      "loss": 2.6138,
+      "step": 14741
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026426097472151373,
+      "loss": 1.8146,
+      "step": 14742
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000264298850104885,
+      "loss": 2.4511,
+      "step": 14743
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026433670811907726,
+      "loss": 2.8096,
+      "step": 14744
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002643745487583375,
+      "loss": 2.4723,
+      "step": 14745
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026441237201691554,
+      "loss": 2.4301,
+      "step": 14746
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026445017788906335,
+      "loss": 2.7408,
+      "step": 14747
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002644879663690362,
+      "loss": 2.5811,
+      "step": 14748
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002645257374510913,
+      "loss": 2.5972,
+      "step": 14749
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026456349112948907,
+      "loss": 2.5388,
+      "step": 14750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026460122739849223,
+      "loss": 2.5691,
+      "step": 14751
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026463894625236643,
+      "loss": 2.9722,
+      "step": 14752
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026467664768537965,
+      "loss": 2.666,
+      "step": 14753
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002647143316918027,
+      "loss": 2.0722,
+      "step": 14754
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000264751998265909,
+      "loss": 1.9641,
+      "step": 14755
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026478964740197483,
+      "loss": 2.2131,
+      "step": 14756
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002648272790942786,
+      "loss": 2.6041,
+      "step": 14757
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002648648933371019,
+      "loss": 2.1489,
+      "step": 14758
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026490249012472864,
+      "loss": 2.0916,
+      "step": 14759
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026494006945144576,
+      "loss": 2.428,
+      "step": 14760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002649776313115422,
+      "loss": 2.4925,
+      "step": 14761
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002650151756993103,
+      "loss": 2.4505,
+      "step": 14762
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026505270260904445,
+      "loss": 2.308,
+      "step": 14763
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002650902120350423,
+      "loss": 2.6167,
+      "step": 14764
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651277039716033,
+      "loss": 2.9206,
+      "step": 14765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651651784130306,
+      "loss": 2.492,
+      "step": 14766
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026520263535362906,
+      "loss": 2.7915,
+      "step": 14767
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652400747877069,
+      "loss": 2.345,
+      "step": 14768
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652774967095746,
+      "loss": 2.1947,
+      "step": 14769
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653149011135454,
+      "loss": 1.9729,
+      "step": 14770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653522879939352,
+      "loss": 2.4833,
+      "step": 14771
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026538965734506265,
+      "loss": 2.5028,
+      "step": 14772
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265427009161249,
+      "loss": 2.3567,
+      "step": 14773
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002654643434368181,
+      "loss": 3.2785,
+      "step": 14774
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026550166016609656,
+      "loss": 2.268,
+      "step": 14775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655389593434137,
+      "loss": 2.5383,
+      "step": 14776
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026557624096310137,
+      "loss": 2.182,
+      "step": 14777
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265613505019494,
+      "loss": 2.3721,
+      "step": 14778
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656507515069291,
+      "loss": 2.4581,
+      "step": 14779
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026568798041974644,
+      "loss": 2.6861,
+      "step": 14780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026572519175228866,
+      "loss": 2.6654,
+      "step": 14781
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002657623854989008,
+      "loss": 2.5556,
+      "step": 14782
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026579956165393115,
+      "loss": 2.4491,
+      "step": 14783
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658367202117302,
+      "loss": 1.5684,
+      "step": 14784
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658738611666511,
+      "loss": 2.4205,
+      "step": 14785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659109845130498,
+      "loss": 2.8331,
+      "step": 14786
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659480902452851,
+      "loss": 2.8427,
+      "step": 14787
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659851783577183,
+      "loss": 2.2774,
+      "step": 14788
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002660222488447132,
+      "loss": 2.6162,
+      "step": 14789
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002660593017006365,
+      "loss": 2.6034,
+      "step": 14790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026609633691985774,
+      "loss": 2.8256,
+      "step": 14791
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026613335449674884,
+      "loss": 2.2768,
+      "step": 14792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661703544256844,
+      "loss": 2.7567,
+      "step": 14793
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026620733670104185,
+      "loss": 2.6099,
+      "step": 14794
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026624430131720134,
+      "loss": 2.232,
+      "step": 14795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002662812482685456,
+      "loss": 2.8204,
+      "step": 14796
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663181775494599,
+      "loss": 2.6789,
+      "step": 14797
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026635508915433265,
+      "loss": 2.6211,
+      "step": 14798
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026639198307755447,
+      "loss": 2.4788,
+      "step": 14799
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664288593135189,
+      "loss": 2.4756,
+      "step": 14800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026646571785662204,
+      "loss": 2.4032,
+      "step": 14801
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665025587012629,
+      "loss": 2.6881,
+      "step": 14802
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026653938184184286,
+      "loss": 2.6975,
+      "step": 14803
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026657618727276655,
+      "loss": 2.3903,
+      "step": 14804
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666129749884404,
+      "loss": 2.9079,
+      "step": 14805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666497449832744,
+      "loss": 2.9649,
+      "step": 14806
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026668649725168067,
+      "loss": 2.3625,
+      "step": 14807
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667232317880744,
+      "loss": 2.6658,
+      "step": 14808
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667599485868732,
+      "loss": 2.6984,
+      "step": 14809
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667966476424976,
+      "loss": 2.5943,
+      "step": 14810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002668333289493705,
+      "loss": 2.3418,
+      "step": 14811
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026686999250191793,
+      "loss": 2.7801,
+      "step": 14812
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002669066382945682,
+      "loss": 3.2285,
+      "step": 14813
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002669432663217527,
+      "loss": 2.2456,
+      "step": 14814
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026697987657790513,
+      "loss": 2.2455,
+      "step": 14815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670164690574622,
+      "loss": 2.459,
+      "step": 14816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026705304375486324,
+      "loss": 2.1624,
+      "step": 14817
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026708960066455025,
+      "loss": 2.3419,
+      "step": 14818
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671261397809678,
+      "loss": 2.2595,
+      "step": 14819
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026716266109856347,
+      "loss": 2.7439,
+      "step": 14820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026719916461178734,
+      "loss": 2.9468,
+      "step": 14821
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672356503150922,
+      "loss": 2.2607,
+      "step": 14822
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672721182029335,
+      "loss": 2.6023,
+      "step": 14823
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026730856826976963,
+      "loss": 2.5966,
+      "step": 14824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026734500051006144,
+      "loss": 2.3793,
+      "step": 14825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026738141491827267,
+      "loss": 2.6458,
+      "step": 14826
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674178114888695,
+      "loss": 2.3916,
+      "step": 14827
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674541902163212,
+      "loss": 2.8978,
+      "step": 14828
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026749055109509946,
+      "loss": 2.4896,
+      "step": 14829
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002675268941196787,
+      "loss": 2.1293,
+      "step": 14830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026756321928453634,
+      "loss": 2.8129,
+      "step": 14831
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026759952658415224,
+      "loss": 2.4695,
+      "step": 14832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026763581601300896,
+      "loss": 2.3961,
+      "step": 14833
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002676720875655917,
+      "loss": 2.5767,
+      "step": 14834
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026770834123638894,
+      "loss": 2.4605,
+      "step": 14835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002677445770198911,
+      "loss": 2.5673,
+      "step": 14836
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026778079491059195,
+      "loss": 2.4264,
+      "step": 14837
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678169949029875,
+      "loss": 2.7319,
+      "step": 14838
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026785317699157686,
+      "loss": 2.2008,
+      "step": 14839
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678893411708615,
+      "loss": 2.641,
+      "step": 14840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002679254874353462,
+      "loss": 1.3037,
+      "step": 14841
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026796161577953754,
+      "loss": 2.4393,
+      "step": 14842
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002679977261979458,
+      "loss": 2.3187,
+      "step": 14843
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002680338186850832,
+      "loss": 2.4286,
+      "step": 14844
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026806989323546537,
+      "loss": 2.7231,
+      "step": 14845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026810594984361,
+      "loss": 2.7,
+      "step": 14846
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026814198850403796,
+      "loss": 2.1281,
+      "step": 14847
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026817800921127274,
+      "loss": 2.6392,
+      "step": 14848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002682140119598406,
+      "loss": 2.4013,
+      "step": 14849
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002682499967442702,
+      "loss": 3.1382,
+      "step": 14850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002682859635590934,
+      "loss": 2.4209,
+      "step": 14851
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026832191239884454,
+      "loss": 2.3183,
+      "step": 14852
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002683578432580608,
+      "loss": 2.427,
+      "step": 14853
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002683937561312818,
+      "loss": 1.7371,
+      "step": 14854
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026842965101305037,
+      "loss": 2.221,
+      "step": 14855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002684655278979116,
+      "loss": 2.1101,
+      "step": 14856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002685013867804138,
+      "loss": 2.4982,
+      "step": 14857
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002685372276551075,
+      "loss": 2.9049,
+      "step": 14858
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002685730505165464,
+      "loss": 2.628,
+      "step": 14859
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026860885535928653,
+      "loss": 2.0938,
+      "step": 14860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002686446421778872,
+      "loss": 2.4894,
+      "step": 14861
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026868041096690995,
+      "loss": 2.6725,
+      "step": 14862
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026871616172091916,
+      "loss": 2.1079,
+      "step": 14863
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002687518944344822,
+      "loss": 2.0849,
+      "step": 14864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002687876091021691,
+      "loss": 2.7088,
+      "step": 14865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002688233057185524,
+      "loss": 2.1192,
+      "step": 14866
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002688589842782075,
+      "loss": 2.8117,
+      "step": 14867
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026889464477571274,
+      "loss": 2.6659,
+      "step": 14868
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000268930287205649,
+      "loss": 2.4591,
+      "step": 14869
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002689659115625999,
+      "loss": 2.6128,
+      "step": 14870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026900151784115184,
+      "loss": 2.7587,
+      "step": 14871
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002690371060358941,
+      "loss": 2.6875,
+      "step": 14872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026907267614141854,
+      "loss": 2.398,
+      "step": 14873
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026910822815231984,
+      "loss": 2.4153,
+      "step": 14874
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002691437620631953,
+      "loss": 2.2239,
+      "step": 14875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002691792778686453,
+      "loss": 2.2621,
+      "step": 14876
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002692147755632726,
+      "loss": 1.4485,
+      "step": 14877
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026925025514168297,
+      "loss": 2.5415,
+      "step": 14878
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002692857165984846,
+      "loss": 2.728,
+      "step": 14879
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000269321159928289,
+      "loss": 3.0324,
+      "step": 14880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002693565851257098,
+      "loss": 1.9373,
+      "step": 14881
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026939199218536404,
+      "loss": 2.7613,
+      "step": 14882
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002694273811018708,
+      "loss": 2.2033,
+      "step": 14883
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026946275186985245,
+      "loss": 2.6947,
+      "step": 14884
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002694981044839339,
+      "loss": 1.3539,
+      "step": 14885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026953343893874296,
+      "loss": 2.3331,
+      "step": 14886
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026956875522891,
+      "loss": 1.9791,
+      "step": 14887
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002696040533490683,
+      "loss": 2.1382,
+      "step": 14888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002696393332938538,
+      "loss": 2.5041,
+      "step": 14889
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002696745950579053,
+      "loss": 2.6935,
+      "step": 14890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026970983863586445,
+      "loss": 2.806,
+      "step": 14891
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026974506402237535,
+      "loss": 2.3315,
+      "step": 14892
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000269780271212085,
+      "loss": 2.2687,
+      "step": 14893
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026981546019964355,
+      "loss": 2.2292,
+      "step": 14894
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026985063097970325,
+      "loss": 2.6315,
+      "step": 14895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002698857835469195,
+      "loss": 2.5301,
+      "step": 14896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026992091789595054,
+      "loss": 2.7896,
+      "step": 14897
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026995603402145727,
+      "loss": 1.9227,
+      "step": 14898
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026999113191810315,
+      "loss": 2.2821,
+      "step": 14899
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002700262115805547,
+      "loss": 2.4806,
+      "step": 14900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002700612730034812,
+      "loss": 2.6529,
+      "step": 14901
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027009631618155457,
+      "loss": 2.2633,
+      "step": 14902
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002701313411094495,
+      "loss": 2.7379,
+      "step": 14903
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027016634778184343,
+      "loss": 2.8512,
+      "step": 14904
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002702013361934169,
+      "loss": 2.516,
+      "step": 14905
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027023630633885276,
+      "loss": 3.267,
+      "step": 14906
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002702712582128369,
+      "loss": 2.747,
+      "step": 14907
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027030619181005785,
+      "loss": 2.3108,
+      "step": 14908
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002703411071252072,
+      "loss": 2.9228,
+      "step": 14909
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027037600415297903,
+      "loss": 2.8923,
+      "step": 14910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002704108828880702,
+      "loss": 2.5375,
+      "step": 14911
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027044574332518045,
+      "loss": 2.6632,
+      "step": 14912
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027048058545901246,
+      "loss": 2.3837,
+      "step": 14913
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002705154092842714,
+      "loss": 2.4185,
+      "step": 14914
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002705502147956653,
+      "loss": 2.3197,
+      "step": 14915
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027058500198790496,
+      "loss": 1.9714,
+      "step": 14916
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027061977085570424,
+      "loss": 2.7311,
+      "step": 14917
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027065452139377944,
+      "loss": 2.8775,
+      "step": 14918
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027068925359684975,
+      "loss": 3.1691,
+      "step": 14919
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027072396745963715,
+      "loss": 2.8978,
+      "step": 14920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002707586629768665,
+      "loss": 2.5439,
+      "step": 14921
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027079334014326535,
+      "loss": 2.5285,
+      "step": 14922
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000270827998953564,
+      "loss": 2.7214,
+      "step": 14923
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002708626394024957,
+      "loss": 2.3372,
+      "step": 14924
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002708972614847962,
+      "loss": 2.2874,
+      "step": 14925
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002709318651952045,
+      "loss": 2.7927,
+      "step": 14926
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002709664505284619,
+      "loss": 2.715,
+      "step": 14927
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027100101747931285,
+      "loss": 2.3979,
+      "step": 14928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027103556604250425,
+      "loss": 3.0512,
+      "step": 14929
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002710700962127864,
+      "loss": 2.4508,
+      "step": 14930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027110460798491153,
+      "loss": 2.3931,
+      "step": 14931
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002711391013536355,
+      "loss": 1.8665,
+      "step": 14932
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002711735763137163,
+      "loss": 2.3981,
+      "step": 14933
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002712080328599153,
+      "loss": 2.2357,
+      "step": 14934
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027124247098699614,
+      "loss": 3.0396,
+      "step": 14935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027127689068972573,
+      "loss": 2.2578,
+      "step": 14936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027131129196287336,
+      "loss": 2.6927,
+      "step": 14937
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027134567480121153,
+      "loss": 2.8291,
+      "step": 14938
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027138003919951505,
+      "loss": 2.2153,
+      "step": 14939
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027141438515256196,
+      "loss": 2.3108,
+      "step": 14940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002714487126551329,
+      "loss": 2.7845,
+      "step": 14941
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002714830217020116,
+      "loss": 2.6843,
+      "step": 14942
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000271517312287984,
+      "loss": 2.3562,
+      "step": 14943
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027155158440783937,
+      "loss": 2.9239,
+      "step": 14944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027158583805636966,
+      "loss": 3.0007,
+      "step": 14945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002716200732283695,
+      "loss": 2.6215,
+      "step": 14946
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027165428991863654,
+      "loss": 2.5419,
+      "step": 14947
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002716884881219711,
+      "loss": 2.4116,
+      "step": 14948
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002717226678331761,
+      "loss": 1.8131,
+      "step": 14949
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002717568290470577,
+      "loss": 2.0589,
+      "step": 14950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002717909717584247,
+      "loss": 2.2752,
+      "step": 14951
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027182509596208843,
+      "loss": 2.5128,
+      "step": 14952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002718592016528636,
+      "loss": 2.6896,
+      "step": 14953
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027189328882556724,
+      "loss": 2.7762,
+      "step": 14954
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027189328882556724,
+      "loss": 4.5926,
+      "step": 14955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027192735747501935,
+      "loss": 2.3477,
+      "step": 14956
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002719614075960428,
+      "loss": 2.6579,
+      "step": 14957
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002719954391834632,
+      "loss": 2.5059,
+      "step": 14958
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027202945223210904,
+      "loss": 2.9328,
+      "step": 14959
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027206344673681173,
+      "loss": 2.199,
+      "step": 14960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002720974226924051,
+      "loss": 3.1871,
+      "step": 14961
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002721313800937263,
+      "loss": 2.525,
+      "step": 14962
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002721653189356149,
+      "loss": 1.9419,
+      "step": 14963
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002721992392129136,
+      "loss": 2.2741,
+      "step": 14964
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027223314092046766,
+      "loss": 2.6178,
+      "step": 14965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002722670240531254,
+      "loss": 2.5154,
+      "step": 14966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027230088860573766,
+      "loss": 2.6107,
+      "step": 14967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002723347345731586,
+      "loss": 2.9512,
+      "step": 14968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027236856195024447,
+      "loss": 2.2909,
+      "step": 14969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027240237073185514,
+      "loss": 3.0132,
+      "step": 14970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027243616091285263,
+      "loss": 2.4152,
+      "step": 14971
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002724699324881024,
+      "loss": 2.5576,
+      "step": 14972
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027250368545247213,
+      "loss": 2.3417,
+      "step": 14973
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002725374198008328,
+      "loss": 2.1615,
+      "step": 14974
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027257113552805795,
+      "loss": 2.1876,
+      "step": 14975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027260483262902423,
+      "loss": 2.4952,
+      "step": 14976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002726385110986106,
+      "loss": 2.0973,
+      "step": 14977
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002726721709316995,
+      "loss": 2.2094,
+      "step": 14978
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027270581212317566,
+      "loss": 2.5183,
+      "step": 14979
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002727394346679271,
+      "loss": 1.6052,
+      "step": 14980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027277303856084427,
+      "loss": 2.758,
+      "step": 14981
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002728066237968206,
+      "loss": 2.6293,
+      "step": 14982
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027284019037075247,
+      "loss": 2.3063,
+      "step": 14983
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000272873738277539,
+      "loss": 2.9008,
+      "step": 14984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027290726751208217,
+      "loss": 2.8302,
+      "step": 14985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002729407780692867,
+      "loss": 2.5709,
+      "step": 14986
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027297426994406026,
+      "loss": 2.7168,
+      "step": 14987
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002730077431313135,
+      "loss": 2.367,
+      "step": 14988
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002730411976259595,
+      "loss": 2.4791,
+      "step": 14989
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027307463342291443,
+      "loss": 2.4685,
+      "step": 14990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002731080505170975,
+      "loss": 3.0711,
+      "step": 14991
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002731414489034304,
+      "loss": 2.1165,
+      "step": 14992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002731748285768378,
+      "loss": 2.1927,
+      "step": 14993
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002732081895322472,
+      "loss": 2.4153,
+      "step": 14994
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002732415317645892,
+      "loss": 3.0963,
+      "step": 14995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002732748552687968,
+      "loss": 2.6074,
+      "step": 14996
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027330816003980616,
+      "loss": 2.6585,
+      "step": 14997
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002733414460725561,
+      "loss": 2.7044,
+      "step": 14998
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027337471336198846,
+      "loss": 2.5356,
+      "step": 14999
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027340796190304786,
+      "loss": 2.2374,
+      "step": 15000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027344119169068175,
+      "loss": 2.2799,
+      "step": 15001
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002734744027198403,
+      "loss": 2.3575,
+      "step": 15002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002735075949854769,
+      "loss": 2.7456,
+      "step": 15003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002735407684825474,
+      "loss": 1.1795,
+      "step": 15004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002735739232060107,
+      "loss": 2.7163,
+      "step": 15005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002736070591508284,
+      "loss": 2.8459,
+      "step": 15006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027364017631196537,
+      "loss": 2.1331,
+      "step": 15007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002736732746843888,
+      "loss": 3.0971,
+      "step": 15008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000273706354263069,
+      "loss": 2.4287,
+      "step": 15009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000273739415042979,
+      "loss": 2.812,
+      "step": 15010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000273772457019095,
+      "loss": 2.778,
+      "step": 15011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027380548018639576,
+      "loss": 1.6897,
+      "step": 15012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002738384845398629,
+      "loss": 2.1959,
+      "step": 15013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027387147007448115,
+      "loss": 2.5214,
+      "step": 15014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027390443678523786,
+      "loss": 2.3532,
+      "step": 15015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027393738466712327,
+      "loss": 2.561,
+      "step": 15016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027397031371513043,
+      "loss": 1.9024,
+      "step": 15017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002740032239242556,
+      "loss": 2.6711,
+      "step": 15018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002740361152894975,
+      "loss": 2.6908,
+      "step": 15019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002740689878058579,
+      "loss": 2.3707,
+      "step": 15020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027410184146834136,
+      "loss": 2.5437,
+      "step": 15021
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002741346762719554,
+      "loss": 2.7319,
+      "step": 15022
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027416749221171027,
+      "loss": 2.3628,
+      "step": 15023
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027420028928261934,
+      "loss": 1.9334,
+      "step": 15024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027423306747969845,
+      "loss": 2.2817,
+      "step": 15025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027426582679796667,
+      "loss": 2.736,
+      "step": 15026
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027429856723244566,
+      "loss": 1.7587,
+      "step": 15027
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002743312887781603,
+      "loss": 2.6139,
+      "step": 15028
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027436399143013803,
+      "loss": 2.5587,
+      "step": 15029
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027439667518340923,
+      "loss": 2.5654,
+      "step": 15030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027442934003300715,
+      "loss": 2.2836,
+      "step": 15031
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002744619859739681,
+      "loss": 1.1067,
+      "step": 15032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000274494613001331,
+      "loss": 2.9912,
+      "step": 15033
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027452722111013774,
+      "loss": 2.6266,
+      "step": 15034
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002745598102954331,
+      "loss": 3.1975,
+      "step": 15035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027459238055226477,
+      "loss": 2.4361,
+      "step": 15036
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027462493187568334,
+      "loss": 2.9333,
+      "step": 15037
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027465746426074214,
+      "loss": 2.6562,
+      "step": 15038
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002746899777024974,
+      "loss": 2.9514,
+      "step": 15039
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002747224721960084,
+      "loss": 2.3505,
+      "step": 15040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002747549477363371,
+      "loss": 2.8315,
+      "step": 15041
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002747874043185485,
+      "loss": 2.4645,
+      "step": 15042
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027481984193771034,
+      "loss": 2.8198,
+      "step": 15043
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002748522605888934,
+      "loss": 2.8843,
+      "step": 15044
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027488466026717116,
+      "loss": 2.495,
+      "step": 15045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027491704096762004,
+      "loss": 2.4274,
+      "step": 15046
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002749494026853194,
+      "loss": 2.9999,
+      "step": 15047
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002749817454153515,
+      "loss": 2.4862,
+      "step": 15048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002750140691528016,
+      "loss": 2.3905,
+      "step": 15049
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002750463738927574,
+      "loss": 2.341,
+      "step": 15050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027507865963030995,
+      "loss": 2.3366,
+      "step": 15051
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002751109263605529,
+      "loss": 2.355,
+      "step": 15052
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002751431740785832,
+      "loss": 2.0454,
+      "step": 15053
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027517540277949997,
+      "loss": 2.4532,
+      "step": 15054
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027520761245840595,
+      "loss": 2.7469,
+      "step": 15055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002752398031104063,
+      "loss": 2.1925,
+      "step": 15056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002752719747306094,
+      "loss": 2.9815,
+      "step": 15057
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002753041273141262,
+      "loss": 2.7727,
+      "step": 15058
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002753362608560708,
+      "loss": 2.8846,
+      "step": 15059
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027536837535155996,
+      "loss": 2.604,
+      "step": 15060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027540047079571375,
+      "loss": 2.2756,
+      "step": 15061
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027543254718365455,
+      "loss": 2.3913,
+      "step": 15062
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002754646045105081,
+      "loss": 2.6469,
+      "step": 15063
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002754966427714028,
+      "loss": 2.9789,
+      "step": 15064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002755286619614702,
+      "loss": 2.5914,
+      "step": 15065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027556066207584434,
+      "loss": 2.4594,
+      "step": 15066
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027559264310966254,
+      "loss": 1.1665,
+      "step": 15067
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002756246050580647,
+      "loss": 3.3214,
+      "step": 15068
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027565654791619407,
+      "loss": 2.6164,
+      "step": 15069
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027568847167919636,
+      "loss": 2.3171,
+      "step": 15070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002757203763422203,
+      "loss": 2.7532,
+      "step": 15071
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002757522619004176,
+      "loss": 2.6326,
+      "step": 15072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002757841283489429,
+      "loss": 2.6177,
+      "step": 15073
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027581597568295365,
+      "loss": 2.5512,
+      "step": 15074
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027584780389761017,
+      "loss": 2.5193,
+      "step": 15075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002758796129880759,
+      "loss": 3.4864,
+      "step": 15076
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002759114029495169,
+      "loss": 2.7869,
+      "step": 15077
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002759431737771024,
+      "loss": 3.2559,
+      "step": 15078
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027597492546600424,
+      "loss": 2.352,
+      "step": 15079
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002760066580113975,
+      "loss": 2.5293,
+      "step": 15080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027603837140845997,
+      "loss": 2.9852,
+      "step": 15081
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027607006565237243,
+      "loss": 2.7278,
+      "step": 15082
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002761017407383184,
+      "loss": 2.3939,
+      "step": 15083
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002761333966614846,
+      "loss": 3.1497,
+      "step": 15084
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002761650334170604,
+      "loss": 2.2955,
+      "step": 15085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027619665100023824,
+      "loss": 2.8055,
+      "step": 15086
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027622824940621335,
+      "loss": 2.644,
+      "step": 15087
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027625982863018405,
+      "loss": 2.9131,
+      "step": 15088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002762913886673514,
+      "loss": 2.4535,
+      "step": 15089
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002763229295129195,
+      "loss": 2.7594,
+      "step": 15090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027635445116209524,
+      "loss": 2.4826,
+      "step": 15091
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027638595361008855,
+      "loss": 2.4149,
+      "step": 15092
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002764174368521123,
+      "loss": 2.5726,
+      "step": 15093
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027644890088338205,
+      "loss": 2.6524,
+      "step": 15094
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002764803456991165,
+      "loss": 2.9318,
+      "step": 15095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027651177129453725,
+      "loss": 2.8547,
+      "step": 15096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002765431776648688,
+      "loss": 2.4383,
+      "step": 15097
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002765745648053385,
+      "loss": 2.7154,
+      "step": 15098
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002766059327111765,
+      "loss": 2.9274,
+      "step": 15099
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027663728137761634,
+      "loss": 2.0013,
+      "step": 15100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000276668610799894,
+      "loss": 2.2617,
+      "step": 15101
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002766999209732487,
+      "loss": 2.9584,
+      "step": 15102
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002767312118929224,
+      "loss": 2.4332,
+      "step": 15103
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027676248355416,
+      "loss": 3.24,
+      "step": 15104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027679373595220933,
+      "loss": 2.4721,
+      "step": 15105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027682496908232134,
+      "loss": 1.3718,
+      "step": 15106
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027685618293974963,
+      "loss": 2.4189,
+      "step": 15107
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027688737751975083,
+      "loss": 2.1319,
+      "step": 15108
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002769185528175847,
+      "loss": 2.7996,
+      "step": 15109
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002769497088285136,
+      "loss": 3.1573,
+      "step": 15110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000276980845547803,
+      "loss": 2.7297,
+      "step": 15111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002770119629707213,
+      "loss": 1.8739,
+      "step": 15112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027704306109253976,
+      "loss": 1.9452,
+      "step": 15113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002770741399085327,
+      "loss": 2.6334,
+      "step": 15114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002771051994139773,
+      "loss": 2.6282,
+      "step": 15115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027713623960415345,
+      "loss": 2.4606,
+      "step": 15116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002771672604743445,
+      "loss": 2.2261,
+      "step": 15117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002771982620198363,
+      "loss": 2.8802,
+      "step": 15118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002772292442359177,
+      "loss": 2.909,
+      "step": 15119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002772602071178806,
+      "loss": 2.9426,
+      "step": 15120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002772911506610199,
+      "loss": 2.2066,
+      "step": 15121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002773220748606332,
+      "loss": 2.5373,
+      "step": 15122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027735297971202117,
+      "loss": 3.1945,
+      "step": 15123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002773838652104875,
+      "loss": 2.4651,
+      "step": 15124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027741473135133867,
+      "loss": 2.7301,
+      "step": 15125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002774455781298842,
+      "loss": 2.7163,
+      "step": 15126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002774764055414366,
+      "loss": 2.8969,
+      "step": 15127
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002775072135813111,
+      "loss": 2.5607,
+      "step": 15128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002775380022448261,
+      "loss": 2.7319,
+      "step": 15129
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027756877152730294,
+      "loss": 2.5146,
+      "step": 15130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002775995214240656,
+      "loss": 2.7042,
+      "step": 15131
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002776302519304416,
+      "loss": 2.5722,
+      "step": 15132
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027766096304176075,
+      "loss": 2.7609,
+      "step": 15133
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027769165475335624,
+      "loss": 2.7902,
+      "step": 15134
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027772232706056396,
+      "loss": 2.666,
+      "step": 15135
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002777529799587229,
+      "loss": 2.7098,
+      "step": 15136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002777836134431751,
+      "loss": 2.6961,
+      "step": 15137
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002778142275092652,
+      "loss": 2.4281,
+      "step": 15138
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027784482215234107,
+      "loss": 2.408,
+      "step": 15139
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002778753973677536,
+      "loss": 2.8358,
+      "step": 15140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002779059531508562,
+      "loss": 2.4883,
+      "step": 15141
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002779364894970059,
+      "loss": 2.9533,
+      "step": 15142
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027796700640156196,
+      "loss": 2.1626,
+      "step": 15143
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002779975038598872,
+      "loss": 3.1164,
+      "step": 15144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002780279818673469,
+      "loss": 2.4448,
+      "step": 15145
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002780584404193098,
+      "loss": 2.9307,
+      "step": 15146
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002780888795111471,
+      "loss": 2.5747,
+      "step": 15147
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027811929913823335,
+      "loss": 2.2452,
+      "step": 15148
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002781496992959458,
+      "loss": 2.7363,
+      "step": 15149
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027818007997966487,
+      "loss": 2.8643,
+      "step": 15150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027821044118477364,
+      "loss": 2.8015,
+      "step": 15151
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002782407829066585,
+      "loss": 2.7441,
+      "step": 15152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002782711051407085,
+      "loss": 2.2559,
+      "step": 15153
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027830140788231603,
+      "loss": 2.4762,
+      "step": 15154
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002783316911268759,
+      "loss": 3.0053,
+      "step": 15155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027836195486978635,
+      "loss": 2.2458,
+      "step": 15156
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002783921991064483,
+      "loss": 2.0391,
+      "step": 15157
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027842242383226595,
+      "loss": 2.5365,
+      "step": 15158
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027845262904264615,
+      "loss": 2.1836,
+      "step": 15159
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027848281473299877,
+      "loss": 2.447,
+      "step": 15160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002785129808987367,
+      "loss": 2.762,
+      "step": 15161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002785431275352759,
+      "loss": 2.7704,
+      "step": 15162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027857325463803527,
+      "loss": 2.1186,
+      "step": 15163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002786033622024363,
+      "loss": 3.1021,
+      "step": 15164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002786334502239042,
+      "loss": 2.1164,
+      "step": 15165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027866351869786635,
+      "loss": 2.0588,
+      "step": 15166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002786935676197536,
+      "loss": 2.8856,
+      "step": 15167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027872359698499956,
+      "loss": 2.354,
+      "step": 15168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027875360678904096,
+      "loss": 2.768,
+      "step": 15169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027878359702731745,
+      "loss": 2.6853,
+      "step": 15170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027881356769527153,
+      "loss": 2.4758,
+      "step": 15171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002788435187883488,
+      "loss": 1.3712,
+      "step": 15172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002788734503019979,
+      "loss": 2.606,
+      "step": 15173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027890336223167016,
+      "loss": 2.287,
+      "step": 15174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002789332545728203,
+      "loss": 2.5521,
+      "step": 15175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027896312732090563,
+      "loss": 2.4993,
+      "step": 15176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027899298047138674,
+      "loss": 2.8509,
+      "step": 15177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002790228140197268,
+      "loss": 2.4166,
+      "step": 15178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027905262796139264,
+      "loss": 2.3274,
+      "step": 15179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027908242229185323,
+      "loss": 2.7594,
+      "step": 15180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002791121970065812,
+      "loss": 2.3237,
+      "step": 15181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027914195210105176,
+      "loss": 2.7057,
+      "step": 15182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027917168757074344,
+      "loss": 2.5478,
+      "step": 15183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002792014034111372,
+      "loss": 2.8303,
+      "step": 15184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002792310996177177,
+      "loss": 2.269,
+      "step": 15185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000279260776185972,
+      "loss": 2.4434,
+      "step": 15186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027929043311139047,
+      "loss": 2.7168,
+      "step": 15187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027932007038946623,
+      "loss": 2.5649,
+      "step": 15188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002793496880156957,
+      "loss": 2.2706,
+      "step": 15189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027937928598557795,
+      "loss": 2.1704,
+      "step": 15190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002794088642946153,
+      "loss": 2.9428,
+      "step": 15191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027943842293831286,
+      "loss": 2.7804,
+      "step": 15192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027946796191217886,
+      "loss": 2.6133,
+      "step": 15193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002794974812117244,
+      "loss": 2.1792,
+      "step": 15194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002795269808324638,
+      "loss": 2.9147,
+      "step": 15195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027955646076991406,
+      "loss": 2.2498,
+      "step": 15196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027958592101959533,
+      "loss": 2.7241,
+      "step": 15197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027961536157703083,
+      "loss": 2.2919,
+      "step": 15198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002796447824377467,
+      "loss": 2.6052,
+      "step": 15199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027967418359727204,
+      "loss": 2.6575,
+      "step": 15200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027970356505113887,
+      "loss": 3.0683,
+      "step": 15201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027973292679488236,
+      "loss": 2.5281,
+      "step": 15202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002797622688240407,
+      "loss": 2.5102,
+      "step": 15203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002797915911341549,
+      "loss": 1.7739,
+      "step": 15204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000279820893720769,
+      "loss": 2.3465,
+      "step": 15205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002798501765794303,
+      "loss": 2.2379,
+      "step": 15206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027987943970568866,
+      "loss": 2.5486,
+      "step": 15207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027990868309509735,
+      "loss": 3.1466,
+      "step": 15208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002799379067432124,
+      "loss": 3.1259,
+      "step": 15209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002799671106455928,
+      "loss": 2.5676,
+      "step": 15210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00027999629479780083,
+      "loss": 2.5287,
+      "step": 15211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002800254591954015,
+      "loss": 3.2912,
+      "step": 15212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028005460383396277,
+      "loss": 2.6235,
+      "step": 15213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028008372870905597,
+      "loss": 2.5048,
+      "step": 15214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028011283381625503,
+      "loss": 2.3824,
+      "step": 15215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002801419191511372,
+      "loss": 2.456,
+      "step": 15216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002801709847092823,
+      "loss": 2.3845,
+      "step": 15217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028020003048627386,
+      "loss": 2.7057,
+      "step": 15218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028022905647769774,
+      "loss": 2.5928,
+      "step": 15219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028025806267914315,
+      "loss": 2.7764,
+      "step": 15220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002802870490862021,
+      "loss": 2.637,
+      "step": 15221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002803160156944699,
+      "loss": 2.9384,
+      "step": 15222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002803449624995446,
+      "loss": 1.5771,
+      "step": 15223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028037388949702735,
+      "loss": 1.4674,
+      "step": 15224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028040279668252243,
+      "loss": 2.2314,
+      "step": 15225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028043168405163693,
+      "loss": 2.8752,
+      "step": 15226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002804605515999811,
+      "loss": 2.7597,
+      "step": 15227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000280489399323168,
+      "loss": 1.8353,
+      "step": 15228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028051822721681405,
+      "loss": 2.4713,
+      "step": 15229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028054703527653833,
+      "loss": 2.5895,
+      "step": 15230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028057582349796326,
+      "loss": 2.3046,
+      "step": 15231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028060459187671387,
+      "loss": 2.2352,
+      "step": 15232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002806333404084186,
+      "loss": 1.3242,
+      "step": 15233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028066206908870865,
+      "loss": 2.5253,
+      "step": 15234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002806907779132185,
+      "loss": 1.1685,
+      "step": 15235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028071946687758517,
+      "loss": 2.8677,
+      "step": 15236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028074813597744923,
+      "loss": 2.3354,
+      "step": 15237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028077678520845397,
+      "loss": 2.2956,
+      "step": 15238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028080541456624593,
+      "loss": 2.2954,
+      "step": 15239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002808340240464742,
+      "loss": 2.6041,
+      "step": 15240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002808626136447915,
+      "loss": 2.1456,
+      "step": 15241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028089118335685305,
+      "loss": 2.3701,
+      "step": 15242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002809197331783175,
+      "loss": 2.9856,
+      "step": 15243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028094826310484617,
+      "loss": 2.8141,
+      "step": 15244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002809767731321037,
+      "loss": 2.8148,
+      "step": 15245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028100526325575755,
+      "loss": 2.1133,
+      "step": 15246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002810337334714783,
+      "loss": 2.0361,
+      "step": 15247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002810621837749396,
+      "loss": 2.3502,
+      "step": 15248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002810906141618179,
+      "loss": 2.2853,
+      "step": 15249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000281119024627793,
+      "loss": 2.6739,
+      "step": 15250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028114741516854753,
+      "loss": 2.3782,
+      "step": 15251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028117578577976715,
+      "loss": 2.707,
+      "step": 15252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028120413645714047,
+      "loss": 2.6722,
+      "step": 15253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002812324671963595,
+      "loss": 2.4806,
+      "step": 15254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028126077799311885,
+      "loss": 3.0061,
+      "step": 15255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002812890688431163,
+      "loss": 2.5353,
+      "step": 15256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028131733974205275,
+      "loss": 2.5527,
+      "step": 15257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002813455906856321,
+      "loss": 2.3201,
+      "step": 15258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002813738216695612,
+      "loss": 2.5374,
+      "step": 15259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002814020326895501,
+      "loss": 3.115,
+      "step": 15260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028143022374131166,
+      "loss": 2.842,
+      "step": 15261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002814583948205619,
+      "loss": 2.4773,
+      "step": 15262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002814865459230198,
+      "loss": 2.8646,
+      "step": 15263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002815146770444077,
+      "loss": 2.5484,
+      "step": 15264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002815427881804504,
+      "loss": 2.5581,
+      "step": 15265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028157087932687626,
+      "loss": 2.3075,
+      "step": 15266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002815989504794163,
+      "loss": 2.5368,
+      "step": 15267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028162700163380497,
+      "loss": 2.1801,
+      "step": 15268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002816550327857793,
+      "loss": 2.6578,
+      "step": 15269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00028168304393107977,
+      "loss": 2.4021,
+      "step": 15270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002817110350654496,
+      "loss": 2.3069,
+      "step": 15271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002817390061846353,
+      "loss": 3.178,
+      "step": 15272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028176695728438617,
+      "loss": 2.7297,
+      "step": 15273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002817948883604549,
+      "loss": 2.6652,
+      "step": 15274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028182279940859665,
+      "loss": 3.001,
+      "step": 15275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002818506904245704,
+      "loss": 3.1053,
+      "step": 15276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002818785614041373,
+      "loss": 2.4352,
+      "step": 15277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002819064123430624,
+      "loss": 2.2988,
+      "step": 15278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028193424323711314,
+      "loss": 2.3736,
+      "step": 15279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002819620540820604,
+      "loss": 2.9524,
+      "step": 15280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002819898448736779,
+      "loss": 1.3395,
+      "step": 15281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028201761560774243,
+      "loss": 2.6613,
+      "step": 15282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028204536628003395,
+      "loss": 2.4554,
+      "step": 15283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002820730968863353,
+      "loss": 2.5394,
+      "step": 15284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002821008074224326,
+      "loss": 2.9673,
+      "step": 15285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002821284978841147,
+      "loss": 2.2036,
+      "step": 15286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028215616826717384,
+      "loss": 3.3769,
+      "step": 15287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002821838185674051,
+      "loss": 2.215,
+      "step": 15288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028221144878060663,
+      "loss": 2.7111,
+      "step": 15289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002822390589025796,
+      "loss": 2.4336,
+      "step": 15290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028226664892912846,
+      "loss": 2.3179,
+      "step": 15291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028229421885606044,
+      "loss": 2.5286,
+      "step": 15292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000282321768679186,
+      "loss": 1.3178,
+      "step": 15293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002823492983943184,
+      "loss": 2.8323,
+      "step": 15294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028237680799727443,
+      "loss": 1.3628,
+      "step": 15295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028240429748387346,
+      "loss": 2.8515,
+      "step": 15296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002824317668499382,
+      "loss": 3.1478,
+      "step": 15297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028245921609129426,
+      "loss": 1.911,
+      "step": 15298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028248664520377043,
+      "loss": 2.7141,
+      "step": 15299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028251405418319855,
+      "loss": 2.3022,
+      "step": 15300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002825414430254133,
+      "loss": 2.3223,
+      "step": 15301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028256881172625275,
+      "loss": 2.6972,
+      "step": 15302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002825961602815578,
+      "loss": 2.365,
+      "step": 15303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028262348868717256,
+      "loss": 3.0933,
+      "step": 15304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028265079693894405,
+      "loss": 1.7755,
+      "step": 15305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002826780850327224,
+      "loss": 2.3649,
+      "step": 15306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000282705352964361,
+      "loss": 2.2802,
+      "step": 15307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000282732600729716,
+      "loss": 2.4953,
+      "step": 15308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028275982832464677,
+      "loss": 2.2552,
+      "step": 15309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028278703574501567,
+      "loss": 2.6855,
+      "step": 15310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002828142229866883,
+      "loss": 2.8751,
+      "step": 15311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002828413900455332,
+      "loss": 2.4875,
+      "step": 15312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028286853691742185,
+      "loss": 2.7602,
+      "step": 15313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028289566359822907,
+      "loss": 2.2674,
+      "step": 15314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028292277008383265,
+      "loss": 2.6895,
+      "step": 15315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002829498563701133,
+      "loss": 2.7862,
+      "step": 15316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028297692245295484,
+      "loss": 1.7905,
+      "step": 15317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028300396832824445,
+      "loss": 2.5749,
+      "step": 15318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028303099399187197,
+      "loss": 2.1832,
+      "step": 15319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028305799943973065,
+      "loss": 2.5293,
+      "step": 15320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002830849846677166,
+      "loss": 2.7734,
+      "step": 15321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000283111949671729,
+      "loss": 2.2527,
+      "step": 15322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002831388944476702,
+      "loss": 2.2332,
+      "step": 15323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002831658189914458,
+      "loss": 2.4166,
+      "step": 15324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028319272329896394,
+      "loss": 2.4736,
+      "step": 15325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002832196073661364,
+      "loss": 2.907,
+      "step": 15326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002832464711888777,
+      "loss": 2.8208,
+      "step": 15327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028327331476310556,
+      "loss": 2.4256,
+      "step": 15328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002833001380847408,
+      "loss": 2.2358,
+      "step": 15329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002833269411497072,
+      "loss": 2.0038,
+      "step": 15330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028335372395393166,
+      "loss": 2.0232,
+      "step": 15331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028338048649334427,
+      "loss": 2.391,
+      "step": 15332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002834072287638782,
+      "loss": 2.6787,
+      "step": 15333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002834339507614694,
+      "loss": 2.6033,
+      "step": 15334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028346065248205723,
+      "loss": 2.4518,
+      "step": 15335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028348733392158403,
+      "loss": 2.7677,
+      "step": 15336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002835139950759953,
+      "loss": 2.0417,
+      "step": 15337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002835406359412394,
+      "loss": 2.7139,
+      "step": 15338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028356725651326785,
+      "loss": 2.4793,
+      "step": 15339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028359385678803544,
+      "loss": 1.927,
+      "step": 15340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028362043676149996,
+      "loss": 1.5269,
+      "step": 15341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028364699642962216,
+      "loss": 2.0392,
+      "step": 15342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028367353578836584,
+      "loss": 2.9078,
+      "step": 15343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002837000548336983,
+      "loss": 2.1406,
+      "step": 15344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028372655356158935,
+      "loss": 2.7836,
+      "step": 15345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002837530319680123,
+      "loss": 2.2585,
+      "step": 15346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002837794900489434,
+      "loss": 2.7922,
+      "step": 15347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000283805927800362,
+      "loss": 2.8058,
+      "step": 15348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028383234521825055,
+      "loss": 2.3415,
+      "step": 15349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002838587422985945,
+      "loss": 2.4846,
+      "step": 15350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028388511903738265,
+      "loss": 2.6885,
+      "step": 15351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028391147543060653,
+      "loss": 2.5086,
+      "step": 15352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002839378114742611,
+      "loss": 2.7235,
+      "step": 15353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028396412716434415,
+      "loss": 1.0051,
+      "step": 15354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028399042249685675,
+      "loss": 2.7264,
+      "step": 15355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002840166974678029,
+      "loss": 2.3968,
+      "step": 15356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002840429520731899,
+      "loss": 2.3901,
+      "step": 15357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002840691863090279,
+      "loss": 2.8858,
+      "step": 15358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028409540017133037,
+      "loss": 2.4148,
+      "step": 15359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028412159365611366,
+      "loss": 2.6649,
+      "step": 15360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028414776675939757,
+      "loss": 2.2313,
+      "step": 15361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002841739194772044,
+      "loss": 2.6603,
+      "step": 15362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028420005180556024,
+      "loss": 2.6811,
+      "step": 15363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002842261637404937,
+      "loss": 2.2928,
+      "step": 15364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000284252255278037,
+      "loss": 2.8853,
+      "step": 15365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002842783264142249,
+      "loss": 1.9945,
+      "step": 15366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028430437714509584,
+      "loss": 2.2116,
+      "step": 15367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028433040746669077,
+      "loss": 2.8947,
+      "step": 15368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002843564173750543,
+      "loss": 2.7208,
+      "step": 15369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028438240686623386,
+      "loss": 2.6153,
+      "step": 15370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028440837593627993,
+      "loss": 2.2395,
+      "step": 15371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002844343245812462,
+      "loss": 3.0118,
+      "step": 15372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028446025279718943,
+      "loss": 2.5182,
+      "step": 15373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002844861605801695,
+      "loss": 2.646,
+      "step": 15374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002845120479262494,
+      "loss": 3.1823,
+      "step": 15375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028453791483149524,
+      "loss": 2.6283,
+      "step": 15376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002845637612919762,
+      "loss": 3.5788,
+      "step": 15377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028458958730376456,
+      "loss": 2.5608,
+      "step": 15378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028461539286293565,
+      "loss": 1.393,
+      "step": 15379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028464117796556815,
+      "loss": 2.8564,
+      "step": 15380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002846669426077436,
+      "loss": 1.4789,
+      "step": 15381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002846926867855467,
+      "loss": 2.206,
+      "step": 15382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028471841049506523,
+      "loss": 1.8564,
+      "step": 15383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028474411373239034,
+      "loss": 2.8922,
+      "step": 15384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000284769796493616,
+      "loss": 2.902,
+      "step": 15385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002847954587748393,
+      "loss": 2.9006,
+      "step": 15386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002848211005721606,
+      "loss": 2.5799,
+      "step": 15387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028484672188168334,
+      "loss": 2.7216,
+      "step": 15388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028487232269951395,
+      "loss": 2.9384,
+      "step": 15389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002848979030217621,
+      "loss": 2.6679,
+      "step": 15390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002849234628445405,
+      "loss": 2.6246,
+      "step": 15391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002849490021639651,
+      "loss": 2.1021,
+      "step": 15392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002849745209761547,
+      "loss": 2.4806,
+      "step": 15393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028500001927723166,
+      "loss": 2.7088,
+      "step": 15394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028502549706332084,
+      "loss": 2.4774,
+      "step": 15395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028505095433055085,
+      "loss": 2.3762,
+      "step": 15396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002850763910750529,
+      "loss": 2.3092,
+      "step": 15397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002851018072929618,
+      "loss": 2.6626,
+      "step": 15398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028512720298041506,
+      "loss": 2.8997,
+      "step": 15399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002851525781335535,
+      "loss": 2.1833,
+      "step": 15400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028517793274852106,
+      "loss": 2.5554,
+      "step": 15401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002852032668214648,
+      "loss": 2.2515,
+      "step": 15402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002852285803485349,
+      "loss": 2.9728,
+      "step": 15403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028525387332588455,
+      "loss": 2.4893,
+      "step": 15404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002852791457496702,
+      "loss": 3.2223,
+      "step": 15405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002853043976160514,
+      "loss": 3.2055,
+      "step": 15406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028532962892119086,
+      "loss": 2.4365,
+      "step": 15407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002853548396612542,
+      "loss": 2.6405,
+      "step": 15408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002853800298324105,
+      "loss": 2.9354,
+      "step": 15409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002854051994308317,
+      "loss": 2.3258,
+      "step": 15410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028543034845269295,
+      "loss": 2.6385,
+      "step": 15411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002854554768941725,
+      "loss": 2.0121,
+      "step": 15412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002854805847514519,
+      "loss": 2.3451,
+      "step": 15413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002855056720207155,
+      "loss": 3.0279,
+      "step": 15414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028553073869815124,
+      "loss": 1.8073,
+      "step": 15415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002855557847799496,
+      "loss": 2.682,
+      "step": 15416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028558081026230464,
+      "loss": 1.1508,
+      "step": 15417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002856058151414135,
+      "loss": 2.8212,
+      "step": 15418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028563079941347627,
+      "loss": 2.3393,
+      "step": 15419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002856557630746962,
+      "loss": 2.3655,
+      "step": 15420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002856807061212799,
+      "loss": 2.7061,
+      "step": 15421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028570562854943687,
+      "loss": 2.3745,
+      "step": 15422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002857305303553799,
+      "loss": 2.4095,
+      "step": 15423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028575541153532476,
+      "loss": 2.3379,
+      "step": 15424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002857802720854904,
+      "loss": 2.4153,
+      "step": 15425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028580511200209903,
+      "loss": 2.9636,
+      "step": 15426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002858299312813759,
+      "loss": 1.0801,
+      "step": 15427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002858547299195493,
+      "loss": 2.8393,
+      "step": 15428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028587950791285094,
+      "loss": 2.8369,
+      "step": 15429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002859042652575153,
+      "loss": 2.7306,
+      "step": 15430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002859290019497803,
+      "loss": 2.598,
+      "step": 15431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002859537179858868,
+      "loss": 2.4189,
+      "step": 15432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000285978413362079,
+      "loss": 2.7306,
+      "step": 15433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028600308807460404,
+      "loss": 2.5962,
+      "step": 15434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002860277421197122,
+      "loss": 2.3052,
+      "step": 15435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002860523754936571,
+      "loss": 3.2869,
+      "step": 15436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028607698819269533,
+      "loss": 2.9417,
+      "step": 15437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028610158021308675,
+      "loss": 2.7756,
+      "step": 15438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028612615155109415,
+      "loss": 3.1929,
+      "step": 15439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028615070220298374,
+      "loss": 2.7094,
+      "step": 15440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002861752321650246,
+      "loss": 2.551,
+      "step": 15441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002861997414334893,
+      "loss": 2.5436,
+      "step": 15442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002862242300046531,
+      "loss": 2.4628,
+      "step": 15443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028624869787479475,
+      "loss": 2.3327,
+      "step": 15444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028627314504019603,
+      "loss": 2.636,
+      "step": 15445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000286297571497142,
+      "loss": 2.8107,
+      "step": 15446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002863219772419205,
+      "loss": 2.5417,
+      "step": 15447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028634636227082297,
+      "loss": 2.7729,
+      "step": 15448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028637072658014364,
+      "loss": 2.5899,
+      "step": 15449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028639507016618026,
+      "loss": 2.7094,
+      "step": 15450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028641939302523323,
+      "loss": 3.1459,
+      "step": 15451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028644369515360655,
+      "loss": 1.1022,
+      "step": 15452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002864679765476072,
+      "loss": 2.8183,
+      "step": 15453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028649223720354533,
+      "loss": 2.4357,
+      "step": 15454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028651647711773405,
+      "loss": 2.681,
+      "step": 15455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028654069628649,
+      "loss": 2.6043,
+      "step": 15456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002865648947061326,
+      "loss": 2.2732,
+      "step": 15457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002865890723729848,
+      "loss": 2.3477,
+      "step": 15458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028661322928337225,
+      "loss": 2.1518,
+      "step": 15459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028663736543362413,
+      "loss": 2.6548,
+      "step": 15460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028666148082007257,
+      "loss": 2.7846,
+      "step": 15461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002866855754390531,
+      "loss": 2.2717,
+      "step": 15462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000286709649286904,
+      "loss": 2.6853,
+      "step": 15463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002867337023599671,
+      "loss": 2.5846,
+      "step": 15464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028675773465458717,
+      "loss": 2.5007,
+      "step": 15465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002867817461671122,
+      "loss": 2.4056,
+      "step": 15466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028680573689389335,
+      "loss": 1.9125,
+      "step": 15467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028682970683128486,
+      "loss": 2.4131,
+      "step": 15468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028685365597564425,
+      "loss": 2.1669,
+      "step": 15469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002868775843233322,
+      "loss": 2.6831,
+      "step": 15470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028690149187071234,
+      "loss": 2.3704,
+      "step": 15471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028692537861415164,
+      "loss": 2.4439,
+      "step": 15472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002869492445500203,
+      "loss": 2.2602,
+      "step": 15473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002869730896746915,
+      "loss": 2.3673,
+      "step": 15474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002869969139845418,
+      "loss": 2.8747,
+      "step": 15475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028702071747595053,
+      "loss": 2.6204,
+      "step": 15476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002870445001453007,
+      "loss": 2.2981,
+      "step": 15477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002870682619889781,
+      "loss": 2.8483,
+      "step": 15478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002870920030033719,
+      "loss": 2.6616,
+      "step": 15479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028711572318487424,
+      "loss": 2.0083,
+      "step": 15480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028713942252988055,
+      "loss": 2.6763,
+      "step": 15481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002871631010347894,
+      "loss": 2.7756,
+      "step": 15482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028718675869600276,
+      "loss": 2.3004,
+      "step": 15483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028721039550992525,
+      "loss": 2.5748,
+      "step": 15484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002872340114729651,
+      "loss": 2.2351,
+      "step": 15485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028725760658153345,
+      "loss": 2.7312,
+      "step": 15486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028728118083204494,
+      "loss": 2.5097,
+      "step": 15487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028730473422091687,
+      "loss": 2.3908,
+      "step": 15488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002873282667445703,
+      "loss": 2.3483,
+      "step": 15489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002873517783994288,
+      "loss": 2.3596,
+      "step": 15490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028737526918191984,
+      "loss": 2.7981,
+      "step": 15491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002873987390884735,
+      "loss": 2.7556,
+      "step": 15492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002874221881155233,
+      "loss": 2.8813,
+      "step": 15493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028744561625950566,
+      "loss": 2.26,
+      "step": 15494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028746902351686065,
+      "loss": 2.3393,
+      "step": 15495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028749240988403115,
+      "loss": 2.7301,
+      "step": 15496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002875157753574632,
+      "loss": 2.5823,
+      "step": 15497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028753911993360624,
+      "loss": 3.0763,
+      "step": 15498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002875624436089127,
+      "loss": 2.4317,
+      "step": 15499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00028758574637983823,
+      "loss": 1.9681,
+      "step": 15500
+    }
+  ],
+  "max_steps": 67871,
+  "num_train_epochs": 1,
+  "total_flos": 7.5831026122752e+16,
+  "trial_name": null,
+  "trial_params": null
+}