diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,44422 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9998648648648648,
+  "eval_steps": 500,
+  "global_step": 7399,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.504504504504505e-08,
+      "loss": 2.3661,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.00900900900901e-08,
+      "loss": 2.5676,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3513513513513515e-07,
+      "loss": 1.2108,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.801801801801802e-07,
+      "loss": 1.3045,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2522522522522524e-07,
+      "loss": 1.1685,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.702702702702703e-07,
+      "loss": 1.179,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.153153153153153e-07,
+      "loss": 1.1451,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.603603603603604e-07,
+      "loss": 1.0036,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0540540540540546e-07,
+      "loss": 1.0644,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.504504504504505e-07,
+      "loss": 1.1227,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.954954954954956e-07,
+      "loss": 0.636,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.405405405405406e-07,
+      "loss": 1.1113,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.855855855855856e-07,
+      "loss": 1.0144,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.306306306306306e-07,
+      "loss": 0.8809,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.756756756756758e-07,
+      "loss": 1.1068,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.207207207207208e-07,
+      "loss": 0.6797,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.657657657657658e-07,
+      "loss": 0.6558,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.108108108108109e-07,
+      "loss": 1.1971,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.55855855855856e-07,
+      "loss": 1.036,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.00900900900901e-07,
+      "loss": 1.1018,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.459459459459461e-07,
+      "loss": 1.1454,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.909909909909911e-07,
+      "loss": 0.9797,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0360360360360361e-06,
+      "loss": 0.7351,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0810810810810812e-06,
+      "loss": 0.9473,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1261261261261262e-06,
+      "loss": 0.9839,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1711711711711712e-06,
+      "loss": 1.1603,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2162162162162164e-06,
+      "loss": 1.1415,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2612612612612613e-06,
+      "loss": 1.4417,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3063063063063065e-06,
+      "loss": 0.7886,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3513513513513515e-06,
+      "loss": 0.7584,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3963963963963963e-06,
+      "loss": 0.8425,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4414414414414416e-06,
+      "loss": 1.0159,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4864864864864868e-06,
+      "loss": 0.8701,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5315315315315316e-06,
+      "loss": 0.8484,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5765765765765766e-06,
+      "loss": 1.2398,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6216216216216219e-06,
+      "loss": 1.1301,
+      "step": 36
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.8964,
+      "step": 37
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.711711711711712e-06,
+      "loss": 1.0256,
+      "step": 38
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.756756756756757e-06,
+      "loss": 1.0027,
+      "step": 39
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.801801801801802e-06,
+      "loss": 0.8898,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.846846846846847e-06,
+      "loss": 1.0345,
+      "step": 41
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8918918918918922e-06,
+      "loss": 1.3515,
+      "step": 42
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9369369369369372e-06,
+      "loss": 1.2354,
+      "step": 43
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9819819819819822e-06,
+      "loss": 0.9391,
+      "step": 44
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0270270270270273e-06,
+      "loss": 1.0027,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0720720720720723e-06,
+      "loss": 0.9059,
+      "step": 46
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1171171171171173e-06,
+      "loss": 0.9469,
+      "step": 47
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1621621621621623e-06,
+      "loss": 1.0413,
+      "step": 48
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2072072072072073e-06,
+      "loss": 0.6037,
+      "step": 49
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2522522522522524e-06,
+      "loss": 0.6035,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.297297297297298e-06,
+      "loss": 0.7158,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3423423423423424e-06,
+      "loss": 1.1222,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3873873873873874e-06,
+      "loss": 0.5705,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.432432432432433e-06,
+      "loss": 0.4162,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4774774774774775e-06,
+      "loss": 0.8682,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5225225225225225e-06,
+      "loss": 1.1469,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5675675675675675e-06,
+      "loss": 1.273,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.612612612612613e-06,
+      "loss": 0.9449,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.657657657657658e-06,
+      "loss": 0.8914,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.702702702702703e-06,
+      "loss": 1.2056,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.747747747747748e-06,
+      "loss": 1.0589,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7927927927927926e-06,
+      "loss": 0.811,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.837837837837838e-06,
+      "loss": 0.8507,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.882882882882883e-06,
+      "loss": 0.8138,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.927927927927928e-06,
+      "loss": 1.2774,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9729729729729736e-06,
+      "loss": 1.0224,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0180180180180186e-06,
+      "loss": 0.8997,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.063063063063063e-06,
+      "loss": 0.756,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1081081081081082e-06,
+      "loss": 0.9939,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1531531531531532e-06,
+      "loss": 1.1182,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1981981981981987e-06,
+      "loss": 0.8227,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2432432432432437e-06,
+      "loss": 0.6076,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2882882882882887e-06,
+      "loss": 0.5397,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.9547,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3783783783783788e-06,
+      "loss": 1.1415,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.423423423423424e-06,
+      "loss": 1.0939,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.468468468468469e-06,
+      "loss": 0.8292,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.513513513513514e-06,
+      "loss": 0.5891,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5585585585585584e-06,
+      "loss": 1.2408,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.603603603603604e-06,
+      "loss": 0.6851,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.648648648648649e-06,
+      "loss": 0.761,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.693693693693694e-06,
+      "loss": 0.9259,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7387387387387394e-06,
+      "loss": 0.8678,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7837837837837844e-06,
+      "loss": 0.8562,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.828828828828829e-06,
+      "loss": 0.9589,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8738738738738744e-06,
+      "loss": 0.7856,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.918918918918919e-06,
+      "loss": 0.8,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9639639639639645e-06,
+      "loss": 0.7404,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.009009009009009e-06,
+      "loss": 0.8665,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0540540540540545e-06,
+      "loss": 0.8353,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.099099099099099e-06,
+      "loss": 0.9005,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1441441441441446e-06,
+      "loss": 1.0996,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.189189189189189e-06,
+      "loss": 0.8704,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.234234234234235e-06,
+      "loss": 0.977,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.27927927927928e-06,
+      "loss": 0.9735,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.324324324324325e-06,
+      "loss": 0.9378,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.369369369369369e-06,
+      "loss": 1.3121,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.414414414414415e-06,
+      "loss": 0.684,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.45945945945946e-06,
+      "loss": 0.9379,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.504504504504505e-06,
+      "loss": 0.9408,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.54954954954955e-06,
+      "loss": 1.3406,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.594594594594596e-06,
+      "loss": 0.6821,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.63963963963964e-06,
+      "loss": 0.675,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.684684684684685e-06,
+      "loss": 1.2011,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.72972972972973e-06,
+      "loss": 0.9231,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.774774774774775e-06,
+      "loss": 0.6436,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.81981981981982e-06,
+      "loss": 1.0293,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.864864864864866e-06,
+      "loss": 0.996,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.90990990990991e-06,
+      "loss": 1.1967,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.954954954954955e-06,
+      "loss": 0.7445,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 0.8975,
+      "step": 111
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.045045045045045e-06,
+      "loss": 1.065,
+      "step": 112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0900900900900905e-06,
+      "loss": 1.154,
+      "step": 113
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.135135135135135e-06,
+      "loss": 0.6337,
+      "step": 114
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.180180180180181e-06,
+      "loss": 0.7103,
+      "step": 115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.225225225225226e-06,
+      "loss": 0.5873,
+      "step": 116
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2702702702702705e-06,
+      "loss": 0.9756,
+      "step": 117
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.315315315315316e-06,
+      "loss": 1.0901,
+      "step": 118
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.360360360360361e-06,
+      "loss": 0.7765,
+      "step": 119
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.405405405405406e-06,
+      "loss": 0.8298,
+      "step": 120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.450450450450451e-06,
+      "loss": 1.262,
+      "step": 121
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.495495495495496e-06,
+      "loss": 1.1442,
+      "step": 122
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.540540540540541e-06,
+      "loss": 0.9434,
+      "step": 123
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.585585585585585e-06,
+      "loss": 0.8778,
+      "step": 124
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6306306306306316e-06,
+      "loss": 0.8167,
+      "step": 125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.675675675675676e-06,
+      "loss": 0.7556,
+      "step": 126
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.720720720720722e-06,
+      "loss": 1.0321,
+      "step": 127
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.765765765765766e-06,
+      "loss": 0.765,
+      "step": 128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.810810810810811e-06,
+      "loss": 0.582,
+      "step": 129
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.855855855855856e-06,
+      "loss": 0.8868,
+      "step": 130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.900900900900901e-06,
+      "loss": 0.7921,
+      "step": 131
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.945945945945947e-06,
+      "loss": 1.2418,
+      "step": 132
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.990990990990992e-06,
+      "loss": 1.2653,
+      "step": 133
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.036036036036037e-06,
+      "loss": 0.6142,
+      "step": 134
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.081081081081082e-06,
+      "loss": 0.87,
+      "step": 135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.126126126126126e-06,
+      "loss": 0.4104,
+      "step": 136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.171171171171172e-06,
+      "loss": 1.0608,
+      "step": 137
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.2162162162162164e-06,
+      "loss": 0.847,
+      "step": 138
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.261261261261262e-06,
+      "loss": 1.0614,
+      "step": 139
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3063063063063065e-06,
+      "loss": 1.1554,
+      "step": 140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.351351351351351e-06,
+      "loss": 1.0325,
+      "step": 141
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.396396396396397e-06,
+      "loss": 1.0874,
+      "step": 142
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.441441441441442e-06,
+      "loss": 1.1499,
+      "step": 143
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.486486486486487e-06,
+      "loss": 0.5193,
+      "step": 144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.531531531531532e-06,
+      "loss": 0.8233,
+      "step": 145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.5765765765765775e-06,
+      "loss": 1.3085,
+      "step": 146
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.621621621621622e-06,
+      "loss": 1.0461,
+      "step": 147
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.9435,
+      "step": 148
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.711711711711713e-06,
+      "loss": 1.2594,
+      "step": 149
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.7567567567567575e-06,
+      "loss": 1.2857,
+      "step": 150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.801801801801803e-06,
+      "loss": 0.8672,
+      "step": 151
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.846846846846848e-06,
+      "loss": 0.7381,
+      "step": 152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.891891891891892e-06,
+      "loss": 0.8319,
+      "step": 153
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.936936936936938e-06,
+      "loss": 1.0941,
+      "step": 154
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.981981981981982e-06,
+      "loss": 0.8002,
+      "step": 155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.027027027027028e-06,
+      "loss": 0.933,
+      "step": 156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.072072072072072e-06,
+      "loss": 0.9462,
+      "step": 157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.117117117117117e-06,
+      "loss": 0.6567,
+      "step": 158
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.162162162162163e-06,
+      "loss": 1.2412,
+      "step": 159
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.207207207207208e-06,
+      "loss": 0.8509,
+      "step": 160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.252252252252253e-06,
+      "loss": 0.7724,
+      "step": 161
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.297297297297298e-06,
+      "loss": 0.9894,
+      "step": 162
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.342342342342343e-06,
+      "loss": 0.9354,
+      "step": 163
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.387387387387388e-06,
+      "loss": 1.2295,
+      "step": 164
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.4324324324324324e-06,
+      "loss": 0.8978,
+      "step": 165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.477477477477479e-06,
+      "loss": 0.6967,
+      "step": 166
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.5225225225225225e-06,
+      "loss": 0.5903,
+      "step": 167
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.567567567567569e-06,
+      "loss": 1.028,
+      "step": 168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.612612612612613e-06,
+      "loss": 1.0248,
+      "step": 169
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.657657657657658e-06,
+      "loss": 0.6847,
+      "step": 170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.702702702702704e-06,
+      "loss": 0.3831,
+      "step": 171
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.747747747747749e-06,
+      "loss": 0.9839,
+      "step": 172
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.792792792792793e-06,
+      "loss": 0.858,
+      "step": 173
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.837837837837838e-06,
+      "loss": 1.0915,
+      "step": 174
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.882882882882884e-06,
+      "loss": 1.2199,
+      "step": 175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.927927927927929e-06,
+      "loss": 1.0432,
+      "step": 176
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.972972972972974e-06,
+      "loss": 0.6438,
+      "step": 177
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.018018018018018e-06,
+      "loss": 0.3585,
+      "step": 178
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.063063063063063e-06,
+      "loss": 1.2815,
+      "step": 179
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.108108108108109e-06,
+      "loss": 1.3676,
+      "step": 180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.153153153153154e-06,
+      "loss": 0.7867,
+      "step": 181
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.198198198198198e-06,
+      "loss": 0.6443,
+      "step": 182
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.243243243243245e-06,
+      "loss": 1.2118,
+      "step": 183
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.288288288288289e-06,
+      "loss": 1.1266,
+      "step": 184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.7312,
+      "step": 185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.378378378378378e-06,
+      "loss": 1.1398,
+      "step": 186
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.423423423423423e-06,
+      "loss": 1.2432,
+      "step": 187
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.46846846846847e-06,
+      "loss": 1.056,
+      "step": 188
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.513513513513514e-06,
+      "loss": 0.6655,
+      "step": 189
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.55855855855856e-06,
+      "loss": 1.1682,
+      "step": 190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.603603603603605e-06,
+      "loss": 0.7053,
+      "step": 191
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.64864864864865e-06,
+      "loss": 1.1364,
+      "step": 192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.693693693693694e-06,
+      "loss": 0.9025,
+      "step": 193
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.738738738738739e-06,
+      "loss": 0.8124,
+      "step": 194
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.783783783783785e-06,
+      "loss": 0.7807,
+      "step": 195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.82882882882883e-06,
+      "loss": 0.7725,
+      "step": 196
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.873873873873876e-06,
+      "loss": 0.6609,
+      "step": 197
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.91891891891892e-06,
+      "loss": 0.7982,
+      "step": 198
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.963963963963965e-06,
+      "loss": 0.7813,
+      "step": 199
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.00900900900901e-06,
+      "loss": 1.2392,
+      "step": 200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.054054054054054e-06,
+      "loss": 1.1609,
+      "step": 201
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.0990990990991e-06,
+      "loss": 0.998,
+      "step": 202
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.144144144144145e-06,
+      "loss": 0.8867,
+      "step": 203
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.189189189189191e-06,
+      "loss": 0.8497,
+      "step": 204
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.234234234234236e-06,
+      "loss": 0.8486,
+      "step": 205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.27927927927928e-06,
+      "loss": 0.7179,
+      "step": 206
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.324324324324325e-06,
+      "loss": 1.1519,
+      "step": 207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.36936936936937e-06,
+      "loss": 0.9863,
+      "step": 208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.414414414414416e-06,
+      "loss": 0.7535,
+      "step": 209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.45945945945946e-06,
+      "loss": 0.6621,
+      "step": 210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.504504504504505e-06,
+      "loss": 1.1324,
+      "step": 211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.54954954954955e-06,
+      "loss": 1.049,
+      "step": 212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.594594594594594e-06,
+      "loss": 0.9393,
+      "step": 213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.63963963963964e-06,
+      "loss": 0.7691,
+      "step": 214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.684684684684685e-06,
+      "loss": 0.6841,
+      "step": 215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.729729729729732e-06,
+      "loss": 0.8068,
+      "step": 216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.774774774774776e-06,
+      "loss": 0.8836,
+      "step": 217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.81981981981982e-06,
+      "loss": 0.8931,
+      "step": 218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.864864864864865e-06,
+      "loss": 1.0101,
+      "step": 219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90990990990991e-06,
+      "loss": 0.6075,
+      "step": 220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.954954954954956e-06,
+      "loss": 0.8576,
+      "step": 221
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.9799,
+      "step": 222
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999952111323e-06,
+      "loss": 0.8809,
+      "step": 223
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998084453014e-06,
+      "loss": 0.8731,
+      "step": 224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999995690019621e-06,
+      "loss": 1.007,
+      "step": 225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999992337813516e-06,
+      "loss": 0.8706,
+      "step": 226
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999988027835339e-06,
+      "loss": 0.8839,
+      "step": 227
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999982760085913e-06,
+      "loss": 1.15,
+      "step": 228
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99997653456625e-06,
+      "loss": 1.1686,
+      "step": 229
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999969351277542e-06,
+      "loss": 0.6826,
+      "step": 230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999961210221168e-06,
+      "loss": 0.784,
+      "step": 231
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999952111398681e-06,
+      "loss": 0.5292,
+      "step": 232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99994205481183e-06,
+      "loss": 1.0509,
+      "step": 233
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999931040462536e-06,
+      "loss": 1.1286,
+      "step": 234
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999919068352912e-06,
+      "loss": 0.9813,
+      "step": 235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99990613848525e-06,
+      "loss": 0.924,
+      "step": 236
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999892250862032e-06,
+      "loss": 0.8261,
+      "step": 237
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99987740548591e-06,
+      "loss": 1.0068,
+      "step": 238
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999861602359732e-06,
+      "loss": 0.5444,
+      "step": 239
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999844841486525e-06,
+      "loss": 1.0323,
+      "step": 240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999827122869499e-06,
+      "loss": 0.9053,
+      "step": 241
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99980844651205e-06,
+      "loss": 1.0402,
+      "step": 242
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999788812417752e-06,
+      "loss": 0.8084,
+      "step": 243
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99976822059037e-06,
+      "loss": 0.9661,
+      "step": 244
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999746671033845e-06,
+      "loss": 1.0094,
+      "step": 245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999724163752307e-06,
+      "loss": 0.8688,
+      "step": 246
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999700698750067e-06,
+      "loss": 1.4656,
+      "step": 247
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99967627603162e-06,
+      "loss": 1.096,
+      "step": 248
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999650895601642e-06,
+      "loss": 0.7428,
+      "step": 249
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999624557464997e-06,
+      "loss": 0.6854,
+      "step": 250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99959726162673e-06,
+      "loss": 0.8097,
+      "step": 251
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999569008092072e-06,
+      "loss": 1.3442,
+      "step": 252
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999539796866428e-06,
+      "loss": 1.1093,
+      "step": 253
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999509627955402e-06,
+      "loss": 0.8681,
+      "step": 254
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999478501364768e-06,
+      "loss": 0.6198,
+      "step": 255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999446417100489e-06,
+      "loss": 0.5527,
+      "step": 256
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999413375168712e-06,
+      "loss": 0.9976,
+      "step": 257
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999379375575764e-06,
+      "loss": 0.9245,
+      "step": 258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999344418328161e-06,
+      "loss": 0.6854,
+      "step": 259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999308503432599e-06,
+      "loss": 1.1792,
+      "step": 260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999271630895955e-06,
+      "loss": 1.2695,
+      "step": 261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999233800725293e-06,
+      "loss": 0.7401,
+      "step": 262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999195012927861e-06,
+      "loss": 0.9438,
+      "step": 263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999155267511087e-06,
+      "loss": 1.0141,
+      "step": 264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999114564482587e-06,
+      "loss": 0.533,
+      "step": 265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999072903850156e-06,
+      "loss": 1.1448,
+      "step": 266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999030285621774e-06,
+      "loss": 0.6979,
+      "step": 267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998986709805604e-06,
+      "loss": 1.06,
+      "step": 268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998942176409998e-06,
+      "loss": 1.0242,
+      "step": 269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998896685443482e-06,
+      "loss": 1.069,
+      "step": 270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99885023691477e-06,
+      "loss": 0.9133,
+      "step": 271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998802830832761e-06,
+      "loss": 0.5658,
+      "step": 272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998754467206535e-06,
+      "loss": 1.024,
+      "step": 273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998705146045358e-06,
+      "loss": 0.8497,
+      "step": 274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998654867358674e-06,
+      "loss": 0.9917,
+      "step": 275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998603631156119e-06,
+      "loss": 0.7124,
+      "step": 276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998551437447503e-06,
+      "loss": 0.6631,
+      "step": 277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998498286242828e-06,
+      "loss": 0.931,
+      "step": 278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998444177552271e-06,
+      "loss": 0.4909,
+      "step": 279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998389111386202e-06,
+      "loss": 0.6657,
+      "step": 280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998333087755164e-06,
+      "loss": 0.8335,
+      "step": 281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998276106669891e-06,
+      "loss": 0.9418,
+      "step": 282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9982181681413e-06,
+      "loss": 0.8933,
+      "step": 283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998159272180485e-06,
+      "loss": 0.7779,
+      "step": 284
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998099418798731e-06,
+      "loss": 0.6149,
+      "step": 285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998038608007504e-06,
+      "loss": 0.7603,
+      "step": 286
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997976839818448e-06,
+      "loss": 0.55,
+      "step": 287
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997914114243399e-06,
+      "loss": 1.0235,
+      "step": 288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997850431294372e-06,
+      "loss": 1.1367,
+      "step": 289
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997785790983563e-06,
+      "loss": 0.9331,
+      "step": 290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997720193323358e-06,
+      "loss": 0.7656,
+      "step": 291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99765363832632e-06,
+      "loss": 1.063,
+      "step": 292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997586126005198e-06,
+      "loss": 0.8746,
+      "step": 293
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997517656372928e-06,
+      "loss": 1.0483,
+      "step": 294
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99744822944262e-06,
+      "loss": 0.7816,
+      "step": 295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997377845227577e-06,
+      "loss": 0.9899,
+      "step": 296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997306503741278e-06,
+      "loss": 1.044,
+      "step": 297
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997234204997392e-06,
+      "loss": 1.1253,
+      "step": 298
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997160949009768e-06,
+      "loss": 1.0866,
+      "step": 299
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997086735792437e-06,
+      "loss": 0.9057,
+      "step": 300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997011565359613e-06,
+      "loss": 0.9395,
+      "step": 301
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996935437725702e-06,
+      "loss": 0.8881,
+      "step": 302
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996858352905279e-06,
+      "loss": 0.9935,
+      "step": 303
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996780310913114e-06,
+      "loss": 0.9663,
+      "step": 304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996701311764157e-06,
+      "loss": 0.8346,
+      "step": 305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996621355473539e-06,
+      "loss": 0.7108,
+      "step": 306
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996540442056576e-06,
+      "loss": 0.6246,
+      "step": 307
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996458571528766e-06,
+      "loss": 1.0693,
+      "step": 308
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996375743905796e-06,
+      "loss": 1.0817,
+      "step": 309
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996291959203528e-06,
+      "loss": 0.9912,
+      "step": 310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996207217438013e-06,
+      "loss": 1.1382,
+      "step": 311
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996121518625483e-06,
+      "loss": 0.8297,
+      "step": 312
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996034862782354e-06,
+      "loss": 1.2069,
+      "step": 313
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995947249925227e-06,
+      "loss": 0.7824,
+      "step": 314
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995858680070882e-06,
+      "loss": 1.1634,
+      "step": 315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995769153236287e-06,
+      "loss": 1.0764,
+      "step": 316
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995678669438591e-06,
+      "loss": 0.6971,
+      "step": 317
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995587228695126e-06,
+      "loss": 0.8335,
+      "step": 318
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99549483102341e-06,
+      "loss": 0.9172,
+      "step": 319
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995401476441137e-06,
+      "loss": 0.9939,
+      "step": 320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995307164966194e-06,
+      "loss": 0.7158,
+      "step": 321
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995211896616645e-06,
+      "loss": 0.689,
+      "step": 322
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995115671410742e-06,
+      "loss": 0.7376,
+      "step": 323
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995018489366911e-06,
+      "loss": 1.0195,
+      "step": 324
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994920350503776e-06,
+      "loss": 1.0317,
+      "step": 325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99482125484013e-06,
+      "loss": 1.0146,
+      "step": 326
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994721202394959e-06,
+      "loss": 0.7935,
+      "step": 327
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994620193187425e-06,
+      "loss": 1.0922,
+      "step": 328
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994518227236878e-06,
+      "loss": 0.6365,
+      "step": 329
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994415304562851e-06,
+      "loss": 0.6641,
+      "step": 330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994311425185058e-06,
+      "loss": 1.1254,
+      "step": 331
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994206589123398e-06,
+      "loss": 0.4834,
+      "step": 332
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994100796397954e-06,
+      "loss": 1.0118,
+      "step": 333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99399404702899e-06,
+      "loss": 1.1223,
+      "step": 334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993886341036957e-06,
+      "loss": 1.2301,
+      "step": 335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99377767844248e-06,
+      "loss": 0.8783,
+      "step": 336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99366805926638e-06,
+      "loss": 0.9033,
+      "step": 337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993557483529652e-06,
+      "loss": 1.112,
+      "step": 338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99344595125348e-06,
+      "loss": 0.9553,
+      "step": 339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993333462459227e-06,
+      "loss": 0.6181,
+      "step": 340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993220017168439e-06,
+      "loss": 0.8756,
+      "step": 341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99310561540285e-06,
+      "loss": 0.7402,
+      "step": 342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992990257184372e-06,
+      "loss": 1.2409,
+      "step": 343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992873942535104e-06,
+      "loss": 0.7908,
+      "step": 344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992756671477325e-06,
+      "loss": 0.7307,
+      "step": 345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9926384440335e-06,
+      "loss": 1.3645,
+      "step": 346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992519260226277e-06,
+      "loss": 1.0515,
+      "step": 347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992399120078484e-06,
+      "loss": 0.8336,
+      "step": 348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992278023613136e-06,
+      "loss": 0.9308,
+      "step": 349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992155970853427e-06,
+      "loss": 1.1748,
+      "step": 350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99203296182274e-06,
+      "loss": 1.166,
+      "step": 351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991908996544637e-06,
+      "loss": 0.6622,
+      "step": 352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991784075042863e-06,
+      "loss": 0.6321,
+      "step": 353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991658197341348e-06,
+      "loss": 1.3269,
+      "step": 354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991531363464207e-06,
+      "loss": 0.595,
+      "step": 355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991403573435729e-06,
+      "loss": 0.705,
+      "step": 356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9912748272804e-06,
+      "loss": 1.1594,
+      "step": 357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99114512502288e-06,
+      "loss": 0.3578,
+      "step": 358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99101446668801e-06,
+      "loss": 0.9165,
+      "step": 359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990882852300823e-06,
+      "loss": 1.251,
+      "step": 360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990750281886528e-06,
+      "loss": 0.6494,
+      "step": 361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99061675547052e-06,
+      "loss": 0.7723,
+      "step": 362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990482273078377e-06,
+      "loss": 1.0315,
+      "step": 363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990346834735857e-06,
+      "loss": 0.9368,
+      "step": 364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99021044046891e-06,
+      "loss": 1.2093,
+      "step": 365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990073090303659e-06,
+      "loss": 0.5562,
+      "step": 366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989934784266412e-06,
+      "loss": 0.7102,
+      "step": 367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989795522383665e-06,
+      "loss": 1.0579,
+      "step": 368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989655304682095e-06,
+      "loss": 0.954,
+      "step": 369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98951413118856e-06,
+      "loss": 0.9138,
+      "step": 370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989372001930101e-06,
+      "loss": 0.689,
+      "step": 371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989228916933946e-06,
+      "loss": 1.1491,
+      "step": 372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989084876227503e-06,
+      "loss": 0.8058,
+      "step": 373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988939879838362e-06,
+      "loss": 0.8375,
+      "step": 374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9887939277943e-06,
+      "loss": 0.8861,
+      "step": 375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988647020123275e-06,
+      "loss": 1.3143,
+      "step": 376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988499156853426e-06,
+      "loss": 0.8301,
+      "step": 377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988350338013077e-06,
+      "loss": 0.8188,
+      "step": 378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988200563630736e-06,
+      "loss": 0.6681,
+      "step": 379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988049833735095e-06,
+      "loss": 1.0567,
+      "step": 380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987898148355021e-06,
+      "loss": 0.783,
+      "step": 381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987745507519574e-06,
+      "loss": 0.6519,
+      "step": 382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987591911257995e-06,
+      "loss": 0.6769,
+      "step": 383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987437359599703e-06,
+      "loss": 0.9492,
+      "step": 384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987281852574303e-06,
+      "loss": 0.8827,
+      "step": 385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987125390211583e-06,
+      "loss": 1.0856,
+      "step": 386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986967972541516e-06,
+      "loss": 0.9668,
+      "step": 387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986809599594258e-06,
+      "loss": 1.0009,
+      "step": 388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986650271400141e-06,
+      "loss": 0.8896,
+      "step": 389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986489987989687e-06,
+      "loss": 1.0889,
+      "step": 390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9863287493936e-06,
+      "loss": 1.1375,
+      "step": 391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986166555642766e-06,
+      "loss": 0.6856,
+      "step": 392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986003406768253e-06,
+      "loss": 0.89,
+      "step": 393
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985839302801314e-06,
+      "loss": 0.597,
+      "step": 394
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985674243773382e-06,
+      "loss": 1.0499,
+      "step": 395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985508229716077e-06,
+      "loss": 0.81,
+      "step": 396
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985341260661197e-06,
+      "loss": 0.7918,
+      "step": 397
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985173336640729e-06,
+      "loss": 1.0794,
+      "step": 398
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98500445768684e-06,
+      "loss": 0.5217,
+      "step": 399
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984834623831874e-06,
+      "loss": 1.1298,
+      "step": 400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98466383510837e-06,
+      "loss": 0.7197,
+      "step": 401
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98449209154904e-06,
+      "loss": 1.2182,
+      "step": 402
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98431939318678e-06,
+      "loss": 0.7866,
+      "step": 403
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984145740054678e-06,
+      "loss": 0.9485,
+      "step": 404
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983971132185994e-06,
+      "loss": 0.8167,
+      "step": 405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983795569614174e-06,
+      "loss": 0.6907,
+      "step": 406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983619052372847e-06,
+      "loss": 0.8638,
+      "step": 407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98344158049583e-06,
+      "loss": 0.6064,
+      "step": 408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983263154017116e-06,
+      "loss": 0.8521,
+      "step": 409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983083772970883e-06,
+      "loss": 0.9409,
+      "step": 410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982903437391494e-06,
+      "loss": 0.8506,
+      "step": 411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98272214731349e-06,
+      "loss": 0.7319,
+      "step": 412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982539902771601e-06,
+      "loss": 1.0331,
+      "step": 413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982356703800737e-06,
+      "loss": 0.6707,
+      "step": 414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98217255043599e-06,
+      "loss": 0.5373,
+      "step": 415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981987442712634e-06,
+      "loss": 0.9105,
+      "step": 416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981801380666128e-06,
+      "loss": 1.0982,
+      "step": 417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981614364332113e-06,
+      "loss": 0.76,
+      "step": 418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981426393746412e-06,
+      "loss": 0.8458,
+      "step": 419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981237468945034e-06,
+      "loss": 1.0121,
+      "step": 420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981047589964169e-06,
+      "loss": 0.9516,
+      "step": 421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980856756840184e-06,
+      "loss": 0.9978,
+      "step": 422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980664969609639e-06,
+      "loss": 1.2066,
+      "step": 423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98047222830927e-06,
+      "loss": 0.9771,
+      "step": 424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980278532975998e-06,
+      "loss": 0.7784,
+      "step": 425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980083883646925e-06,
+      "loss": 0.9427,
+      "step": 426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979888280359338e-06,
+      "loss": 0.8165,
+      "step": 427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979691723150707e-06,
+      "loss": 1.0337,
+      "step": 428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97949421205868e-06,
+      "loss": 0.9154,
+      "step": 429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979295747121094e-06,
+      "loss": 1.2343,
+      "step": 430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979096328375966e-06,
+      "loss": 0.9874,
+      "step": 431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978895955861494e-06,
+      "loss": 1.0055,
+      "step": 432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97869462961606e-06,
+      "loss": 1.0063,
+      "step": 433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978492349678232e-06,
+      "loss": 0.9728,
+      "step": 434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978289116086755e-06,
+      "loss": 0.9621,
+      "step": 435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97808492888056e-06,
+      "loss": 0.5343,
+      "step": 436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977879788098759e-06,
+      "loss": 0.5128,
+      "step": 437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977673693780649e-06,
+      "loss": 1.0888,
+      "step": 438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97746664596571e-06,
+      "loss": 1.1135,
+      "step": 439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977258644693599e-06,
+      "loss": 0.6881,
+      "step": 440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977049690004162e-06,
+      "loss": 1.1507,
+      "step": 441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976839781937426e-06,
+      "loss": 0.9149,
+      "step": 442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976628920533598e-06,
+      "loss": 1.0412,
+      "step": 443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97641710583307e-06,
+      "loss": 0.7587,
+      "step": 444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976204337876418e-06,
+      "loss": 0.5421,
+      "step": 445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975990616704396e-06,
+      "loss": 0.3849,
+      "step": 446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975775942357944e-06,
+      "loss": 0.933,
+      "step": 447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975560314878185e-06,
+      "loss": 1.1257,
+      "step": 448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975343734306423e-06,
+      "loss": 1.2571,
+      "step": 449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975126200684145e-06,
+      "loss": 0.8951,
+      "step": 450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974907714053019e-06,
+      "loss": 0.8468,
+      "step": 451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9746882744549e-06,
+      "loss": 1.0278,
+      "step": 452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97446788193182e-06,
+      "loss": 0.6844,
+      "step": 453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974246536525998e-06,
+      "loss": 0.99,
+      "step": 454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974024238279833e-06,
+      "loss": 0.7992,
+      "step": 455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973800987235908e-06,
+      "loss": 1.0582,
+      "step": 456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973576783436987e-06,
+      "loss": 1.152,
+      "step": 457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973351626926016e-06,
+      "loss": 0.9606,
+      "step": 458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973125517746129e-06,
+      "loss": 0.6483,
+      "step": 459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972898455940634e-06,
+      "loss": 1.0067,
+      "step": 460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972670441553027e-06,
+      "loss": 0.7687,
+      "step": 461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972441474626987e-06,
+      "loss": 0.6311,
+      "step": 462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972211555206372e-06,
+      "loss": 0.7677,
+      "step": 463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971980683335223e-06,
+      "loss": 0.4959,
+      "step": 464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971748859057769e-06,
+      "loss": 1.0437,
+      "step": 465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971516082418411e-06,
+      "loss": 0.8767,
+      "step": 466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971282353461743e-06,
+      "loss": 1.0986,
+      "step": 467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971047672232535e-06,
+      "loss": 1.0199,
+      "step": 468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970812038775741e-06,
+      "loss": 0.9444,
+      "step": 469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9705754531365e-06,
+      "loss": 0.7275,
+      "step": 470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970337915360128e-06,
+      "loss": 0.5996,
+      "step": 471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97009942549213e-06,
+      "loss": 0.7855,
+      "step": 472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969859983578186e-06,
+      "loss": 0.6925,
+      "step": 473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969619589664165e-06,
+      "loss": 1.238,
+      "step": 474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969378243796114e-06,
+      "loss": 0.6377,
+      "step": 475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969135946020265e-06,
+      "loss": 0.8494,
+      "step": 476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96889269638303e-06,
+      "loss": 1.062,
+      "step": 477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968648494931007e-06,
+      "loss": 1.0424,
+      "step": 478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968403341710971e-06,
+      "loss": 0.9495,
+      "step": 479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968157236769884e-06,
+      "loss": 1.0932,
+      "step": 480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96791018015489e-06,
+      "loss": 1.1115,
+      "step": 481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967662171913311e-06,
+      "loss": 0.9254,
+      "step": 482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967413212092653e-06,
+      "loss": 0.6472,
+      "step": 483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96716330074061e-06,
+      "loss": 0.6548,
+      "step": 484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96691243790505e-06,
+      "loss": 1.0796,
+      "step": 485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96666062363403e-06,
+      "loss": 1.4496,
+      "step": 486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966407857975784e-06,
+      "loss": 0.7589,
+      "step": 487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96615414097873e-06,
+      "loss": 1.1153,
+      "step": 488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96589947269147e-06,
+      "loss": 0.8483,
+      "step": 489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965643853162787e-06,
+      "loss": 0.6724,
+      "step": 490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965387282441646e-06,
+      "loss": 1.046,
+      "step": 491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965129760577194e-06,
+      "loss": 0.7642,
+      "step": 492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96487128761876e-06,
+      "loss": 0.528,
+      "step": 493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964611863615856e-06,
+      "loss": 0.9637,
+      "step": 494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964351488618177e-06,
+      "loss": 1.0592,
+      "step": 495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964090162675597e-06,
+      "loss": 0.9474,
+      "step": 496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963827885838176e-06,
+      "loss": 0.7017,
+      "step": 497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963564658156156e-06,
+      "loss": 0.9424,
+      "step": 498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963300479679954e-06,
+      "loss": 0.9678,
+      "step": 499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96303535046018e-06,
+      "loss": 1.1911,
+      "step": 500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962769270547616e-06,
+      "loss": 1.1378,
+      "step": 501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962502239993235e-06,
+      "loss": 0.8081,
+      "step": 502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962234258848189e-06,
+      "loss": 0.9825,
+      "step": 503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961965327163806e-06,
+      "loss": 0.8279,
+      "step": 504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961695444991604e-06,
+      "loss": 0.8931,
+      "step": 505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961424612383279e-06,
+      "loss": 1.1198,
+      "step": 506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961152829390712e-06,
+      "loss": 1.093,
+      "step": 507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960880096065964e-06,
+      "loss": 0.7394,
+      "step": 508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960606412461276e-06,
+      "loss": 0.383,
+      "step": 509
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960331778629077e-06,
+      "loss": 1.1416,
+      "step": 510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960056194621972e-06,
+      "loss": 0.7321,
+      "step": 511
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959779660492752e-06,
+      "loss": 0.9663,
+      "step": 512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959502176294384e-06,
+      "loss": 0.9955,
+      "step": 513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959223742080028e-06,
+      "loss": 1.1555,
+      "step": 514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958944357903015e-06,
+      "loss": 0.7828,
+      "step": 515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958664023816864e-06,
+      "loss": 1.0883,
+      "step": 516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958382739875273e-06,
+      "loss": 0.9767,
+      "step": 517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958100506132127e-06,
+      "loss": 0.8694,
+      "step": 518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957817322641483e-06,
+      "loss": 0.8403,
+      "step": 519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957533189457592e-06,
+      "loss": 0.8948,
+      "step": 520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957248106634877e-06,
+      "loss": 0.7887,
+      "step": 521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956962074227948e-06,
+      "loss": 0.8266,
+      "step": 522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956675092291596e-06,
+      "loss": 0.7375,
+      "step": 523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956387160880796e-06,
+      "loss": 1.053,
+      "step": 524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9560982800507e-06,
+      "loss": 1.245,
+      "step": 525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955808449856646e-06,
+      "loss": 1.0802,
+      "step": 526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95551767035415e-06,
+      "loss": 0.8212,
+      "step": 527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955225941598913e-06,
+      "loss": 0.8458,
+      "step": 528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95493326364682e-06,
+      "loss": 0.716,
+      "step": 529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954639636553931e-06,
+      "loss": 0.7832,
+      "step": 530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954345060376494e-06,
+      "loss": 0.7786,
+      "step": 531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954049535170934e-06,
+      "loss": 1.0799,
+      "step": 532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953753060993863e-06,
+      "loss": 0.9293,
+      "step": 533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953455637902072e-06,
+      "loss": 0.812,
+      "step": 534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953157265952532e-06,
+      "loss": 0.8227,
+      "step": 535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952857945202397e-06,
+      "loss": 0.8841,
+      "step": 536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952557675709005e-06,
+      "loss": 0.6258,
+      "step": 537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952256457529874e-06,
+      "loss": 0.7807,
+      "step": 538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951954290722704e-06,
+      "loss": 1.2959,
+      "step": 539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951651175345376e-06,
+      "loss": 0.6945,
+      "step": 540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95134711145595e-06,
+      "loss": 0.5745,
+      "step": 541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951042099112677e-06,
+      "loss": 0.6464,
+      "step": 542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95073613837398e-06,
+      "loss": 1.0292,
+      "step": 543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950429229298468e-06,
+      "loss": 0.9573,
+      "step": 544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950121371944929e-06,
+      "loss": 0.759,
+      "step": 545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949812566372336e-06,
+      "loss": 0.6118,
+      "step": 546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949502812639844e-06,
+      "loss": 0.7274,
+      "step": 547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949192110806784e-06,
+      "loss": 1.0543,
+      "step": 548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948880460932674e-06,
+      "loss": 0.8597,
+      "step": 549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948567863077215e-06,
+      "loss": 0.864,
+      "step": 550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948254317300284e-06,
+      "loss": 0.9568,
+      "step": 551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947939823661941e-06,
+      "loss": 0.8584,
+      "step": 552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94762438222243e-06,
+      "loss": 0.6857,
+      "step": 553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947307993042174e-06,
+      "loss": 1.2844,
+      "step": 554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.946990656181782e-06,
+      "loss": 1.0336,
+      "step": 555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946672371702038e-06,
+      "loss": 0.6975,
+      "step": 556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946353139663912e-06,
+      "loss": 0.7109,
+      "step": 557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946032960128556e-06,
+      "loss": 0.889,
+      "step": 558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9457118331573e-06,
+      "loss": 0.7297,
+      "step": 559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945389758811657e-06,
+      "loss": 0.6754,
+      "step": 560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945066737153324e-06,
+      "loss": 1.1451,
+      "step": 561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944742768244175e-06,
+      "loss": 1.1348,
+      "step": 562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94441785214627e-06,
+      "loss": 0.9754,
+      "step": 563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944091988921848e-06,
+      "loss": 0.8784,
+      "step": 564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943765178633327e-06,
+      "loss": 1.1028,
+      "step": 565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943437421343312e-06,
+      "loss": 0.7337,
+      "step": 566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943108717114585e-06,
+      "loss": 0.9664,
+      "step": 567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94277906601011e-06,
+      "loss": 0.5131,
+      "step": 568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942448468093035e-06,
+      "loss": 1.0208,
+      "step": 569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942116923426687e-06,
+      "loss": 0.7394,
+      "step": 570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941784432074574e-06,
+      "loss": 0.9505,
+      "step": 571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941450994100389e-06,
+      "loss": 0.5422,
+      "step": 572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941116609568e-06,
+      "loss": 0.8676,
+      "step": 573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940781278541463e-06,
+      "loss": 0.7919,
+      "step": 574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94044500108501e-06,
+      "loss": 0.9621,
+      "step": 575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940107777263058e-06,
+      "loss": 0.5457,
+      "step": 576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939769607140202e-06,
+      "loss": 0.7754,
+      "step": 577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939430490781224e-06,
+      "loss": 0.9272,
+      "step": 578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93909042825108e-06,
+      "loss": 0.5137,
+      "step": 579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938749419614909e-06,
+      "loss": 0.8178,
+      "step": 580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938407464938035e-06,
+      "loss": 0.9325,
+      "step": 581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938064564285964e-06,
+      "loss": 1.0217,
+      "step": 582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937720717724375e-06,
+      "loss": 0.724,
+      "step": 583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937375925319136e-06,
+      "loss": 1.0682,
+      "step": 584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937030187136294e-06,
+      "loss": 0.6731,
+      "step": 585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936683503242075e-06,
+      "loss": 0.9841,
+      "step": 586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936335873702892e-06,
+      "loss": 0.8879,
+      "step": 587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93598729858533e-06,
+      "loss": 0.7579,
+      "step": 588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935637777956162e-06,
+      "loss": 0.4593,
+      "step": 589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935287311882341e-06,
+      "loss": 0.6574,
+      "step": 590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934935900431e-06,
+      "loss": 0.9981,
+      "step": 591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934583543669454e-06,
+      "loss": 1.0436,
+      "step": 592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934230241665197e-06,
+      "loss": 0.9223,
+      "step": 593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933875994485908e-06,
+      "loss": 0.5531,
+      "step": 594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933520802199446e-06,
+      "loss": 0.6977,
+      "step": 595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933164664873845e-06,
+      "loss": 0.8682,
+      "step": 596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932807582577326e-06,
+      "loss": 1.0364,
+      "step": 597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932449555378291e-06,
+      "loss": 0.985,
+      "step": 598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932090583345323e-06,
+      "loss": 0.8884,
+      "step": 599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931730666547184e-06,
+      "loss": 0.7476,
+      "step": 600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931369805052815e-06,
+      "loss": 0.5483,
+      "step": 601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931007998931346e-06,
+      "loss": 0.5233,
+      "step": 602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930645248252078e-06,
+      "loss": 0.8397,
+      "step": 603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930281553084498e-06,
+      "loss": 0.9363,
+      "step": 604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929916913498276e-06,
+      "loss": 1.0691,
+      "step": 605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929551329563259e-06,
+      "loss": 0.8623,
+      "step": 606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929184801349476e-06,
+      "loss": 0.7546,
+      "step": 607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928817328927138e-06,
+      "loss": 0.948,
+      "step": 608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928448912366637e-06,
+      "loss": 0.8551,
+      "step": 609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928079551738542e-06,
+      "loss": 0.5595,
+      "step": 610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92770924711361e-06,
+      "loss": 0.6546,
+      "step": 611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92733799856277e-06,
+      "loss": 1.1394,
+      "step": 612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926965806157139e-06,
+      "loss": 0.9028,
+      "step": 613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926592669968012e-06,
+      "loss": 0.9726,
+      "step": 614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926218590066862e-06,
+      "loss": 1.2096,
+      "step": 615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925843566525352e-06,
+      "loss": 0.5007,
+      "step": 616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925467599415315e-06,
+      "loss": 0.6363,
+      "step": 617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92509068880877e-06,
+      "loss": 0.6997,
+      "step": 618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924712834777917e-06,
+      "loss": 1.0362,
+      "step": 619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924334037395133e-06,
+      "loss": 0.8084,
+      "step": 620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923954296732983e-06,
+      "loss": 0.7591,
+      "step": 621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923573612864203e-06,
+      "loss": 0.9736,
+      "step": 622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923191985861717e-06,
+      "loss": 0.9902,
+      "step": 623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92280941579863e-06,
+      "loss": 0.214,
+      "step": 624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922425902748223e-06,
+      "loss": 1.3241,
+      "step": 625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922041446783958e-06,
+      "loss": 0.9198,
+      "step": 626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921656047979482e-06,
+      "loss": 0.8802,
+      "step": 627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921269706408617e-06,
+      "loss": 0.6909,
+      "step": 628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920882422145372e-06,
+      "loss": 0.8899,
+      "step": 629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920494195263932e-06,
+      "loss": 1.111,
+      "step": 630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920105025838663e-06,
+      "loss": 0.8218,
+      "step": 631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91971491394411e-06,
+      "loss": 1.1865,
+      "step": 632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919323859655005e-06,
+      "loss": 0.8212,
+      "step": 633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918931863046255e-06,
+      "loss": 1.1388,
+      "step": 634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918538924192948e-06,
+      "loss": 1.3067,
+      "step": 635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918145043170354e-06,
+      "loss": 1.0191,
+      "step": 636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917750220053922e-06,
+      "loss": 0.7972,
+      "step": 637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917354454919282e-06,
+      "loss": 0.8026,
+      "step": 638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916957747842246e-06,
+      "loss": 1.0769,
+      "step": 639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916560098898804e-06,
+      "loss": 0.7667,
+      "step": 640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916161508165128e-06,
+      "loss": 0.626,
+      "step": 641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91576197571757e-06,
+      "loss": 0.4273,
+      "step": 642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915361501632662e-06,
+      "loss": 0.8488,
+      "step": 643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914960085987117e-06,
+      "loss": 0.5344,
+      "step": 644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914557728857828e-06,
+      "loss": 0.9835,
+      "step": 645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914154430321868e-06,
+      "loss": 0.9352,
+      "step": 646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913750190456491e-06,
+      "loss": 0.7781,
+      "step": 647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913345009339133e-06,
+      "loss": 0.4665,
+      "step": 648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912938887047402e-06,
+      "loss": 0.5267,
+      "step": 649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912531823659102e-06,
+      "loss": 0.876,
+      "step": 650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9121238192522e-06,
+      "loss": 0.9045,
+      "step": 651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911714873904856e-06,
+      "loss": 0.6884,
+      "step": 652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911304987695403e-06,
+      "loss": 0.6458,
+      "step": 653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910894160702356e-06,
+      "loss": 1.0946,
+      "step": 654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910482393004412e-06,
+      "loss": 0.8536,
+      "step": 655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910069684680449e-06,
+      "loss": 0.8783,
+      "step": 656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90965603580952e-06,
+      "loss": 0.6397,
+      "step": 657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909241446470864e-06,
+      "loss": 0.532,
+      "step": 658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908825916743893e-06,
+      "loss": 0.854,
+      "step": 659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90840944670821e-06,
+      "loss": 0.856,
+      "step": 660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907992036443588e-06,
+      "loss": 1.2318,
+      "step": 661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907573686029983e-06,
+      "loss": 0.893,
+      "step": 662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907154395547535e-06,
+      "loss": 1.2827,
+      "step": 663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90673416507656e-06,
+      "loss": 0.8333,
+      "step": 664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906312994697554e-06,
+      "loss": 0.6706,
+      "step": 665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905890884491196e-06,
+      "loss": 0.6113,
+      "step": 666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905467834538342e-06,
+      "loss": 0.4621,
+      "step": 667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905043844920029e-06,
+      "loss": 0.8177,
+      "step": 668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904618915717474e-06,
+      "loss": 0.5307,
+      "step": 669
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904193047012075e-06,
+      "loss": 0.7455,
+      "step": 670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903766238885409e-06,
+      "loss": 1.1035,
+      "step": 671
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903338491419234e-06,
+      "loss": 0.7861,
+      "step": 672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902909804695484e-06,
+      "loss": 1.0159,
+      "step": 673
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902480178796279e-06,
+      "loss": 0.8174,
+      "step": 674
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902049613803916e-06,
+      "loss": 1.1595,
+      "step": 675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90161810980087e-06,
+      "loss": 1.1182,
+      "step": 676
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901185666869798e-06,
+      "loss": 1.0087,
+      "step": 677
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900752285093537e-06,
+      "loss": 0.7609,
+      "step": 678
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900317964555103e-06,
+      "loss": 0.9424,
+      "step": 679
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899882705337692e-06,
+      "loss": 0.7604,
+      "step": 680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899446507524681e-06,
+      "loss": 0.5887,
+      "step": 681
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899009371199623e-06,
+      "loss": 0.7853,
+      "step": 682
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898571296446257e-06,
+      "loss": 0.9126,
+      "step": 683
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898132283348496e-06,
+      "loss": 0.7996,
+      "step": 684
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897692331990437e-06,
+      "loss": 0.845,
+      "step": 685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89725144245635e-06,
+      "loss": 1.0441,
+      "step": 686
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896809614830697e-06,
+      "loss": 1.0814,
+      "step": 687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896366849198107e-06,
+      "loss": 1.0648,
+      "step": 688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895923145643394e-06,
+      "loss": 0.3799,
+      "step": 689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895478504251553e-06,
+      "loss": 1.1858,
+      "step": 690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895032925107757e-06,
+      "loss": 0.772,
+      "step": 691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894586408297357e-06,
+      "loss": 0.9046,
+      "step": 692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894138953905889e-06,
+      "loss": 0.8208,
+      "step": 693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89369056201906e-06,
+      "loss": 1.0719,
+      "step": 694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893241232722767e-06,
+      "loss": 1.0333,
+      "step": 695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892790966103078e-06,
+      "loss": 0.8998,
+      "step": 696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892339762246244e-06,
+      "loss": 0.9987,
+      "step": 697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891887621238694e-06,
+      "loss": 0.727,
+      "step": 698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89143454316704e-06,
+      "loss": 0.7553,
+      "step": 699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89098052811807e-06,
+      "loss": 1.0307,
+      "step": 700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890525576178756e-06,
+      "loss": 1.337,
+      "step": 701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89006968743624e-06,
+      "loss": 0.5028,
+      "step": 702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889612861977855e-06,
+      "loss": 1.1676,
+      "step": 703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889155099891104e-06,
+      "loss": 1.0415,
+      "step": 704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888696401263676e-06,
+      "loss": 0.5409,
+      "step": 705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888236766183437e-06,
+      "loss": 0.7758,
+      "step": 706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887776194738433e-06,
+      "loss": 0.8334,
+      "step": 707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887314687016885e-06,
+      "loss": 0.5538,
+      "step": 708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8868522431072e-06,
+      "loss": 0.7818,
+      "step": 709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88638886309796e-06,
+      "loss": 0.6484,
+      "step": 710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885924547077929e-06,
+      "loss": 1.0821,
+      "step": 711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885459295136049e-06,
+      "loss": 1.0019,
+      "step": 712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884993107361438e-06,
+      "loss": 1.0072,
+      "step": 713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8845259838434e-06,
+      "loss": 0.7903,
+      "step": 714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884057924671413e-06,
+      "loss": 0.8519,
+      "step": 715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883588929935137e-06,
+      "loss": 0.7915,
+      "step": 716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883118999724408e-06,
+      "loss": 0.9514,
+      "step": 717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882648134129247e-06,
+      "loss": 0.7518,
+      "step": 718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882176333239847e-06,
+      "loss": 1.0917,
+      "step": 719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881703597146585e-06,
+      "loss": 0.604,
+      "step": 720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881229925940018e-06,
+      "loss": 1.2455,
+      "step": 721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880755319710876e-06,
+      "loss": 0.7099,
+      "step": 722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880279778550074e-06,
+      "loss": 1.2002,
+      "step": 723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879803302548705e-06,
+      "loss": 0.6645,
+      "step": 724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87932589179804e-06,
+      "loss": 0.7224,
+      "step": 725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878847546389526e-06,
+      "loss": 0.87,
+      "step": 726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878368266414795e-06,
+      "loss": 1.1468,
+      "step": 727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877888051965659e-06,
+      "loss": 0.9486,
+      "step": 728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877406903134099e-06,
+      "loss": 0.7768,
+      "step": 729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876924820012284e-06,
+      "loss": 0.8107,
+      "step": 730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876441802692558e-06,
+      "loss": 0.7528,
+      "step": 731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875957851267449e-06,
+      "loss": 1.1243,
+      "step": 732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875472965829654e-06,
+      "loss": 1.3151,
+      "step": 733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874987146472062e-06,
+      "loss": 1.2582,
+      "step": 734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87450039328773e-06,
+      "loss": 0.9472,
+      "step": 735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874012706369896e-06,
+      "loss": 0.596,
+      "step": 736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873524085811985e-06,
+      "loss": 0.8041,
+      "step": 737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873034531707589e-06,
+      "loss": 0.6678,
+      "step": 738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872544044150486e-06,
+      "loss": 1.4206,
+      "step": 739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872052623234632e-06,
+      "loss": 0.7388,
+      "step": 740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871560269054159e-06,
+      "loss": 0.8548,
+      "step": 741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871066981703383e-06,
+      "loss": 0.7746,
+      "step": 742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870572761276794e-06,
+      "loss": 0.8671,
+      "step": 743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87007760786906e-06,
+      "loss": 0.9876,
+      "step": 744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869581521575033e-06,
+      "loss": 0.7214,
+      "step": 745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86908450248974e-06,
+      "loss": 0.9608,
+      "step": 746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868586550708385e-06,
+      "loss": 1.0384,
+      "step": 747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868087666326356e-06,
+      "loss": 1.1826,
+      "step": 748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867587849439215e-06,
+      "loss": 0.862,
+      "step": 749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867087100142707e-06,
+      "loss": 0.9746,
+      "step": 750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866585418532749e-06,
+      "loss": 0.8815,
+      "step": 751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866082804705443e-06,
+      "loss": 0.7723,
+      "step": 752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865579258757066e-06,
+      "loss": 1.0226,
+      "step": 753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865074780784075e-06,
+      "loss": 1.2054,
+      "step": 754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864569370883104e-06,
+      "loss": 0.9596,
+      "step": 755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864063029150967e-06,
+      "loss": 1.1194,
+      "step": 756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863555755684659e-06,
+      "loss": 0.5769,
+      "step": 757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863047550581349e-06,
+      "loss": 0.7887,
+      "step": 758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862538413938382e-06,
+      "loss": 0.7382,
+      "step": 759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862028345853291e-06,
+      "loss": 0.5134,
+      "step": 760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861517346423782e-06,
+      "loss": 0.8533,
+      "step": 761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861005415747733e-06,
+      "loss": 0.8663,
+      "step": 762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860492553923213e-06,
+      "loss": 0.5481,
+      "step": 763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859978761048461e-06,
+      "loss": 0.8242,
+      "step": 764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859464037221898e-06,
+      "loss": 1.1168,
+      "step": 765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85894838254212e-06,
+      "loss": 1.1676,
+      "step": 766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8584317971079e-06,
+      "loss": 0.9099,
+      "step": 767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8579142810182e-06,
+      "loss": 0.9513,
+      "step": 768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857395834372147e-06,
+      "loss": 0.9945,
+      "step": 769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856876457269054e-06,
+      "loss": 0.8949,
+      "step": 770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856356149808408e-06,
+      "loss": 0.9941,
+      "step": 771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85583491208988e-06,
+      "loss": 0.6979,
+      "step": 772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85531274421331e-06,
+      "loss": 1.0583,
+      "step": 773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854789646278726e-06,
+      "loss": 0.9125,
+      "step": 774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85426561838633e-06,
+      "loss": 0.5385,
+      "step": 775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853740660636501e-06,
+      "loss": 1.015,
+      "step": 776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853214773129796e-06,
+      "loss": 0.8212,
+      "step": 777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852687955966953e-06,
+      "loss": 0.3944,
+      "step": 778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852160209248887e-06,
+      "loss": 0.5963,
+      "step": 779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851631533076686e-06,
+      "loss": 0.686,
+      "step": 780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851101927551625e-06,
+      "loss": 1.0175,
+      "step": 781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850571392775152e-06,
+      "loss": 0.7722,
+      "step": 782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85003992884889e-06,
+      "loss": 0.9795,
+      "step": 783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849507535874647e-06,
+      "loss": 0.9995,
+      "step": 784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848974213954402e-06,
+      "loss": 0.7029,
+      "step": 785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84843996319032e-06,
+      "loss": 0.9328,
+      "step": 786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847904783684736e-06,
+      "loss": 0.907,
+      "step": 787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847368675540167e-06,
+      "loss": 0.9678,
+      "step": 788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846831638859305e-06,
+      "loss": 0.7602,
+      "step": 789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846293673745026e-06,
+      "loss": 0.589,
+      "step": 790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845754780300375e-06,
+      "loss": 0.9694,
+      "step": 791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845214958628584e-06,
+      "loss": 1.0075,
+      "step": 792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844674208833056e-06,
+      "loss": 1.1848,
+      "step": 793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844132531017374e-06,
+      "loss": 0.6933,
+      "step": 794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8435899252853e-06,
+      "loss": 1.1212,
+      "step": 795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84304639174077e-06,
+      "loss": 1.0715,
+      "step": 796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842501930487907e-06,
+      "loss": 0.9248,
+      "step": 797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841956541631e-06,
+      "loss": 1.3539,
+      "step": 798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841410225274519e-06,
+      "loss": 0.9047,
+      "step": 799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840862981523116e-06,
+      "loss": 0.8181,
+      "step": 800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840314810481618e-06,
+      "loss": 0.4776,
+      "step": 801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839765712255033e-06,
+      "loss": 0.4892,
+      "step": 802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839215686948538e-06,
+      "loss": 0.5184,
+      "step": 803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838664734667496e-06,
+      "loss": 0.8588,
+      "step": 804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838112855517442e-06,
+      "loss": 0.7959,
+      "step": 805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837560049604094e-06,
+      "loss": 0.8675,
+      "step": 806
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837006317033343e-06,
+      "loss": 0.883,
+      "step": 807
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836451657911258e-06,
+      "loss": 1.1488,
+      "step": 808
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83589607234409e-06,
+      "loss": 0.9192,
+      "step": 809
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835339560438262e-06,
+      "loss": 0.8721,
+      "step": 810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834782122300375e-06,
+      "loss": 0.8588,
+      "step": 811
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83422375803721e-06,
+      "loss": 0.6745,
+      "step": 812
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833664467755727e-06,
+      "loss": 0.9076,
+      "step": 813
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833104251563058e-06,
+      "loss": 0.7565,
+      "step": 814
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832543109566512e-06,
+      "loss": 0.5676,
+      "step": 815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831981041873585e-06,
+      "loss": 1.0683,
+      "step": 816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83141804859194e-06,
+      "loss": 1.1415,
+      "step": 817
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83085412982942e-06,
+      "loss": 0.9838,
+      "step": 818
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83028928569405e-06,
+      "loss": 1.0977,
+      "step": 819
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829723516294025e-06,
+      "loss": 1.0048,
+      "step": 820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829156821737721e-06,
+      "loss": 0.7266,
+      "step": 821
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828589202133695e-06,
+      "loss": 0.8012,
+      "step": 822
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828020657590674e-06,
+      "loss": 1.0607,
+      "step": 823
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827451188217563e-06,
+      "loss": 1.1139,
+      "step": 824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826880794123452e-06,
+      "loss": 0.7748,
+      "step": 825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826309475417598e-06,
+      "loss": 0.9398,
+      "step": 826
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825737232209442e-06,
+      "loss": 1.1726,
+      "step": 827
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8251640646086e-06,
+      "loss": 0.9058,
+      "step": 828
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824589972724865e-06,
+      "loss": 0.5538,
+      "step": 829
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824014956668206e-06,
+      "loss": 1.1838,
+      "step": 830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823439016548772e-06,
+      "loss": 0.6924,
+      "step": 831
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822862152476884e-06,
+      "loss": 1.1165,
+      "step": 832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822284364563045e-06,
+      "loss": 0.78,
+      "step": 833
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821705652917932e-06,
+      "loss": 0.7107,
+      "step": 834
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821126017652403e-06,
+      "loss": 0.5942,
+      "step": 835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820545458877485e-06,
+      "loss": 0.6833,
+      "step": 836
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819963976704389e-06,
+      "loss": 0.7568,
+      "step": 837
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819381571244504e-06,
+      "loss": 0.7148,
+      "step": 838
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818798242609387e-06,
+      "loss": 1.0044,
+      "step": 839
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81821399091078e-06,
+      "loss": 1.0086,
+      "step": 840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8176288162606e-06,
+      "loss": 0.6862,
+      "step": 841
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817042718770937e-06,
+      "loss": 0.8914,
+      "step": 842
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816455698554064e-06,
+      "loss": 1.1491,
+      "step": 843
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815867755722426e-06,
+      "loss": 1.0445,
+      "step": 844
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815278890388646e-06,
+      "loss": 0.8669,
+      "step": 845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814689102665524e-06,
+      "loss": 0.9428,
+      "step": 846
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814098392666035e-06,
+      "loss": 0.8762,
+      "step": 847
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813506760503338e-06,
+      "loss": 0.9525,
+      "step": 848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812914206290757e-06,
+      "loss": 1.2006,
+      "step": 849
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8123207301418e-06,
+      "loss": 0.925,
+      "step": 850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811726332170153e-06,
+      "loss": 0.8126,
+      "step": 851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.81113101248967e-06,
+      "loss": 0.6725,
+      "step": 852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810534771214395e-06,
+      "loss": 1.1322,
+      "step": 853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809937608458534e-06,
+      "loss": 0.7004,
+      "step": 854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809339524336479e-06,
+      "loss": 0.8507,
+      "step": 855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808740518962797e-06,
+      "loss": 1.0524,
+      "step": 856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808140592452226e-06,
+      "loss": 0.434,
+      "step": 857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80753974491969e-06,
+      "loss": 0.6685,
+      "step": 858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806937976480282e-06,
+      "loss": 1.1402,
+      "step": 859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806335287249273e-06,
+      "loss": 0.8625,
+      "step": 860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805731677342114e-06,
+      "loss": 0.6995,
+      "step": 861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805127146874423e-06,
+      "loss": 1.3234,
+      "step": 862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804521695962007e-06,
+      "loss": 0.7609,
+      "step": 863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80391532472084e-06,
+      "loss": 0.9118,
+      "step": 864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803308033267077e-06,
+      "loss": 0.4854,
+      "step": 865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802699821717042e-06,
+      "loss": 1.1936,
+      "step": 866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80209069018725e-06,
+      "loss": 0.9729,
+      "step": 867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801480638794376e-06,
+      "loss": 1.0319,
+      "step": 868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800869667655279e-06,
+      "loss": 0.6202,
+      "step": 869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800257776886997e-06,
+      "loss": 1.2058,
+      "step": 870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799644966606736e-06,
+      "loss": 0.7687,
+      "step": 871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799031236931887e-06,
+      "loss": 0.8308,
+      "step": 872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798416587980009e-06,
+      "loss": 0.9572,
+      "step": 873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797801019868844e-06,
+      "loss": 0.3484,
+      "step": 874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797184532716305e-06,
+      "loss": 1.3287,
+      "step": 875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796567126640482e-06,
+      "loss": 0.9088,
+      "step": 876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795948801759646e-06,
+      "loss": 0.9951,
+      "step": 877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795329558192237e-06,
+      "loss": 0.8032,
+      "step": 878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794709396056876e-06,
+      "loss": 0.9703,
+      "step": 879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794088315472355e-06,
+      "loss": 0.8195,
+      "step": 880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793466316557646e-06,
+      "loss": 0.9233,
+      "step": 881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792843399431898e-06,
+      "loss": 0.7159,
+      "step": 882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792219564214431e-06,
+      "loss": 0.8759,
+      "step": 883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791594811024746e-06,
+      "loss": 1.0812,
+      "step": 884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790969139982515e-06,
+      "loss": 0.705,
+      "step": 885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79034255120759e-06,
+      "loss": 1.093,
+      "step": 886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789715044819996e-06,
+      "loss": 0.7424,
+      "step": 887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789086620939936e-06,
+      "loss": 1.053,
+      "step": 888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788457279687786e-06,
+      "loss": 0.6572,
+      "step": 889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787827021184101e-06,
+      "loss": 0.5893,
+      "step": 890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78719584554961e-06,
+      "loss": 0.6839,
+      "step": 891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786563752905214e-06,
+      "loss": 0.7147,
+      "step": 892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785930743371996e-06,
+      "loss": 0.9055,
+      "step": 893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785296817071213e-06,
+      "loss": 0.6515,
+      "step": 894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784661974124296e-06,
+      "loss": 0.9122,
+      "step": 895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784026214652851e-06,
+      "loss": 0.8724,
+      "step": 896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783389538778662e-06,
+      "loss": 0.7265,
+      "step": 897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782751946623686e-06,
+      "loss": 0.8057,
+      "step": 898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782113438310056e-06,
+      "loss": 0.6942,
+      "step": 899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781474013960086e-06,
+      "loss": 1.0472,
+      "step": 900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780833673696255e-06,
+      "loss": 0.9588,
+      "step": 901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780192417641228e-06,
+      "loss": 0.9667,
+      "step": 902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779550245917836e-06,
+      "loss": 0.7577,
+      "step": 903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778907158649093e-06,
+      "loss": 0.6149,
+      "step": 904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778263155958186e-06,
+      "loss": 0.943,
+      "step": 905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777618237968472e-06,
+      "loss": 0.9835,
+      "step": 906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776972404803495e-06,
+      "loss": 0.6627,
+      "step": 907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776325656586962e-06,
+      "loss": 0.8504,
+      "step": 908
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775677993442762e-06,
+      "loss": 0.6592,
+      "step": 909
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775029415494961e-06,
+      "loss": 1.1381,
+      "step": 910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774379922867793e-06,
+      "loss": 0.413,
+      "step": 911
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773729515685673e-06,
+      "loss": 1.0431,
+      "step": 912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77307819407319e-06,
+      "loss": 0.9672,
+      "step": 913
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772425958155108e-06,
+      "loss": 0.7452,
+      "step": 914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771772808056364e-06,
+      "loss": 0.8912,
+      "step": 915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771118743902073e-06,
+      "loss": 0.8834,
+      "step": 916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770463765817524e-06,
+      "loss": 0.5571,
+      "step": 917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769807873928181e-06,
+      "loss": 0.7153,
+      "step": 918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769151068359686e-06,
+      "loss": 1.0675,
+      "step": 919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76849334923785e-06,
+      "loss": 0.8759,
+      "step": 920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76783471668866e-06,
+      "loss": 1.0361,
+      "step": 921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767175170838285e-06,
+      "loss": 0.6302,
+      "step": 922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766514711813064e-06,
+      "loss": 0.8162,
+      "step": 923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765853339739507e-06,
+      "loss": 1.3011,
+      "step": 924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765191054744305e-06,
+      "loss": 1.2434,
+      "step": 925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764527856954322e-06,
+      "loss": 0.7647,
+      "step": 926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763863746496598e-06,
+      "loss": 0.761,
+      "step": 927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763198723498343e-06,
+      "loss": 0.7407,
+      "step": 928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76253278808695e-06,
+      "loss": 0.5768,
+      "step": 929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76186594038998e-06,
+      "loss": 0.8951,
+      "step": 930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761198180535168e-06,
+      "loss": 0.9081,
+      "step": 931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760529508650429e-06,
+      "loss": 0.7699,
+      "step": 932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75985992486385e-06,
+      "loss": 0.7777,
+      "step": 933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759189429303692e-06,
+      "loss": 0.5933,
+      "step": 934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758518022098394e-06,
+      "loss": 1.1449,
+      "step": 935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757845703376566e-06,
+      "loss": 0.557,
+      "step": 936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757172473266993e-06,
+      "loss": 0.8711,
+      "step": 937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756498331898636e-06,
+      "loss": 0.9407,
+      "step": 938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755823279400631e-06,
+      "loss": 0.8398,
+      "step": 939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755147315902286e-06,
+      "loss": 1.0807,
+      "step": 940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754470441533084e-06,
+      "loss": 0.8777,
+      "step": 941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753792656422687e-06,
+      "loss": 0.5376,
+      "step": 942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753113960700924e-06,
+      "loss": 0.9402,
+      "step": 943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752434354497805e-06,
+      "loss": 0.8072,
+      "step": 944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751753837943511e-06,
+      "loss": 0.9349,
+      "step": 945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751072411168397e-06,
+      "loss": 0.8984,
+      "step": 946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750390074302996e-06,
+      "loss": 0.6087,
+      "step": 947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749706827478012e-06,
+      "loss": 0.715,
+      "step": 948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749022670824322e-06,
+      "loss": 1.0578,
+      "step": 949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748337604472981e-06,
+      "loss": 0.8493,
+      "step": 950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747651628555218e-06,
+      "loss": 0.6873,
+      "step": 951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746964743202432e-06,
+      "loss": 0.6648,
+      "step": 952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746276948546203e-06,
+      "loss": 1.1307,
+      "step": 953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745588244718277e-06,
+      "loss": 0.5654,
+      "step": 954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744898631850581e-06,
+      "loss": 0.9557,
+      "step": 955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744208110075212e-06,
+      "loss": 0.5153,
+      "step": 956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743516679524448e-06,
+      "loss": 0.6488,
+      "step": 957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742824340330728e-06,
+      "loss": 0.544,
+      "step": 958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742131092626677e-06,
+      "loss": 1.1169,
+      "step": 959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741436936545088e-06,
+      "loss": 0.6182,
+      "step": 960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740741872218933e-06,
+      "loss": 0.472,
+      "step": 961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740045899781353e-06,
+      "loss": 0.4112,
+      "step": 962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739349019365664e-06,
+      "loss": 0.8428,
+      "step": 963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738651231105358e-06,
+      "loss": 0.7863,
+      "step": 964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737952535134099e-06,
+      "loss": 1.0156,
+      "step": 965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737252931585725e-06,
+      "loss": 1.1325,
+      "step": 966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73655242059425e-06,
+      "loss": 1.1295,
+      "step": 967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735851002293859e-06,
+      "loss": 1.1062,
+      "step": 968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735148676818912e-06,
+      "loss": 1.0078,
+      "step": 969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734445444303942e-06,
+      "loss": 1.2328,
+      "step": 970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73374130488366e-06,
+      "loss": 0.7339,
+      "step": 971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73303625869294e-06,
+      "loss": 1.0848,
+      "step": 972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732330305866847e-06,
+      "loss": 0.8908,
+      "step": 973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7316234465406e-06,
+      "loss": 0.7187,
+      "step": 974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730915680849606e-06,
+      "loss": 0.997,
+      "step": 975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730207008929441e-06,
+      "loss": 0.4803,
+      "step": 976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729497430915854e-06,
+      "loss": 0.4967,
+      "step": 977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728786946944767e-06,
+      "loss": 0.7069,
+      "step": 978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728075557152278e-06,
+      "loss": 0.7409,
+      "step": 979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727363261674656e-06,
+      "loss": 0.9827,
+      "step": 980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726650060648345e-06,
+      "loss": 0.8174,
+      "step": 981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72593595420996e-06,
+      "loss": 0.8544,
+      "step": 982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725220942496294e-06,
+      "loss": 0.497,
+      "step": 983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72450502564431e-06,
+      "loss": 1.2493,
+      "step": 984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723788203791147e-06,
+      "loss": 0.5994,
+      "step": 985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723070477074114e-06,
+      "loss": 1.1727,
+      "step": 986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722351845630694e-06,
+      "loss": 0.7106,
+      "step": 987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721632309598546e-06,
+      "loss": 0.7104,
+      "step": 988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720911869115496e-06,
+      "loss": 0.7522,
+      "step": 989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720190524319556e-06,
+      "loss": 0.7363,
+      "step": 990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719468275348896e-06,
+      "loss": 0.9552,
+      "step": 991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71874512234187e-06,
+      "loss": 0.9395,
+      "step": 992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718021065437e-06,
+      "loss": 0.6577,
+      "step": 993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717296104772981e-06,
+      "loss": 1.2169,
+      "step": 994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716570240488685e-06,
+      "loss": 1.1289,
+      "step": 995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715843472723154e-06,
+      "loss": 1.0722,
+      "step": 996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715115801615602e-06,
+      "loss": 1.0545,
+      "step": 997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714387227305422e-06,
+      "loss": 1.1627,
+      "step": 998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713657749932172e-06,
+      "loss": 0.4835,
+      "step": 999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712927369635587e-06,
+      "loss": 0.89,
+      "step": 1000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712196086555579e-06,
+      "loss": 0.8509,
+      "step": 1001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711463900832223e-06,
+      "loss": 0.8813,
+      "step": 1002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710730812605775e-06,
+      "loss": 1.0144,
+      "step": 1003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709996822016664e-06,
+      "loss": 1.0149,
+      "step": 1004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709261929205484e-06,
+      "loss": 0.7972,
+      "step": 1005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708526134313013e-06,
+      "loss": 1.0874,
+      "step": 1006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707789437480193e-06,
+      "loss": 0.8926,
+      "step": 1007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70705183884814e-06,
+      "loss": 0.814,
+      "step": 1008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706313338558149e-06,
+      "loss": 0.7113,
+      "step": 1009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705573936751678e-06,
+      "loss": 0.7158,
+      "step": 1010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704833633570368e-06,
+      "loss": 0.5926,
+      "step": 1011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704092429156025e-06,
+      "loss": 0.4782,
+      "step": 1012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70335032365063e-06,
+      "loss": 1.0283,
+      "step": 1013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702607317196336e-06,
+      "loss": 1.2056,
+      "step": 1014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70186340993547e-06,
+      "loss": 1.2174,
+      "step": 1015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701118602010533e-06,
+      "loss": 0.5362,
+      "step": 1016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700372893564195e-06,
+      "loss": 0.508,
+      "step": 1017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699626284739299e-06,
+      "loss": 0.948,
+      "step": 1018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698878775678864e-06,
+      "loss": 0.8598,
+      "step": 1019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698130366526076e-06,
+      "loss": 0.627,
+      "step": 1020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697381057424298e-06,
+      "loss": 0.4071,
+      "step": 1021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69663084851706e-06,
+      "loss": 0.5336,
+      "step": 1022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695879739948075e-06,
+      "loss": 1.1489,
+      "step": 1023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695127731861219e-06,
+      "loss": 0.6623,
+      "step": 1024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694374824400537e-06,
+      "loss": 0.9642,
+      "step": 1025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693621017710258e-06,
+      "loss": 0.9696,
+      "step": 1026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692866311934777e-06,
+      "loss": 0.9578,
+      "step": 1027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692110707218659e-06,
+      "loss": 0.7435,
+      "step": 1028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691354203706644e-06,
+      "loss": 0.5427,
+      "step": 1029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690596801543643e-06,
+      "loss": 0.2173,
+      "step": 1030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689838500874745e-06,
+      "loss": 1.0293,
+      "step": 1031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6890793018452e-06,
+      "loss": 1.1193,
+      "step": 1032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688319204600438e-06,
+      "loss": 0.907,
+      "step": 1033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687558209286061e-06,
+      "loss": 0.9386,
+      "step": 1034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68679631604784e-06,
+      "loss": 0.8087,
+      "step": 1035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68603352503172e-06,
+      "loss": 0.7875,
+      "step": 1036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685269836383816e-06,
+      "loss": 0.9136,
+      "step": 1037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684505250250415e-06,
+      "loss": 1.0891,
+      "step": 1038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68373976677798e-06,
+      "loss": 0.7684,
+      "step": 1039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68297338611314e-06,
+      "loss": 1.2054,
+      "step": 1040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682206108402702e-06,
+      "loss": 0.7007,
+      "step": 1041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68143793379364e-06,
+      "loss": 0.7333,
+      "step": 1042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680668862433102e-06,
+      "loss": 1.2872,
+      "step": 1043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679898894468407e-06,
+      "loss": 0.4657,
+      "step": 1044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679128030047046e-06,
+      "loss": 0.862,
+      "step": 1045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67835626931668e-06,
+      "loss": 0.9239,
+      "step": 1046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677583612425145e-06,
+      "loss": 0.4619,
+      "step": 1047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676810059520447e-06,
+      "loss": 0.4342,
+      "step": 1048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676035610750765e-06,
+      "loss": 0.6208,
+      "step": 1049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675260266264445e-06,
+      "loss": 1.2591,
+      "step": 1050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674484026210013e-06,
+      "loss": 0.7816,
+      "step": 1051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673706890736156e-06,
+      "loss": 1.0562,
+      "step": 1052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672928859991742e-06,
+      "loss": 1.2345,
+      "step": 1053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672149934125802e-06,
+      "loss": 0.8196,
+      "step": 1054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671370113287548e-06,
+      "loss": 0.9023,
+      "step": 1055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670589397626356e-06,
+      "loss": 1.0233,
+      "step": 1056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669807787291774e-06,
+      "loss": 0.6869,
+      "step": 1057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669025282433526e-06,
+      "loss": 1.1051,
+      "step": 1058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668241883201504e-06,
+      "loss": 0.5653,
+      "step": 1059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66745758974577e-06,
+      "loss": 0.7677,
+      "step": 1060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66667240221656e-06,
+      "loss": 0.6749,
+      "step": 1061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665886320764282e-06,
+      "loss": 0.8854,
+      "step": 1062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665099345539512e-06,
+      "loss": 0.4973,
+      "step": 1063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664311476692997e-06,
+      "loss": 1.0088,
+      "step": 1064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663522714375662e-06,
+      "loss": 0.9126,
+      "step": 1065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662733058738594e-06,
+      "loss": 1.2037,
+      "step": 1066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661942509933058e-06,
+      "loss": 0.8689,
+      "step": 1067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661151068110484e-06,
+      "loss": 0.547,
+      "step": 1068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660358733422478e-06,
+      "loss": 0.6606,
+      "step": 1069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659565506020816e-06,
+      "loss": 0.9528,
+      "step": 1070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658771386057443e-06,
+      "loss": 1.1948,
+      "step": 1071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65797637368448e-06,
+      "loss": 0.9141,
+      "step": 1072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657180469054213e-06,
+      "loss": 0.474,
+      "step": 1073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656383672319098e-06,
+      "loss": 0.8879,
+      "step": 1074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65558598363177e-06,
+      "loss": 0.3788,
+      "step": 1075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65478740314503e-06,
+      "loss": 1.0446,
+      "step": 1076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653987931011846e-06,
+      "loss": 1.0825,
+      "step": 1077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653187567385364e-06,
+      "loss": 0.8898,
+      "step": 1078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652386312418896e-06,
+      "loss": 0.925,
+      "step": 1079
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651584166265927e-06,
+      "loss": 0.9722,
+      "step": 1080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650781129080112e-06,
+      "loss": 0.8094,
+      "step": 1081
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649977201015273e-06,
+      "loss": 0.8727,
+      "step": 1082
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64917238222541e-06,
+      "loss": 0.9172,
+      "step": 1083
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64836667286469e-06,
+      "loss": 1.0877,
+      "step": 1084
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64756007308745e-06,
+      "loss": 0.6778,
+      "step": 1085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646752583048196e-06,
+      "loss": 0.5955,
+      "step": 1086
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645944202901607e-06,
+      "loss": 1.0142,
+      "step": 1087
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645134932802533e-06,
+      "loss": 1.0215,
+      "step": 1088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644324772905995e-06,
+      "loss": 0.8969,
+      "step": 1089
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643513723367181e-06,
+      "loss": 0.9485,
+      "step": 1090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64270178434145e-06,
+      "loss": 0.7435,
+      "step": 1091
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641888955984336e-06,
+      "loss": 0.5949,
+      "step": 1092
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641075238451538e-06,
+      "loss": 0.9986,
+      "step": 1093
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640260631898927e-06,
+      "loss": 0.9842,
+      "step": 1094
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639445136482549e-06,
+      "loss": 1.1986,
+      "step": 1095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638628752358609e-06,
+      "loss": 0.9297,
+      "step": 1096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637811479683494e-06,
+      "loss": 1.0415,
+      "step": 1097
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636993318613754e-06,
+      "loss": 0.8798,
+      "step": 1098
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636174269306114e-06,
+      "loss": 0.4874,
+      "step": 1099
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635354331917465e-06,
+      "loss": 1.2775,
+      "step": 1100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634533506604872e-06,
+      "loss": 0.4841,
+      "step": 1101
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633711793525567e-06,
+      "loss": 1.1232,
+      "step": 1102
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63288919283695e-06,
+      "loss": 1.1598,
+      "step": 1103
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632065704696599e-06,
+      "loss": 0.3766,
+      "step": 1104
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631241329262251e-06,
+      "loss": 0.8519,
+      "step": 1105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630416066691826e-06,
+      "loss": 0.9287,
+      "step": 1106
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6295899171434e-06,
+      "loss": 1.0379,
+      "step": 1107
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628762880775233e-06,
+      "loss": 0.856,
+      "step": 1108
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627934957745742e-06,
+      "loss": 1.0414,
+      "step": 1109
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627106148213521e-06,
+      "loss": 0.8934,
+      "step": 1110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626276452337334e-06,
+      "loss": 0.6241,
+      "step": 1111
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625445870276112e-06,
+      "loss": 0.905,
+      "step": 1112
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624614402188956e-06,
+      "loss": 0.6624,
+      "step": 1113
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62378204823514e-06,
+      "loss": 0.6978,
+      "step": 1114
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622948808574104e-06,
+      "loss": 0.7638,
+      "step": 1115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622114683365457e-06,
+      "loss": 1.2867,
+      "step": 1116
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621279672768984e-06,
+      "loss": 1.2041,
+      "step": 1117
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62044377694463e-06,
+      "loss": 0.7469,
+      "step": 1118
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61960699605252e-06,
+      "loss": 1.006,
+      "step": 1119
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618769330252939e-06,
+      "loss": 0.7097,
+      "step": 1120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617930779706347e-06,
+      "loss": 0.7929,
+      "step": 1121
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617091344573374e-06,
+      "loss": 0.9529,
+      "step": 1122
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616251025014818e-06,
+      "loss": 0.8732,
+      "step": 1123
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615409821191643e-06,
+      "loss": 0.8596,
+      "step": 1124
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61456773326499e-06,
+      "loss": 0.9014,
+      "step": 1125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61372476139616e-06,
+      "loss": 0.797,
+      "step": 1126
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612880905746631e-06,
+      "loss": 0.7716,
+      "step": 1127
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612036166478046e-06,
+      "loss": 0.771,
+      "step": 1128
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611190543752223e-06,
+      "loss": 0.9656,
+      "step": 1129
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610344037731139e-06,
+      "loss": 0.6564,
+      "step": 1130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60949664857695e-06,
+      "loss": 0.9552,
+      "step": 1131
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608648376451977e-06,
+      "loss": 1.008,
+      "step": 1132
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60779922151871e-06,
+      "loss": 0.7395,
+      "step": 1133
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606949183939806e-06,
+      "loss": 1.0102,
+      "step": 1134
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606098263878099e-06,
+      "loss": 0.8999,
+      "step": 1135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605246461496584e-06,
+      "loss": 0.7882,
+      "step": 1136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604393776958426e-06,
+      "loss": 0.7428,
+      "step": 1137
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603540210426965e-06,
+      "loss": 0.7412,
+      "step": 1138
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602685762065703e-06,
+      "loss": 0.7417,
+      "step": 1139
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601830432038312e-06,
+      "loss": 0.7208,
+      "step": 1140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600974220508637e-06,
+      "loss": 1.0164,
+      "step": 1141
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600117127640689e-06,
+      "loss": 0.7093,
+      "step": 1142
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599259153598648e-06,
+      "loss": 0.8602,
+      "step": 1143
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598400298546862e-06,
+      "loss": 0.8267,
+      "step": 1144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.59754056264985e-06,
+      "loss": 0.7241,
+      "step": 1145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596679946072298e-06,
+      "loss": 1.1275,
+      "step": 1146
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.595818448979061e-06,
+      "loss": 0.79,
+      "step": 1147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594956071535165e-06,
+      "loss": 1.0798,
+      "step": 1148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594092813905797e-06,
+      "loss": 0.9072,
+      "step": 1149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593228676256324e-06,
+      "loss": 1.0141,
+      "step": 1150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592363658752274e-06,
+      "loss": 1.0687,
+      "step": 1151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59149776155934e-06,
+      "loss": 0.7827,
+      "step": 1152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590630984843398e-06,
+      "loss": 0.7581,
+      "step": 1153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589763328770476e-06,
+      "loss": 1.0607,
+      "step": 1154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58889479350678e-06,
+      "loss": 1.0357,
+      "step": 1155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58802537921868e-06,
+      "loss": 0.7411,
+      "step": 1156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587155086072722e-06,
+      "loss": 0.6334,
+      "step": 1157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586283914235608e-06,
+      "loss": 1.3024,
+      "step": 1158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585411863874219e-06,
+      "loss": 0.8831,
+      "step": 1159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584538935155599e-06,
+      "loss": 0.6707,
+      "step": 1160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583665128246962e-06,
+      "loss": 0.8695,
+      "step": 1161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58279044331569e-06,
+      "loss": 0.7451,
+      "step": 1162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581914880529333e-06,
+      "loss": 0.8976,
+      "step": 1163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581038440055608e-06,
+      "loss": 1.0083,
+      "step": 1164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580161122062403e-06,
+      "loss": 1.0858,
+      "step": 1165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579282926717772e-06,
+      "loss": 0.712,
+      "step": 1166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578403854189936e-06,
+      "loss": 0.6994,
+      "step": 1167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577523904647289e-06,
+      "loss": 0.7682,
+      "step": 1168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576643078258384e-06,
+      "loss": 0.7198,
+      "step": 1169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575761375191952e-06,
+      "loss": 0.8203,
+      "step": 1170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574878795616887e-06,
+      "loss": 0.6439,
+      "step": 1171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573995339702248e-06,
+      "loss": 1.1759,
+      "step": 1172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573111007617269e-06,
+      "loss": 0.3833,
+      "step": 1173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572225799531346e-06,
+      "loss": 0.9103,
+      "step": 1174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571339715614045e-06,
+      "loss": 1.0636,
+      "step": 1175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5704527560351e-06,
+      "loss": 1.2104,
+      "step": 1176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569564920964411e-06,
+      "loss": 0.7665,
+      "step": 1177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568676210572048e-06,
+      "loss": 0.8416,
+      "step": 1178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567786625028248e-06,
+      "loss": 0.2392,
+      "step": 1179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566896164503414e-06,
+      "loss": 0.6484,
+      "step": 1180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56600482916812e-06,
+      "loss": 0.8614,
+      "step": 1181
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565112619193104e-06,
+      "loss": 0.7991,
+      "step": 1182
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564219534749273e-06,
+      "loss": 0.9251,
+      "step": 1183
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563325576007702e-06,
+      "loss": 0.8826,
+      "step": 1184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562430743139632e-06,
+      "loss": 1.0053,
+      "step": 1185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561535036316474e-06,
+      "loss": 0.6821,
+      "step": 1186
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560638455709802e-06,
+      "loss": 0.835,
+      "step": 1187
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559741001491366e-06,
+      "loss": 0.4819,
+      "step": 1188
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558842673833072e-06,
+      "loss": 0.3448,
+      "step": 1189
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557943472907e-06,
+      "loss": 1.0135,
+      "step": 1190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557043398885399e-06,
+      "loss": 0.7778,
+      "step": 1191
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55614245194068e-06,
+      "loss": 1.0124,
+      "step": 1192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555240632245423e-06,
+      "loss": 0.5928,
+      "step": 1193
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554337939972378e-06,
+      "loss": 1.1324,
+      "step": 1194
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553434375294456e-06,
+      "loss": 0.6965,
+      "step": 1195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552529938384744e-06,
+      "loss": 1.0476,
+      "step": 1196
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551624629416489e-06,
+      "loss": 1.0813,
+      "step": 1197
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550718448563108e-06,
+      "loss": 0.6466,
+      "step": 1198
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549811395998182e-06,
+      "loss": 0.9897,
+      "step": 1199
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548903471895463e-06,
+      "loss": 0.8911,
+      "step": 1200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547994676428868e-06,
+      "loss": 0.9192,
+      "step": 1201
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547085009772481e-06,
+      "loss": 0.8122,
+      "step": 1202
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546174472100551e-06,
+      "loss": 0.686,
+      "step": 1203
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5452630635875e-06,
+      "loss": 0.6846,
+      "step": 1204
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544350784407908e-06,
+      "loss": 0.9753,
+      "step": 1205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54343763473653e-06,
+      "loss": 1.047,
+      "step": 1206
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542523614748282e-06,
+      "loss": 0.6565,
+      "step": 1207
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541608724618249e-06,
+      "loss": 0.6404,
+      "step": 1208
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540692964521684e-06,
+      "loss": 0.7611,
+      "step": 1209
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539776334634002e-06,
+      "loss": 0.8611,
+      "step": 1210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53885883513079e-06,
+      "loss": 0.9496,
+      "step": 1211
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537940466187803e-06,
+      "loss": 0.5993,
+      "step": 1212
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537021227980952e-06,
+      "loss": 1.0129,
+      "step": 1213
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536101120686324e-06,
+      "loss": 0.7609,
+      "step": 1214
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535180144480172e-06,
+      "loss": 0.525,
+      "step": 1215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53425829953891e-06,
+      "loss": 0.6241,
+      "step": 1216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533335586039124e-06,
+      "loss": 0.7318,
+      "step": 1217
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532412004157565e-06,
+      "loss": 0.7625,
+      "step": 1218
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531487554071147e-06,
+      "loss": 1.0035,
+      "step": 1219
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530562235956954e-06,
+      "loss": 0.559,
+      "step": 1220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529636049992235e-06,
+      "loss": 0.9103,
+      "step": 1221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528708996354406e-06,
+      "loss": 0.655,
+      "step": 1222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527781075221046e-06,
+      "loss": 0.7555,
+      "step": 1223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526852286769903e-06,
+      "loss": 0.8209,
+      "step": 1224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525922631178894e-06,
+      "loss": 0.6834,
+      "step": 1225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524992108626096e-06,
+      "loss": 0.9091,
+      "step": 1226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524060719289758e-06,
+      "loss": 0.6561,
+      "step": 1227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523128463348288e-06,
+      "loss": 0.6606,
+      "step": 1228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522195340980267e-06,
+      "loss": 0.7074,
+      "step": 1229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521261352364437e-06,
+      "loss": 0.89,
+      "step": 1230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520326497679711e-06,
+      "loss": 1.0782,
+      "step": 1231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51939077710516e-06,
+      "loss": 0.6273,
+      "step": 1232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518454190820031e-06,
+      "loss": 1.1741,
+      "step": 1233
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517516739003726e-06,
+      "loss": 0.7169,
+      "step": 1234
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516578421835823e-06,
+      "loss": 0.8182,
+      "step": 1235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515639239496058e-06,
+      "loss": 0.8116,
+      "step": 1236
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514699192164338e-06,
+      "loss": 0.938,
+      "step": 1237
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513758280020733e-06,
+      "loss": 0.8876,
+      "step": 1238
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512816503245476e-06,
+      "loss": 0.7381,
+      "step": 1239
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511873862018975e-06,
+      "loss": 1.1015,
+      "step": 1240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510930356521793e-06,
+      "loss": 0.8122,
+      "step": 1241
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509985986934663e-06,
+      "loss": 0.946,
+      "step": 1242
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509040753438484e-06,
+      "loss": 0.6131,
+      "step": 1243
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50809465621432e-06,
+      "loss": 0.7344,
+      "step": 1244
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507147695443402e-06,
+      "loss": 1.1855,
+      "step": 1245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506199871307122e-06,
+      "loss": 0.9389,
+      "step": 1246
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505251183987041e-06,
+      "loss": 0.8541,
+      "step": 1247
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504301633664886e-06,
+      "loss": 0.7611,
+      "step": 1248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503351220522548e-06,
+      "loss": 0.7975,
+      "step": 1249
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50239994474208e-06,
+      "loss": 0.8758,
+      "step": 1250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501447806505706e-06,
+      "loss": 0.9387,
+      "step": 1251
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50049480599581e-06,
+      "loss": 0.7101,
+      "step": 1252
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49954094339495e-06,
+      "loss": 0.9974,
+      "step": 1253
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498586218885835e-06,
+      "loss": 0.815,
+      "step": 1254
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497630632651352e-06,
+      "loss": 0.5935,
+      "step": 1255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496674184874543e-06,
+      "loss": 0.9934,
+      "step": 1256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495716875738628e-06,
+      "loss": 1.0424,
+      "step": 1257
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494758705426978e-06,
+      "loss": 0.6323,
+      "step": 1258
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493799674123136e-06,
+      "loss": 0.8528,
+      "step": 1259
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49283978201081e-06,
+      "loss": 0.844,
+      "step": 1260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491879029273871e-06,
+      "loss": 0.6444,
+      "step": 1261
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490917416096357e-06,
+      "loss": 0.8069,
+      "step": 1262
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489954942662469e-06,
+      "loss": 0.7733,
+      "step": 1263
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488991609156572e-06,
+      "loss": 1.0491,
+      "step": 1264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488027415763198e-06,
+      "loss": 0.9631,
+      "step": 1265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487062362667045e-06,
+      "loss": 0.7605,
+      "step": 1266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48609645005297e-06,
+      "loss": 0.6872,
+      "step": 1267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485129678106e-06,
+      "loss": 0.638,
+      "step": 1268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484162047011324e-06,
+      "loss": 0.8496,
+      "step": 1269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483193556954295e-06,
+      "loss": 1.0076,
+      "step": 1270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482224208120435e-06,
+      "loss": 0.7387,
+      "step": 1271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481254000695425e-06,
+      "loss": 0.9184,
+      "step": 1272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480282934865113e-06,
+      "loss": 0.6324,
+      "step": 1273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479311010815512e-06,
+      "loss": 0.6467,
+      "step": 1274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4783382287328e-06,
+      "loss": 0.9529,
+      "step": 1275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477364588803313e-06,
+      "loss": 0.6535,
+      "step": 1276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476390091213562e-06,
+      "loss": 0.81,
+      "step": 1277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.475414736150213e-06,
+      "loss": 1.1219,
+      "step": 1278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474438523800102e-06,
+      "loss": 0.5334,
+      "step": 1279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473461454350226e-06,
+      "loss": 0.8144,
+      "step": 1280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472483527987747e-06,
+      "loss": 0.829,
+      "step": 1281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471504744899991e-06,
+      "loss": 0.9584,
+      "step": 1282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47052510527445e-06,
+      "loss": 0.7461,
+      "step": 1283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46954460929878e-06,
+      "loss": 0.6331,
+      "step": 1284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468563257160795e-06,
+      "loss": 0.6089,
+      "step": 1285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46758104904848e-06,
+      "loss": 0.8268,
+      "step": 1286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466597985149982e-06,
+      "loss": 0.9259,
+      "step": 1287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465614065653611e-06,
+      "loss": 0.7074,
+      "step": 1288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464629290747844e-06,
+      "loss": 0.7204,
+      "step": 1289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463643660621314e-06,
+      "loss": 0.5231,
+      "step": 1290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462657175462827e-06,
+      "loss": 0.8309,
+      "step": 1291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461669835461349e-06,
+      "loss": 0.641,
+      "step": 1292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460681640806007e-06,
+      "loss": 0.6462,
+      "step": 1293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459692591686095e-06,
+      "loss": 0.6694,
+      "step": 1294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458702688291072e-06,
+      "loss": 0.603,
+      "step": 1295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457711930810558e-06,
+      "loss": 0.9896,
+      "step": 1296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456720319434337e-06,
+      "loss": 0.6832,
+      "step": 1297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455727854352356e-06,
+      "loss": 0.4687,
+      "step": 1298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454734535754727e-06,
+      "loss": 1.1759,
+      "step": 1299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453740363831725e-06,
+      "loss": 0.749,
+      "step": 1300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452745338773787e-06,
+      "loss": 0.4832,
+      "step": 1301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451749460771517e-06,
+      "loss": 1.004,
+      "step": 1302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450752730015677e-06,
+      "loss": 0.6229,
+      "step": 1303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4497551466972e-06,
+      "loss": 0.6282,
+      "step": 1304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448756711007174e-06,
+      "loss": 0.5743,
+      "step": 1305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447757423136856e-06,
+      "loss": 1.0535,
+      "step": 1306
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446757283277663e-06,
+      "loss": 0.9805,
+      "step": 1307
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445756291621177e-06,
+      "loss": 0.6783,
+      "step": 1308
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444754448359144e-06,
+      "loss": 1.1517,
+      "step": 1309
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443751753683467e-06,
+      "loss": 0.8523,
+      "step": 1310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442748207786225e-06,
+      "loss": 0.8278,
+      "step": 1311
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441743810859645e-06,
+      "loss": 0.5689,
+      "step": 1312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440738563096126e-06,
+      "loss": 1.0016,
+      "step": 1313
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439732464688229e-06,
+      "loss": 0.6784,
+      "step": 1314
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438725515828675e-06,
+      "loss": 0.8687,
+      "step": 1315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437717716710353e-06,
+      "loss": 0.9754,
+      "step": 1316
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436709067526309e-06,
+      "loss": 1.2597,
+      "step": 1317
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435699568469755e-06,
+      "loss": 1.0691,
+      "step": 1318
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434689219734063e-06,
+      "loss": 1.1289,
+      "step": 1319
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433678021512774e-06,
+      "loss": 0.4729,
+      "step": 1320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432665973999587e-06,
+      "loss": 0.9728,
+      "step": 1321
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431653077388364e-06,
+      "loss": 0.9091,
+      "step": 1322
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430639331873129e-06,
+      "loss": 0.9709,
+      "step": 1323
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42962473764807e-06,
+      "loss": 0.4992,
+      "step": 1324
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428609294907539e-06,
+      "loss": 0.7106,
+      "step": 1325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427593003846046e-06,
+      "loss": 0.6576,
+      "step": 1326
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42657586465827e-06,
+      "loss": 0.9015,
+      "step": 1327
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425557877539046e-06,
+      "loss": 0.5918,
+      "step": 1328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424539042683375e-06,
+      "loss": 0.4935,
+      "step": 1329
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423519360286419e-06,
+      "loss": 1.1137,
+      "step": 1330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422498830543506e-06,
+      "loss": 1.0143,
+      "step": 1331
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421477453650118e-06,
+      "loss": 0.6772,
+      "step": 1332
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42045522980191e-06,
+      "loss": 0.7899,
+      "step": 1333
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41943215919469e-06,
+      "loss": 0.5874,
+      "step": 1334
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418408242024433e-06,
+      "loss": 0.6487,
+      "step": 1335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417383478487275e-06,
+      "loss": 0.5056,
+      "step": 1336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416357868779516e-06,
+      "loss": 0.8424,
+      "step": 1337
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415331413097614e-06,
+      "loss": 1.1857,
+      "step": 1338
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414304111638194e-06,
+      "loss": 1.0542,
+      "step": 1339
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413275964598037e-06,
+      "loss": 0.9315,
+      "step": 1340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41224697217409e-06,
+      "loss": 1.0176,
+      "step": 1341
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411217134563466e-06,
+      "loss": 0.9225,
+      "step": 1342
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41018645196343e-06,
+      "loss": 0.7161,
+      "step": 1343
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409154924571416e-06,
+      "loss": 0.7249,
+      "step": 1344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408122552585019e-06,
+      "loss": 0.7425,
+      "step": 1345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407089336201993e-06,
+      "loss": 0.8569,
+      "step": 1346
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406055275620257e-06,
+      "loss": 0.8851,
+      "step": 1347
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405020371037887e-06,
+      "loss": 0.9227,
+      "step": 1348
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403984622653128e-06,
+      "loss": 1.0702,
+      "step": 1349
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402948030664379e-06,
+      "loss": 1.1319,
+      "step": 1350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401910595270208e-06,
+      "loss": 0.6795,
+      "step": 1351
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400872316669336e-06,
+      "loss": 0.6159,
+      "step": 1352
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399833195060654e-06,
+      "loss": 0.9216,
+      "step": 1353
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398793230643209e-06,
+      "loss": 0.3606,
+      "step": 1354
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39775242361621e-06,
+      "loss": 0.8156,
+      "step": 1355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39671077417903e-06,
+      "loss": 0.7462,
+      "step": 1356
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395668282531202e-06,
+      "loss": 0.8179,
+      "step": 1357
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39462494887242e-06,
+      "loss": 0.4831,
+      "step": 1358
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393580773402538e-06,
+      "loss": 1.1654,
+      "step": 1359
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392535756321574e-06,
+      "loss": 0.9547,
+      "step": 1360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391489897829706e-06,
+      "loss": 1.0499,
+      "step": 1361
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390443198127274e-06,
+      "loss": 0.868,
+      "step": 1362
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389395657414778e-06,
+      "loss": 0.8952,
+      "step": 1363
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388347275892874e-06,
+      "loss": 0.7829,
+      "step": 1364
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387298053762394e-06,
+      "loss": 0.4734,
+      "step": 1365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386247991224315e-06,
+      "loss": 0.7338,
+      "step": 1366
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385197088479782e-06,
+      "loss": 0.6195,
+      "step": 1367
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384145345730101e-06,
+      "loss": 1.17,
+      "step": 1368
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38309276317674e-06,
+      "loss": 0.5068,
+      "step": 1369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382039341021324e-06,
+      "loss": 0.5829,
+      "step": 1370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38098507946564e-06,
+      "loss": 0.8676,
+      "step": 1371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37992997871164e-06,
+      "loss": 0.7961,
+      "step": 1372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378874038961432e-06,
+      "loss": 0.9936,
+      "step": 1373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377817260417286e-06,
+      "loss": 0.6669,
+      "step": 1374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376759643281633e-06,
+      "loss": 0.9096,
+      "step": 1375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375701187757062e-06,
+      "loss": 0.9195,
+      "step": 1376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37464189404633e-06,
+      "loss": 0.641,
+      "step": 1377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373581762352349e-06,
+      "loss": 0.4717,
+      "step": 1378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372520792878187e-06,
+      "loss": 1.0136,
+      "step": 1379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371458985827085e-06,
+      "loss": 0.8411,
+      "step": 1380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37039634140243e-06,
+      "loss": 0.9216,
+      "step": 1381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369332859807782e-06,
+      "loss": 1.0462,
+      "step": 1382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368268541246854e-06,
+      "loss": 0.8469,
+      "step": 1383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36720338592352e-06,
+      "loss": 1.3339,
+      "step": 1384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366137394041817e-06,
+      "loss": 0.8735,
+      "step": 1385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365070565805941e-06,
+      "loss": 1.0422,
+      "step": 1386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364002901420248e-06,
+      "loss": 0.921,
+      "step": 1387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362934401089251e-06,
+      "loss": 0.6619,
+      "step": 1388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361865065017633e-06,
+      "loss": 0.907,
+      "step": 1389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360794893410223e-06,
+      "loss": 0.4789,
+      "step": 1390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35972388647202e-06,
+      "loss": 0.6113,
+      "step": 1391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358652044408183e-06,
+      "loss": 0.5659,
+      "step": 1392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357579367424025e-06,
+      "loss": 0.7502,
+      "step": 1393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356505855725024e-06,
+      "loss": 0.8365,
+      "step": 1394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355431509516816e-06,
+      "loss": 0.4929,
+      "step": 1395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354356329005197e-06,
+      "loss": 0.7516,
+      "step": 1396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353280314396124e-06,
+      "loss": 0.6257,
+      "step": 1397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352203465895709e-06,
+      "loss": 1.1452,
+      "step": 1398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351125783710232e-06,
+      "loss": 0.9393,
+      "step": 1399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350047268046125e-06,
+      "loss": 0.8839,
+      "step": 1400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348967919109985e-06,
+      "loss": 1.0797,
+      "step": 1401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347887737108566e-06,
+      "loss": 0.7106,
+      "step": 1402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346806722248778e-06,
+      "loss": 0.684,
+      "step": 1403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3457248747377e-06,
+      "loss": 0.4193,
+      "step": 1404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34464219478256e-06,
+      "loss": 0.9363,
+      "step": 1405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343558682590757e-06,
+      "loss": 0.8091,
+      "step": 1406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342474338369836e-06,
+      "loss": 0.3598,
+      "step": 1407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341389162327512e-06,
+      "loss": 0.8974,
+      "step": 1408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340303154671654e-06,
+      "loss": 0.8359,
+      "step": 1409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339216315610293e-06,
+      "loss": 0.7162,
+      "step": 1410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338128645351619e-06,
+      "loss": 0.9476,
+      "step": 1411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337040144103977e-06,
+      "loss": 1.1742,
+      "step": 1412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335950812075877e-06,
+      "loss": 1.157,
+      "step": 1413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334860649475987e-06,
+      "loss": 0.8925,
+      "step": 1414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33376965651313e-06,
+      "loss": 0.4252,
+      "step": 1415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332677833396291e-06,
+      "loss": 1.3273,
+      "step": 1416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331585180334615e-06,
+      "loss": 1.2196,
+      "step": 1417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330491697537407e-06,
+      "loss": 0.9344,
+      "step": 1418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329397385214124e-06,
+      "loss": 1.0966,
+      "step": 1419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328302243574389e-06,
+      "loss": 0.6816,
+      "step": 1420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327206272827983e-06,
+      "loss": 0.589,
+      "step": 1421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326109473184842e-06,
+      "loss": 0.5566,
+      "step": 1422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325011844855066e-06,
+      "loss": 0.4448,
+      "step": 1423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323913388048906e-06,
+      "loss": 0.5863,
+      "step": 1424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32281410297678e-06,
+      "loss": 0.9298,
+      "step": 1425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321713989849263e-06,
+      "loss": 0.5475,
+      "step": 1426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320613048877083e-06,
+      "loss": 0.4859,
+      "step": 1427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319511280271132e-06,
+      "loss": 0.6835,
+      "step": 1428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31840868424246e-06,
+      "loss": 0.6906,
+      "step": 1429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317305261002273e-06,
+      "loss": 0.9142,
+      "step": 1430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316201010761937e-06,
+      "loss": 0.7998,
+      "step": 1431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315095933732978e-06,
+      "loss": 1.0624,
+      "step": 1432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313990030127075e-06,
+      "loss": 0.5058,
+      "step": 1433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312883300156075e-06,
+      "loss": 0.6056,
+      "step": 1434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31177574403197e-06,
+      "loss": 1.2737,
+      "step": 1435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310667361966923e-06,
+      "loss": 1.0151,
+      "step": 1436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309558154173249e-06,
+      "loss": 0.4315,
+      "step": 1437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308448120863421e-06,
+      "loss": 0.7839,
+      "step": 1438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30733726225007e-06,
+      "loss": 1.2252,
+      "step": 1439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306225578545989e-06,
+      "loss": 0.4657,
+      "step": 1440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305113069964124e-06,
+      "loss": 1.0066,
+      "step": 1441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303999736717582e-06,
+      "loss": 0.6317,
+      "step": 1442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302885579019626e-06,
+      "loss": 1.2422,
+      "step": 1443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301770597083681e-06,
+      "loss": 0.9085,
+      "step": 1444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300654791123324e-06,
+      "loss": 0.8271,
+      "step": 1445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.299538161352297e-06,
+      "loss": 0.9793,
+      "step": 1446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298420707984492e-06,
+      "loss": 0.8102,
+      "step": 1447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297302431233962e-06,
+      "loss": 0.6626,
+      "step": 1448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29618333131492e-06,
+      "loss": 0.7632,
+      "step": 1449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295063408441736e-06,
+      "loss": 0.9572,
+      "step": 1450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293942662828934e-06,
+      "loss": 0.6642,
+      "step": 1451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292821094691198e-06,
+      "loss": 0.8856,
+      "step": 1452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291698704243371e-06,
+      "loss": 0.961,
+      "step": 1453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.290575491700453e-06,
+      "loss": 0.5692,
+      "step": 1454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289451457277598e-06,
+      "loss": 0.7721,
+      "step": 1455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288326601190123e-06,
+      "loss": 0.7146,
+      "step": 1456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2872009236535e-06,
+      "loss": 0.6568,
+      "step": 1457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28607442488335e-06,
+      "loss": 0.9581,
+      "step": 1458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284947105095469e-06,
+      "loss": 0.6741,
+      "step": 1459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283818964505797e-06,
+      "loss": 0.8712,
+      "step": 1460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282690003330435e-06,
+      "loss": 0.8225,
+      "step": 1461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281560221785636e-06,
+      "loss": 0.5574,
+      "step": 1462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280429620087822e-06,
+      "loss": 0.4541,
+      "step": 1463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279298198453562e-06,
+      "loss": 0.5724,
+      "step": 1464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278165957099584e-06,
+      "loss": 1.0968,
+      "step": 1465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277032896242777e-06,
+      "loss": 0.768,
+      "step": 1466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275899016100181e-06,
+      "loss": 0.6644,
+      "step": 1467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274764316888998e-06,
+      "loss": 0.7904,
+      "step": 1468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273628798826585e-06,
+      "loss": 0.7144,
+      "step": 1469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272492462130454e-06,
+      "loss": 0.6851,
+      "step": 1470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271355307018279e-06,
+      "loss": 1.1679,
+      "step": 1471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270217333707884e-06,
+      "loss": 0.9669,
+      "step": 1472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269078542417254e-06,
+      "loss": 0.8381,
+      "step": 1473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267938933364531e-06,
+      "loss": 0.9276,
+      "step": 1474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266798506768012e-06,
+      "loss": 0.8054,
+      "step": 1475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265657262846152e-06,
+      "loss": 0.7549,
+      "step": 1476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264515201817558e-06,
+      "loss": 0.9973,
+      "step": 1477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263372323901001e-06,
+      "loss": 0.8608,
+      "step": 1478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262228629315406e-06,
+      "loss": 0.6801,
+      "step": 1479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261084118279846e-06,
+      "loss": 1.0755,
+      "step": 1480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259938791013566e-06,
+      "loss": 0.9771,
+      "step": 1481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258792647735953e-06,
+      "loss": 1.0579,
+      "step": 1482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257645688666557e-06,
+      "loss": 0.5763,
+      "step": 1483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256497914025085e-06,
+      "loss": 1.0517,
+      "step": 1484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2553493240314e-06,
+      "loss": 0.7668,
+      "step": 1485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254199918905515e-06,
+      "loss": 1.1062,
+      "step": 1486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253049698867608e-06,
+      "loss": 0.9081,
+      "step": 1487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251898664138008e-06,
+      "loss": 0.6516,
+      "step": 1488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2507468149372e-06,
+      "loss": 0.5956,
+      "step": 1489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249594151485828e-06,
+      "loss": 1.0502,
+      "step": 1490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248440674004688e-06,
+      "loss": 0.9925,
+      "step": 1491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247286382714738e-06,
+      "loss": 1.1596,
+      "step": 1492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246131277837082e-06,
+      "loss": 0.6221,
+      "step": 1493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244975359592991e-06,
+      "loss": 0.7315,
+      "step": 1494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243818628203885e-06,
+      "loss": 0.9679,
+      "step": 1495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242661083891338e-06,
+      "loss": 0.8024,
+      "step": 1496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241502726877089e-06,
+      "loss": 0.8927,
+      "step": 1497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240343557383022e-06,
+      "loss": 0.7853,
+      "step": 1498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239183575631183e-06,
+      "loss": 0.8087,
+      "step": 1499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238022781843773e-06,
+      "loss": 0.4597,
+      "step": 1500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236861176243146e-06,
+      "loss": 1.0914,
+      "step": 1501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235698759051813e-06,
+      "loss": 1.0767,
+      "step": 1502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234535530492442e-06,
+      "loss": 0.6976,
+      "step": 1503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233371490787853e-06,
+      "loss": 0.992,
+      "step": 1504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232206640161026e-06,
+      "loss": 0.7938,
+      "step": 1505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23104097883509e-06,
+      "loss": 1.0518,
+      "step": 1506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229874507033336e-06,
+      "loss": 1.094,
+      "step": 1507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228707224979206e-06,
+      "loss": 0.5993,
+      "step": 1508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.227539132896297e-06,
+      "loss": 0.8554,
+      "step": 1509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226370231008363e-06,
+      "loss": 1.0187,
+      "step": 1510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225200519539316e-06,
+      "loss": 1.1097,
+      "step": 1511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224029998713217e-06,
+      "loss": 0.7152,
+      "step": 1512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222858668754284e-06,
+      "loss": 0.7436,
+      "step": 1513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221686529886893e-06,
+      "loss": 0.957,
+      "step": 1514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220513582335571e-06,
+      "loss": 0.8968,
+      "step": 1515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219339826325001e-06,
+      "loss": 0.7622,
+      "step": 1516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218165262080024e-06,
+      "loss": 0.8936,
+      "step": 1517
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216989889825631e-06,
+      "loss": 0.8917,
+      "step": 1518
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215813709786972e-06,
+      "loss": 0.9586,
+      "step": 1519
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214636722189347e-06,
+      "loss": 1.0104,
+      "step": 1520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213458927258217e-06,
+      "loss": 0.7307,
+      "step": 1521
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212280325219193e-06,
+      "loss": 0.737,
+      "step": 1522
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.21110091629804e-06,
+      "loss": 1.0477,
+      "step": 1523
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20992070072068e-06,
+      "loss": 0.7221,
+      "step": 1524
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208739678713193e-06,
+      "loss": 0.3767,
+      "step": 1525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207557850501803e-06,
+      "loss": 0.8739,
+      "step": 1526
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206375216312898e-06,
+      "loss": 0.9618,
+      "step": 1527
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205191776373017e-06,
+      "loss": 0.7006,
+      "step": 1528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.204007530908851e-06,
+      "loss": 0.7798,
+      "step": 1529
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202822480147253e-06,
+      "loss": 0.8736,
+      "step": 1530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20163662431522e-06,
+      "loss": 0.9771,
+      "step": 1531
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200449963639912e-06,
+      "loss": 0.8743,
+      "step": 1532
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199262498348637e-06,
+      "loss": 1.2335,
+      "step": 1533
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.198074228668861e-06,
+      "loss": 0.3436,
+      "step": 1534
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196885154828201e-06,
+      "loss": 0.6558,
+      "step": 1535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19569527705443e-06,
+      "loss": 0.3636,
+      "step": 1536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194504595575478e-06,
+      "loss": 0.6676,
+      "step": 1537
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193313110619419e-06,
+      "loss": 0.9302,
+      "step": 1538
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192120822414495e-06,
+      "loss": 1.1884,
+      "step": 1539
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19092773118909e-06,
+      "loss": 0.6018,
+      "step": 1540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.189733837171747e-06,
+      "loss": 0.8575,
+      "step": 1541
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188539140591162e-06,
+      "loss": 1.1859,
+      "step": 1542
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187343641676186e-06,
+      "loss": 0.9403,
+      "step": 1543
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18614734065582e-06,
+      "loss": 0.3785,
+      "step": 1544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184950237759225e-06,
+      "loss": 0.841,
+      "step": 1545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183752333215706e-06,
+      "loss": 0.9077,
+      "step": 1546
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182553627254734e-06,
+      "loss": 0.982,
+      "step": 1547
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18135412010592e-06,
+      "loss": 0.7308,
+      "step": 1548
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18015381199904e-06,
+      "loss": 0.8112,
+      "step": 1549
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178952703164017e-06,
+      "loss": 1.1539,
+      "step": 1550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177750793830928e-06,
+      "loss": 0.7298,
+      "step": 1551
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176548084230007e-06,
+      "loss": 0.6867,
+      "step": 1552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175344574591637e-06,
+      "loss": 0.749,
+      "step": 1553
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174140265146356e-06,
+      "loss": 0.6064,
+      "step": 1554
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172935156124856e-06,
+      "loss": 0.9529,
+      "step": 1555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17172924775798e-06,
+      "loss": 0.7062,
+      "step": 1556
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170522540276728e-06,
+      "loss": 1.254,
+      "step": 1557
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169315033912246e-06,
+      "loss": 0.6558,
+      "step": 1558
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16810672889584e-06,
+      "loss": 0.8502,
+      "step": 1559
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166897625458968e-06,
+      "loss": 0.5891,
+      "step": 1560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165687723833239e-06,
+      "loss": 0.8094,
+      "step": 1561
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164477024250413e-06,
+      "loss": 1.0524,
+      "step": 1562
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163265526942407e-06,
+      "loss": 1.0272,
+      "step": 1563
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162053232141288e-06,
+      "loss": 1.0207,
+      "step": 1564
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16084014007928e-06,
+      "loss": 1.0305,
+      "step": 1565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159626250988751e-06,
+      "loss": 1.2841,
+      "step": 1566
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.158411565102232e-06,
+      "loss": 0.9123,
+      "step": 1567
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1571960826524e-06,
+      "loss": 0.8264,
+      "step": 1568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155979803872085e-06,
+      "loss": 0.623,
+      "step": 1569
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154762728994271e-06,
+      "loss": 0.6602,
+      "step": 1570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153544858252099e-06,
+      "loss": 1.001,
+      "step": 1571
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152326191878851e-06,
+      "loss": 1.0652,
+      "step": 1572
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151106730107975e-06,
+      "loss": 0.5507,
+      "step": 1573
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14988647317306e-06,
+      "loss": 1.1284,
+      "step": 1574
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.148665421307852e-06,
+      "loss": 0.8295,
+      "step": 1575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147443574746254e-06,
+      "loss": 1.067,
+      "step": 1576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14622093372231e-06,
+      "loss": 0.8924,
+      "step": 1577
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144997498470228e-06,
+      "loss": 1.0305,
+      "step": 1578
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143773269224359e-06,
+      "loss": 0.8572,
+      "step": 1579
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142548246219212e-06,
+      "loss": 0.7564,
+      "step": 1580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141322429689445e-06,
+      "loss": 0.7492,
+      "step": 1581
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14009581986987e-06,
+      "loss": 1.2181,
+      "step": 1582
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138868416995447e-06,
+      "loss": 0.5831,
+      "step": 1583
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137640221301296e-06,
+      "loss": 1.0275,
+      "step": 1584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13641123302268e-06,
+      "loss": 0.6659,
+      "step": 1585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135181452395017e-06,
+      "loss": 0.9083,
+      "step": 1586
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133950879653878e-06,
+      "loss": 0.9396,
+      "step": 1587
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132719515034985e-06,
+      "loss": 0.6716,
+      "step": 1588
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131487358774213e-06,
+      "loss": 0.5967,
+      "step": 1589
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130254411107587e-06,
+      "loss": 1.1339,
+      "step": 1590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.129020672271283e-06,
+      "loss": 0.9742,
+      "step": 1591
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12778614250163e-06,
+      "loss": 0.631,
+      "step": 1592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126550822035108e-06,
+      "loss": 0.8183,
+      "step": 1593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125314711108348e-06,
+      "loss": 0.8861,
+      "step": 1594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124077809958132e-06,
+      "loss": 0.9989,
+      "step": 1595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122840118821397e-06,
+      "loss": 1.0007,
+      "step": 1596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121601637935224e-06,
+      "loss": 1.1235,
+      "step": 1597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120362367536857e-06,
+      "loss": 0.9115,
+      "step": 1598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119122307863677e-06,
+      "loss": 0.791,
+      "step": 1599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117881459153226e-06,
+      "loss": 0.6666,
+      "step": 1600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116639821643196e-06,
+      "loss": 1.1231,
+      "step": 1601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115397395571426e-06,
+      "loss": 0.7078,
+      "step": 1602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11415418117591e-06,
+      "loss": 0.6815,
+      "step": 1603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.112910178694791e-06,
+      "loss": 0.6806,
+      "step": 1604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111665388366364e-06,
+      "loss": 0.8112,
+      "step": 1605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110419810429074e-06,
+      "loss": 0.7733,
+      "step": 1606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109173445121518e-06,
+      "loss": 1.2444,
+      "step": 1607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107926292682442e-06,
+      "loss": 0.3813,
+      "step": 1608
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106678353350745e-06,
+      "loss": 0.8678,
+      "step": 1609
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105429627365474e-06,
+      "loss": 1.049,
+      "step": 1610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10418011496583e-06,
+      "loss": 0.8039,
+      "step": 1611
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102929816391164e-06,
+      "loss": 0.8802,
+      "step": 1612
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101678731880974e-06,
+      "loss": 0.7961,
+      "step": 1613
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10042686167491e-06,
+      "loss": 0.8367,
+      "step": 1614
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09917420601278e-06,
+      "loss": 0.5302,
+      "step": 1615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09792076513453e-06,
+      "loss": 0.6874,
+      "step": 1616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096666539280264e-06,
+      "loss": 0.428,
+      "step": 1617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.095411528690237e-06,
+      "loss": 1.0295,
+      "step": 1618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.094155733604848e-06,
+      "loss": 0.4697,
+      "step": 1619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092899154264654e-06,
+      "loss": 0.7448,
+      "step": 1620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091641790910358e-06,
+      "loss": 1.0814,
+      "step": 1621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090383643782815e-06,
+      "loss": 1.0652,
+      "step": 1622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.089124713123025e-06,
+      "loss": 0.5728,
+      "step": 1623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087864999172145e-06,
+      "loss": 1.1686,
+      "step": 1624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086604502171478e-06,
+      "loss": 0.8696,
+      "step": 1625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085343222362481e-06,
+      "loss": 0.9646,
+      "step": 1626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084081159986754e-06,
+      "loss": 0.953,
+      "step": 1627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082818315286054e-06,
+      "loss": 0.8007,
+      "step": 1628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081554688502285e-06,
+      "loss": 1.1211,
+      "step": 1629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0802902798775e-06,
+      "loss": 0.867,
+      "step": 1630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0790250896539e-06,
+      "loss": 0.7906,
+      "step": 1631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077759118073839e-06,
+      "loss": 0.8518,
+      "step": 1632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076492365379824e-06,
+      "loss": 0.7597,
+      "step": 1633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0752248318145e-06,
+      "loss": 0.8486,
+      "step": 1634
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073956517620677e-06,
+      "loss": 0.6313,
+      "step": 1635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072687423041302e-06,
+      "loss": 1.0295,
+      "step": 1636
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071417548319476e-06,
+      "loss": 0.916,
+      "step": 1637
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070146893698451e-06,
+      "loss": 0.5709,
+      "step": 1638
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.068875459421625e-06,
+      "loss": 0.775,
+      "step": 1639
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06760324573255e-06,
+      "loss": 0.7762,
+      "step": 1640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066330252874922e-06,
+      "loss": 0.936,
+      "step": 1641
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065056481092592e-06,
+      "loss": 0.7002,
+      "step": 1642
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063781930629552e-06,
+      "loss": 0.8313,
+      "step": 1643
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062506601729952e-06,
+      "loss": 0.8338,
+      "step": 1644
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061230494638087e-06,
+      "loss": 0.9568,
+      "step": 1645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0599536095984e-06,
+      "loss": 0.5942,
+      "step": 1646
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.058675946855484e-06,
+      "loss": 0.9734,
+      "step": 1647
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057397506654085e-06,
+      "loss": 1.0701,
+      "step": 1648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056118289239087e-06,
+      "loss": 0.8582,
+      "step": 1649
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054838294855537e-06,
+      "loss": 0.67,
+      "step": 1650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053557523748621e-06,
+      "loss": 0.6062,
+      "step": 1651
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.052275976163679e-06,
+      "loss": 0.8157,
+      "step": 1652
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050993652346193e-06,
+      "loss": 0.6175,
+      "step": 1653
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049710552541802e-06,
+      "loss": 0.8974,
+      "step": 1654
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04842667699629e-06,
+      "loss": 0.7023,
+      "step": 1655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047142025955585e-06,
+      "loss": 0.6459,
+      "step": 1656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045856599665773e-06,
+      "loss": 0.8497,
+      "step": 1657
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044570398373081e-06,
+      "loss": 0.7202,
+      "step": 1658
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043283422323887e-06,
+      "loss": 1.0734,
+      "step": 1659
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04199567176472e-06,
+      "loss": 1.059,
+      "step": 1660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040707146942248e-06,
+      "loss": 0.698,
+      "step": 1661
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0394178481033e-06,
+      "loss": 1.0207,
+      "step": 1662
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038127775494846e-06,
+      "loss": 0.9272,
+      "step": 1663
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.036836929364007e-06,
+      "loss": 0.8049,
+      "step": 1664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.035545309958048e-06,
+      "loss": 0.8483,
+      "step": 1665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034252917524384e-06,
+      "loss": 1.3311,
+      "step": 1666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032959752310582e-06,
+      "loss": 0.6728,
+      "step": 1667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03166581456435e-06,
+      "loss": 0.8834,
+      "step": 1668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030371104533552e-06,
+      "loss": 0.9571,
+      "step": 1669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029075622466195e-06,
+      "loss": 0.8295,
+      "step": 1670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02777936861043e-06,
+      "loss": 0.891,
+      "step": 1671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026482343214567e-06,
+      "loss": 1.0396,
+      "step": 1672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025184546527054e-06,
+      "loss": 0.6949,
+      "step": 1673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023885978796492e-06,
+      "loss": 0.7563,
+      "step": 1674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022586640271624e-06,
+      "loss": 0.6692,
+      "step": 1675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021286531201347e-06,
+      "loss": 1.023,
+      "step": 1676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019985651834703e-06,
+      "loss": 0.6396,
+      "step": 1677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.018684002420883e-06,
+      "loss": 0.5067,
+      "step": 1678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017381583209219e-06,
+      "loss": 0.7361,
+      "step": 1679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016078394449201e-06,
+      "loss": 0.8945,
+      "step": 1680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01477443639046e-06,
+      "loss": 1.0121,
+      "step": 1681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013469709282772e-06,
+      "loss": 0.846,
+      "step": 1682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012164213376066e-06,
+      "loss": 0.6968,
+      "step": 1683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010857948920417e-06,
+      "loss": 0.431,
+      "step": 1684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009550916166046e-06,
+      "loss": 1.0497,
+      "step": 1685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.008243115363318e-06,
+      "loss": 0.6902,
+      "step": 1686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006934546762749e-06,
+      "loss": 0.866,
+      "step": 1687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005625210615006e-06,
+      "loss": 0.7743,
+      "step": 1688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004315107170893e-06,
+      "loss": 0.8942,
+      "step": 1689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003004236681369e-06,
+      "loss": 1.0881,
+      "step": 1690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00169259939754e-06,
+      "loss": 0.5686,
+      "step": 1691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000380195570651e-06,
+      "loss": 0.8261,
+      "step": 1692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.999067025452103e-06,
+      "loss": 0.7115,
+      "step": 1693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997753089293438e-06,
+      "loss": 0.8049,
+      "step": 1694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996438387346347e-06,
+      "loss": 0.5863,
+      "step": 1695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99512291986267e-06,
+      "loss": 0.814,
+      "step": 1696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993806687094387e-06,
+      "loss": 0.7317,
+      "step": 1697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.992489689293629e-06,
+      "loss": 0.489,
+      "step": 1698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991171926712676e-06,
+      "loss": 0.6993,
+      "step": 1699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.989853399603949e-06,
+      "loss": 0.8245,
+      "step": 1700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988534108220021e-06,
+      "loss": 1.2239,
+      "step": 1701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987214052813605e-06,
+      "loss": 0.7957,
+      "step": 1702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985893233637566e-06,
+      "loss": 0.6097,
+      "step": 1703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98457165094491e-06,
+      "loss": 0.9219,
+      "step": 1704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9832493049888e-06,
+      "loss": 0.8989,
+      "step": 1705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981926196022529e-06,
+      "loss": 0.9018,
+      "step": 1706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980602324299548e-06,
+      "loss": 1.0271,
+      "step": 1707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979277690073452e-06,
+      "loss": 0.9223,
+      "step": 1708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97795229359798e-06,
+      "loss": 0.7291,
+      "step": 1709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976626135127018e-06,
+      "loss": 0.9998,
+      "step": 1710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9752992149146e-06,
+      "loss": 1.2814,
+      "step": 1711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973971533214897e-06,
+      "loss": 0.7935,
+      "step": 1712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.972643090282242e-06,
+      "loss": 0.8808,
+      "step": 1713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971313886371098e-06,
+      "loss": 0.8771,
+      "step": 1714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96998392173608e-06,
+      "loss": 0.9117,
+      "step": 1715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.968653196631954e-06,
+      "loss": 1.0658,
+      "step": 1716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967321711313624e-06,
+      "loss": 0.7869,
+      "step": 1717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96598946603614e-06,
+      "loss": 0.8646,
+      "step": 1718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.964656461054702e-06,
+      "loss": 0.9689,
+      "step": 1719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.963322696624654e-06,
+      "loss": 0.8481,
+      "step": 1720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961988173001486e-06,
+      "loss": 0.6101,
+      "step": 1721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960652890440829e-06,
+      "loss": 1.1758,
+      "step": 1722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.959316849198463e-06,
+      "loss": 0.8139,
+      "step": 1723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957980049530314e-06,
+      "loss": 0.5795,
+      "step": 1724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956642491692453e-06,
+      "loss": 0.8428,
+      "step": 1725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955304175941096e-06,
+      "loss": 0.9521,
+      "step": 1726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953965102532602e-06,
+      "loss": 0.959,
+      "step": 1727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952625271723477e-06,
+      "loss": 0.591,
+      "step": 1728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951284683770372e-06,
+      "loss": 1.0581,
+      "step": 1729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.949943338930083e-06,
+      "loss": 0.8507,
+      "step": 1730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948601237459551e-06,
+      "loss": 0.7786,
+      "step": 1731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947258379615864e-06,
+      "loss": 1.0252,
+      "step": 1732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94591476565625e-06,
+      "loss": 0.7746,
+      "step": 1733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944570395838083e-06,
+      "loss": 1.2565,
+      "step": 1734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943225270418886e-06,
+      "loss": 1.107,
+      "step": 1735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941879389656327e-06,
+      "loss": 0.7494,
+      "step": 1736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.940532753808209e-06,
+      "loss": 1.3095,
+      "step": 1737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939185363132492e-06,
+      "loss": 0.9042,
+      "step": 1738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.937837217887273e-06,
+      "loss": 0.9962,
+      "step": 1739
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.936488318330794e-06,
+      "loss": 0.7008,
+      "step": 1740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935138664721447e-06,
+      "loss": 0.6114,
+      "step": 1741
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93378825731776e-06,
+      "loss": 0.9917,
+      "step": 1742
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.932437096378413e-06,
+      "loss": 0.4566,
+      "step": 1743
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931085182162227e-06,
+      "loss": 1.1239,
+      "step": 1744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929732514928167e-06,
+      "loss": 0.7278,
+      "step": 1745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92837909493534e-06,
+      "loss": 1.0885,
+      "step": 1746
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927024922443003e-06,
+      "loss": 1.0623,
+      "step": 1747
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925669997710556e-06,
+      "loss": 0.8133,
+      "step": 1748
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924314320997539e-06,
+      "loss": 0.8982,
+      "step": 1749
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.922957892563636e-06,
+      "loss": 0.9427,
+      "step": 1750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921600712668681e-06,
+      "loss": 1.0024,
+      "step": 1751
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920242781572645e-06,
+      "loss": 0.8806,
+      "step": 1752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91888409953565e-06,
+      "loss": 1.1989,
+      "step": 1753
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917524666817954e-06,
+      "loss": 0.9779,
+      "step": 1754
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916164483679963e-06,
+      "loss": 0.9897,
+      "step": 1755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91480355038223e-06,
+      "loss": 0.7829,
+      "step": 1756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913441867185446e-06,
+      "loss": 0.8124,
+      "step": 1757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912079434350447e-06,
+      "loss": 1.0114,
+      "step": 1758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910716252138215e-06,
+      "loss": 0.7291,
+      "step": 1759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909352320809873e-06,
+      "loss": 0.8029,
+      "step": 1760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90798764062669e-06,
+      "loss": 0.6499,
+      "step": 1761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.906622211850073e-06,
+      "loss": 1.0227,
+      "step": 1762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90525603474158e-06,
+      "loss": 1.0479,
+      "step": 1763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.903889109562908e-06,
+      "loss": 0.7939,
+      "step": 1764
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902521436575898e-06,
+      "loss": 0.695,
+      "step": 1765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.901153016042533e-06,
+      "loss": 0.6138,
+      "step": 1766
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89978384822494e-06,
+      "loss": 0.7491,
+      "step": 1767
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898413933385393e-06,
+      "loss": 0.5142,
+      "step": 1768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897043271786304e-06,
+      "loss": 1.0493,
+      "step": 1769
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895671863690227e-06,
+      "loss": 0.8514,
+      "step": 1770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894299709359864e-06,
+      "loss": 0.7277,
+      "step": 1771
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892926809058058e-06,
+      "loss": 0.5376,
+      "step": 1772
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891553163047792e-06,
+      "loss": 0.7476,
+      "step": 1773
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890178771592198e-06,
+      "loss": 1.1927,
+      "step": 1774
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.888803634954546e-06,
+      "loss": 1.0118,
+      "step": 1775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887427753398249e-06,
+      "loss": 0.8934,
+      "step": 1776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.886051127186861e-06,
+      "loss": 0.4946,
+      "step": 1777
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884673756584087e-06,
+      "loss": 0.8291,
+      "step": 1778
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883295641853764e-06,
+      "loss": 0.4486,
+      "step": 1779
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88191678325988e-06,
+      "loss": 0.7858,
+      "step": 1780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880537181066559e-06,
+      "loss": 0.8218,
+      "step": 1781
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.879156835538072e-06,
+      "loss": 1.1701,
+      "step": 1782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87777574693883e-06,
+      "loss": 1.0023,
+      "step": 1783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.876393915533387e-06,
+      "loss": 1.0929,
+      "step": 1784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87501134158644e-06,
+      "loss": 0.6682,
+      "step": 1785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873628025362825e-06,
+      "loss": 0.5756,
+      "step": 1786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872243967127527e-06,
+      "loss": 0.5596,
+      "step": 1787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870859167145665e-06,
+      "loss": 1.0318,
+      "step": 1788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869473625682506e-06,
+      "loss": 1.0311,
+      "step": 1789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868087343003458e-06,
+      "loss": 0.8424,
+      "step": 1790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866700319374067e-06,
+      "loss": 1.033,
+      "step": 1791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.865312555060024e-06,
+      "loss": 0.5801,
+      "step": 1792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.863924050327166e-06,
+      "loss": 0.7879,
+      "step": 1793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862534805441463e-06,
+      "loss": 0.7934,
+      "step": 1794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861144820669034e-06,
+      "loss": 0.8919,
+      "step": 1795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859754096276137e-06,
+      "loss": 1.1483,
+      "step": 1796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85836263252917e-06,
+      "loss": 1.1119,
+      "step": 1797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856970429694676e-06,
+      "loss": 0.8358,
+      "step": 1798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.855577488039338e-06,
+      "loss": 0.8399,
+      "step": 1799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85418380782998e-06,
+      "loss": 0.569,
+      "step": 1800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852789389333567e-06,
+      "loss": 0.5842,
+      "step": 1801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.851394232817207e-06,
+      "loss": 0.7202,
+      "step": 1802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.849998338548152e-06,
+      "loss": 0.9328,
+      "step": 1803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848601706793786e-06,
+      "loss": 1.2413,
+      "step": 1804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.847204337821647e-06,
+      "loss": 0.7787,
+      "step": 1805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845806231899404e-06,
+      "loss": 0.5537,
+      "step": 1806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84440738929487e-06,
+      "loss": 1.123,
+      "step": 1807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843007810276002e-06,
+      "loss": 1.1566,
+      "step": 1808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841607495110896e-06,
+      "loss": 0.8921,
+      "step": 1809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840206444067787e-06,
+      "loss": 0.9581,
+      "step": 1810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838804657415054e-06,
+      "loss": 1.0035,
+      "step": 1811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837402135421215e-06,
+      "loss": 0.9956,
+      "step": 1812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83599887835493e-06,
+      "loss": 1.2419,
+      "step": 1813
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834594886485001e-06,
+      "loss": 1.0681,
+      "step": 1814
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83319016008037e-06,
+      "loss": 0.8054,
+      "step": 1815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.831784699410116e-06,
+      "loss": 0.7371,
+      "step": 1816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.830378504743461e-06,
+      "loss": 0.7424,
+      "step": 1817
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828971576349771e-06,
+      "loss": 0.904,
+      "step": 1818
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.827563914498551e-06,
+      "loss": 0.7229,
+      "step": 1819
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82615551945944e-06,
+      "loss": 0.6633,
+      "step": 1820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824746391502228e-06,
+      "loss": 0.2481,
+      "step": 1821
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823336530896835e-06,
+      "loss": 0.493,
+      "step": 1822
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82192593791333e-06,
+      "loss": 0.8102,
+      "step": 1823
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820514612821922e-06,
+      "loss": 0.6685,
+      "step": 1824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819102555892948e-06,
+      "loss": 0.3147,
+      "step": 1825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817689767396902e-06,
+      "loss": 0.7868,
+      "step": 1826
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.816276247604407e-06,
+      "loss": 1.0885,
+      "step": 1827
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.814861996786229e-06,
+      "loss": 1.0187,
+      "step": 1828
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.813447015213274e-06,
+      "loss": 0.6907,
+      "step": 1829
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812031303156591e-06,
+      "loss": 0.9299,
+      "step": 1830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.810614860887365e-06,
+      "loss": 0.9025,
+      "step": 1831
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809197688676921e-06,
+      "loss": 0.8601,
+      "step": 1832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807779786796728e-06,
+      "loss": 0.9836,
+      "step": 1833
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806361155518387e-06,
+      "loss": 0.8887,
+      "step": 1834
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80494179511365e-06,
+      "loss": 1.1028,
+      "step": 1835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803521705854395e-06,
+      "loss": 0.8436,
+      "step": 1836
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802100888012652e-06,
+      "loss": 0.8456,
+      "step": 1837
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.800679341860583e-06,
+      "loss": 1.1032,
+      "step": 1838
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799257067670492e-06,
+      "loss": 0.6411,
+      "step": 1839
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.797834065714824e-06,
+      "loss": 0.7731,
+      "step": 1840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796410336266161e-06,
+      "loss": 1.0094,
+      "step": 1841
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.794985879597222e-06,
+      "loss": 0.8841,
+      "step": 1842
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793560695980875e-06,
+      "loss": 0.458,
+      "step": 1843
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792134785690113e-06,
+      "loss": 0.8205,
+      "step": 1844
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79070814899808e-06,
+      "loss": 0.7624,
+      "step": 1845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789280786178056e-06,
+      "loss": 0.6074,
+      "step": 1846
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.787852697503456e-06,
+      "loss": 0.7758,
+      "step": 1847
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786423883247838e-06,
+      "loss": 0.9214,
+      "step": 1848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7849943436849e-06,
+      "loss": 1.0393,
+      "step": 1849
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783564079088478e-06,
+      "loss": 0.3115,
+      "step": 1850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782133089732541e-06,
+      "loss": 0.7303,
+      "step": 1851
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780701375891204e-06,
+      "loss": 0.6605,
+      "step": 1852
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77926893783872e-06,
+      "loss": 0.8498,
+      "step": 1853
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.777835775849477e-06,
+      "loss": 0.9894,
+      "step": 1854
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776401890198006e-06,
+      "loss": 0.9035,
+      "step": 1855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774967281158973e-06,
+      "loss": 0.6797,
+      "step": 1856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.773531949007186e-06,
+      "loss": 0.9123,
+      "step": 1857
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772095894017587e-06,
+      "loss": 0.8831,
+      "step": 1858
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770659116465262e-06,
+      "loss": 0.7089,
+      "step": 1859
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76922161662543e-06,
+      "loss": 0.7902,
+      "step": 1860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.767783394773452e-06,
+      "loss": 0.7846,
+      "step": 1861
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766344451184825e-06,
+      "loss": 0.4916,
+      "step": 1862
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764904786135187e-06,
+      "loss": 0.7969,
+      "step": 1863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763464399900314e-06,
+      "loss": 0.9873,
+      "step": 1864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762023292756114e-06,
+      "loss": 1.0612,
+      "step": 1865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760581464978642e-06,
+      "loss": 0.5267,
+      "step": 1866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759138916844086e-06,
+      "loss": 0.9273,
+      "step": 1867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757695648628772e-06,
+      "loss": 0.9177,
+      "step": 1868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756251660609165e-06,
+      "loss": 1.1813,
+      "step": 1869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75480695306187e-06,
+      "loss": 1.0681,
+      "step": 1870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.753361526263622e-06,
+      "loss": 0.7178,
+      "step": 1871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751915380491306e-06,
+      "loss": 0.6298,
+      "step": 1872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.750468516021932e-06,
+      "loss": 0.9185,
+      "step": 1873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749020933132659e-06,
+      "loss": 0.6769,
+      "step": 1874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747572632100775e-06,
+      "loss": 0.9152,
+      "step": 1875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746123613203709e-06,
+      "loss": 0.94,
+      "step": 1876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74467387671903e-06,
+      "loss": 1.0502,
+      "step": 1877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.743223422924439e-06,
+      "loss": 1.1915,
+      "step": 1878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741772252097778e-06,
+      "loss": 1.0342,
+      "step": 1879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740320364517025e-06,
+      "loss": 0.9296,
+      "step": 1880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.738867760460298e-06,
+      "loss": 0.5966,
+      "step": 1881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737414440205849e-06,
+      "loss": 0.9965,
+      "step": 1882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735960404032066e-06,
+      "loss": 0.7934,
+      "step": 1883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734505652217481e-06,
+      "loss": 0.628,
+      "step": 1884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.733050185040757e-06,
+      "loss": 0.9274,
+      "step": 1885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731594002780694e-06,
+      "loss": 0.4325,
+      "step": 1886
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.730137105716231e-06,
+      "loss": 0.9184,
+      "step": 1887
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728679494126444e-06,
+      "loss": 0.6656,
+      "step": 1888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727221168290545e-06,
+      "loss": 1.0658,
+      "step": 1889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725762128487884e-06,
+      "loss": 0.6138,
+      "step": 1890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724302374997946e-06,
+      "loss": 0.6851,
+      "step": 1891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.722841908100354e-06,
+      "loss": 0.979,
+      "step": 1892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721380728074869e-06,
+      "loss": 1.0072,
+      "step": 1893
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719918835201385e-06,
+      "loss": 0.9417,
+      "step": 1894
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718456229759933e-06,
+      "loss": 0.5407,
+      "step": 1895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716992912030684e-06,
+      "loss": 0.8083,
+      "step": 1896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715528882293944e-06,
+      "loss": 0.6325,
+      "step": 1897
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714064140830153e-06,
+      "loss": 0.5294,
+      "step": 1898
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.712598687919892e-06,
+      "loss": 0.6656,
+      "step": 1899
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71113252384387e-06,
+      "loss": 0.9574,
+      "step": 1900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709665648882943e-06,
+      "loss": 0.9947,
+      "step": 1901
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708198063318097e-06,
+      "loss": 0.5956,
+      "step": 1902
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706729767430451e-06,
+      "loss": 1.145,
+      "step": 1903
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.705260761501268e-06,
+      "loss": 1.1564,
+      "step": 1904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.703791045811943e-06,
+      "loss": 0.5103,
+      "step": 1905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702320620644005e-06,
+      "loss": 0.6525,
+      "step": 1906
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700849486279122e-06,
+      "loss": 0.7976,
+      "step": 1907
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699377642999096e-06,
+      "loss": 0.7333,
+      "step": 1908
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697905091085865e-06,
+      "loss": 0.9148,
+      "step": 1909
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696431830821506e-06,
+      "loss": 0.9622,
+      "step": 1910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694957862488225e-06,
+      "loss": 1.0761,
+      "step": 1911
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69348318636837e-06,
+      "loss": 0.6898,
+      "step": 1912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692007802744422e-06,
+      "loss": 0.7969,
+      "step": 1913
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690531711898997e-06,
+      "loss": 0.6841,
+      "step": 1914
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689054914114848e-06,
+      "loss": 1.189,
+      "step": 1915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687577409674863e-06,
+      "loss": 0.7366,
+      "step": 1916
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.686099198862064e-06,
+      "loss": 0.7595,
+      "step": 1917
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684620281959609e-06,
+      "loss": 1.013,
+      "step": 1918
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68314065925079e-06,
+      "loss": 0.9164,
+      "step": 1919
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68166033101904e-06,
+      "loss": 0.5276,
+      "step": 1920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.680179297547919e-06,
+      "loss": 0.7701,
+      "step": 1921
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678697559121128e-06,
+      "loss": 0.743,
+      "step": 1922
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.677215116022499e-06,
+      "loss": 0.9875,
+      "step": 1923
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675731968536004e-06,
+      "loss": 1.2318,
+      "step": 1924
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674248116945743e-06,
+      "loss": 1.1358,
+      "step": 1925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672763561535957e-06,
+      "loss": 0.8401,
+      "step": 1926
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67127830259102e-06,
+      "loss": 0.8341,
+      "step": 1927
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669792340395441e-06,
+      "loss": 1.0606,
+      "step": 1928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66830567523386e-06,
+      "loss": 1.0671,
+      "step": 1929
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666818307391058e-06,
+      "loss": 0.9874,
+      "step": 1930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665330237151946e-06,
+      "loss": 1.2755,
+      "step": 1931
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66384146480157e-06,
+      "loss": 0.5684,
+      "step": 1932
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.662351990625113e-06,
+      "loss": 0.8272,
+      "step": 1933
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660861814907888e-06,
+      "loss": 1.1468,
+      "step": 1934
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65937093793535e-06,
+      "loss": 0.8785,
+      "step": 1935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.657879359993078e-06,
+      "loss": 0.7846,
+      "step": 1936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656387081366793e-06,
+      "loss": 0.8145,
+      "step": 1937
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654894102342349e-06,
+      "loss": 0.7153,
+      "step": 1938
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653400423205734e-06,
+      "loss": 0.6358,
+      "step": 1939
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651906044243066e-06,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650410965740602e-06,
+      "loss": 0.4982,
+      "step": 1941
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.648915187984733e-06,
+      "loss": 0.9667,
+      "step": 1942
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647418711261977e-06,
+      "loss": 0.7527,
+      "step": 1943
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645921535859e-06,
+      "loss": 0.7239,
+      "step": 1944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644423662062584e-06,
+      "loss": 1.0603,
+      "step": 1945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642925090159659e-06,
+      "loss": 0.781,
+      "step": 1946
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.641425820437282e-06,
+      "loss": 0.7066,
+      "step": 1947
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639925853182645e-06,
+      "loss": 1.0401,
+      "step": 1948
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638425188683072e-06,
+      "loss": 0.8562,
+      "step": 1949
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636923827226029e-06,
+      "loss": 0.7312,
+      "step": 1950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.635421769099102e-06,
+      "loss": 0.9672,
+      "step": 1951
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633919014590018e-06,
+      "loss": 0.9975,
+      "step": 1952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63241556398664e-06,
+      "loss": 0.5192,
+      "step": 1953
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630911417576959e-06,
+      "loss": 0.8793,
+      "step": 1954
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.629406575649102e-06,
+      "loss": 0.7625,
+      "step": 1955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627901038491327e-06,
+      "loss": 0.4613,
+      "step": 1956
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626394806392028e-06,
+      "loss": 0.7504,
+      "step": 1957
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624887879639733e-06,
+      "loss": 0.9148,
+      "step": 1958
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.623380258523097e-06,
+      "loss": 0.789,
+      "step": 1959
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.621871943330914e-06,
+      "loss": 0.6926,
+      "step": 1960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.620362934352109e-06,
+      "loss": 0.8566,
+      "step": 1961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.618853231875738e-06,
+      "loss": 0.4178,
+      "step": 1962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617342836190995e-06,
+      "loss": 0.7879,
+      "step": 1963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615831747587198e-06,
+      "loss": 1.0364,
+      "step": 1964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.614319966353808e-06,
+      "loss": 0.9183,
+      "step": 1965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612807492780413e-06,
+      "loss": 0.6997,
+      "step": 1966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611294327156734e-06,
+      "loss": 0.6935,
+      "step": 1967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.609780469772623e-06,
+      "loss": 0.506,
+      "step": 1968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608265920918069e-06,
+      "loss": 0.9806,
+      "step": 1969
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60675068088319e-06,
+      "loss": 0.807,
+      "step": 1970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.605234749958238e-06,
+      "loss": 0.9238,
+      "step": 1971
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603718128433596e-06,
+      "loss": 0.5932,
+      "step": 1972
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.602200816599779e-06,
+      "loss": 0.6864,
+      "step": 1973
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600682814747439e-06,
+      "loss": 0.9071,
+      "step": 1974
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599164123167351e-06,
+      "loss": 0.783,
+      "step": 1975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.597644742150433e-06,
+      "loss": 1.0403,
+      "step": 1976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596124671987725e-06,
+      "loss": 0.9878,
+      "step": 1977
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.594603912970405e-06,
+      "loss": 0.8942,
+      "step": 1978
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593082465389784e-06,
+      "loss": 0.8429,
+      "step": 1979
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5915603295373e-06,
+      "loss": 0.6074,
+      "step": 1980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590037505704526e-06,
+      "loss": 0.3539,
+      "step": 1981
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588513994183165e-06,
+      "loss": 0.7216,
+      "step": 1982
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586989795265056e-06,
+      "loss": 0.8451,
+      "step": 1983
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58546490924216e-06,
+      "loss": 0.9245,
+      "step": 1984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.583939336406583e-06,
+      "loss": 0.6793,
+      "step": 1985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.582413077050554e-06,
+      "loss": 0.904,
+      "step": 1986
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580886131466431e-06,
+      "loss": 0.5624,
+      "step": 1987
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579358499946713e-06,
+      "loss": 0.705,
+      "step": 1988
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.577830182784021e-06,
+      "loss": 0.7365,
+      "step": 1989
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576301180271116e-06,
+      "loss": 0.8717,
+      "step": 1990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.57477149270088e-06,
+      "loss": 0.8954,
+      "step": 1991
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573241120366336e-06,
+      "loss": 1.2035,
+      "step": 1992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571710063560633e-06,
+      "loss": 0.8126,
+      "step": 1993
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.57017832257705e-06,
+      "loss": 0.8723,
+      "step": 1994
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.568645897709002e-06,
+      "loss": 0.7384,
+      "step": 1995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56711278925003e-06,
+      "loss": 1.4421,
+      "step": 1996
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565578997493811e-06,
+      "loss": 0.6832,
+      "step": 1997
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564044522734147e-06,
+      "loss": 0.8771,
+      "step": 1998
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562509365264975e-06,
+      "loss": 0.8187,
+      "step": 1999
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.560973525380362e-06,
+      "loss": 1.1009,
+      "step": 2000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559437003374507e-06,
+      "loss": 0.8208,
+      "step": 2001
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.557899799541735e-06,
+      "loss": 0.8347,
+      "step": 2002
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556361914176505e-06,
+      "loss": 1.0774,
+      "step": 2003
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554823347573408e-06,
+      "loss": 0.903,
+      "step": 2004
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.553284100027162e-06,
+      "loss": 0.7711,
+      "step": 2005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551744171832617e-06,
+      "loss": 1.1452,
+      "step": 2006
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.550203563284757e-06,
+      "loss": 1.0167,
+      "step": 2007
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548662274678688e-06,
+      "loss": 0.925,
+      "step": 2008
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547120306309653e-06,
+      "loss": 0.9169,
+      "step": 2009
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545577658473024e-06,
+      "loss": 0.9504,
+      "step": 2010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544034331464302e-06,
+      "loss": 0.9478,
+      "step": 2011
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542490325579118e-06,
+      "loss": 0.732,
+      "step": 2012
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540945641113234e-06,
+      "loss": 0.7891,
+      "step": 2013
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539400278362545e-06,
+      "loss": 0.9469,
+      "step": 2014
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537854237623065e-06,
+      "loss": 0.9101,
+      "step": 2015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536307519190951e-06,
+      "loss": 1.1349,
+      "step": 2016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.534760123362482e-06,
+      "loss": 0.6695,
+      "step": 2017
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53321205043407e-06,
+      "loss": 0.5826,
+      "step": 2018
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.531663300702258e-06,
+      "loss": 0.8136,
+      "step": 2019
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53011387446371e-06,
+      "loss": 0.8759,
+      "step": 2020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528563772015232e-06,
+      "loss": 0.7768,
+      "step": 2021
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52701299365375e-06,
+      "loss": 0.5129,
+      "step": 2022
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.525461539676324e-06,
+      "loss": 0.827,
+      "step": 2023
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523909410380142e-06,
+      "loss": 0.6983,
+      "step": 2024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522356606062522e-06,
+      "loss": 0.7476,
+      "step": 2025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520803127020913e-06,
+      "loss": 1.1908,
+      "step": 2026
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.519248973552887e-06,
+      "loss": 0.7835,
+      "step": 2027
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.517694145956152e-06,
+      "loss": 0.8199,
+      "step": 2028
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51613864452854e-06,
+      "loss": 0.4771,
+      "step": 2029
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514582469568019e-06,
+      "loss": 0.7632,
+      "step": 2030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513025621372679e-06,
+      "loss": 0.9136,
+      "step": 2031
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511468100240742e-06,
+      "loss": 0.8113,
+      "step": 2032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509909906470556e-06,
+      "loss": 1.0009,
+      "step": 2033
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.508351040360607e-06,
+      "loss": 0.7482,
+      "step": 2034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506791502209497e-06,
+      "loss": 0.7726,
+      "step": 2035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505231292315964e-06,
+      "loss": 0.4674,
+      "step": 2036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.503670410978874e-06,
+      "loss": 1.0328,
+      "step": 2037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.502108858497223e-06,
+      "loss": 1.0995,
+      "step": 2038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.500546635170133e-06,
+      "loss": 0.7885,
+      "step": 2039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.498983741296852e-06,
+      "loss": 1.1231,
+      "step": 2040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497420177176762e-06,
+      "loss": 0.6228,
+      "step": 2041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.495855943109375e-06,
+      "loss": 1.0658,
+      "step": 2042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49429103939432e-06,
+      "loss": 0.549,
+      "step": 2043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492725466331366e-06,
+      "loss": 1.0008,
+      "step": 2044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491159224220406e-06,
+      "loss": 1.0255,
+      "step": 2045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489592313361459e-06,
+      "loss": 1.0351,
+      "step": 2046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.488024734054676e-06,
+      "loss": 0.7672,
+      "step": 2047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.486456486600334e-06,
+      "loss": 0.4347,
+      "step": 2048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484887571298837e-06,
+      "loss": 0.7015,
+      "step": 2049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48331798845072e-06,
+      "loss": 0.9083,
+      "step": 2050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.481747738356642e-06,
+      "loss": 0.6285,
+      "step": 2051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480176821317394e-06,
+      "loss": 1.111,
+      "step": 2052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47860523763389e-06,
+      "loss": 0.7889,
+      "step": 2053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.477032987607175e-06,
+      "loss": 0.4105,
+      "step": 2054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475460071538422e-06,
+      "loss": 0.7645,
+      "step": 2055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.473886489728931e-06,
+      "loss": 0.7447,
+      "step": 2056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472312242480125e-06,
+      "loss": 0.7064,
+      "step": 2057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470737330093564e-06,
+      "loss": 0.7379,
+      "step": 2058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469161752870927e-06,
+      "loss": 1.0724,
+      "step": 2059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467585511114021e-06,
+      "loss": 0.6955,
+      "step": 2060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466008605124786e-06,
+      "loss": 0.5268,
+      "step": 2061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464431035205288e-06,
+      "loss": 0.7145,
+      "step": 2062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462852801657711e-06,
+      "loss": 0.8318,
+      "step": 2063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461273904784378e-06,
+      "loss": 1.1635,
+      "step": 2064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459694344887732e-06,
+      "loss": 0.9595,
+      "step": 2065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.458114122270347e-06,
+      "loss": 0.8355,
+      "step": 2066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45653323723492e-06,
+      "loss": 1.1903,
+      "step": 2067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454951690084278e-06,
+      "loss": 0.4495,
+      "step": 2068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.453369481121376e-06,
+      "loss": 0.8319,
+      "step": 2069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45178661064929e-06,
+      "loss": 0.5441,
+      "step": 2070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.450203078971227e-06,
+      "loss": 0.8374,
+      "step": 2071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448618886390523e-06,
+      "loss": 0.7022,
+      "step": 2072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447034033210632e-06,
+      "loss": 0.548,
+      "step": 2073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445448519735144e-06,
+      "loss": 0.8999,
+      "step": 2074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443862346267772e-06,
+      "loss": 0.3431,
+      "step": 2075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442275513112355e-06,
+      "loss": 1.2767,
+      "step": 2076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440688020572856e-06,
+      "loss": 0.2273,
+      "step": 2077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439099868953368e-06,
+      "loss": 0.5549,
+      "step": 2078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.437511058558109e-06,
+      "loss": 0.9834,
+      "step": 2079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.435921589691423e-06,
+      "loss": 0.7696,
+      "step": 2080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43433146265778e-06,
+      "loss": 0.8885,
+      "step": 2081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.432740677761776e-06,
+      "loss": 0.9248,
+      "step": 2082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431149235308133e-06,
+      "loss": 0.7424,
+      "step": 2083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.429557135601703e-06,
+      "loss": 0.7338,
+      "step": 2084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.427964378947454e-06,
+      "loss": 0.7574,
+      "step": 2085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426370965650492e-06,
+      "loss": 1.0198,
+      "step": 2086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42477689601604e-06,
+      "loss": 0.869,
+      "step": 2087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423182170349449e-06,
+      "loss": 0.9349,
+      "step": 2088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.421586788956197e-06,
+      "loss": 0.8519,
+      "step": 2089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.419990752141886e-06,
+      "loss": 0.5729,
+      "step": 2090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418394060212244e-06,
+      "loss": 0.5967,
+      "step": 2091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41679671347313e-06,
+      "loss": 0.8319,
+      "step": 2092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415198712230517e-06,
+      "loss": 0.949,
+      "step": 2093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.413600056790512e-06,
+      "loss": 0.7523,
+      "step": 2094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.412000747459344e-06,
+      "loss": 0.8491,
+      "step": 2095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410400784543371e-06,
+      "loss": 1.1791,
+      "step": 2096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.408800168349069e-06,
+      "loss": 1.127,
+      "step": 2097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407198899183049e-06,
+      "loss": 0.9381,
+      "step": 2098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.405596977352038e-06,
+      "loss": 0.7392,
+      "step": 2099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40399440316289e-06,
+      "loss": 0.7171,
+      "step": 2100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40239117692259e-06,
+      "loss": 0.9815,
+      "step": 2101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.400787298938243e-06,
+      "loss": 0.3836,
+      "step": 2102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.399182769517076e-06,
+      "loss": 1.1091,
+      "step": 2103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.397577588966446e-06,
+      "loss": 1.2457,
+      "step": 2104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395971757593835e-06,
+      "loss": 0.7619,
+      "step": 2105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.394365275706844e-06,
+      "loss": 1.0751,
+      "step": 2106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392758143613205e-06,
+      "loss": 0.8815,
+      "step": 2107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39115036162077e-06,
+      "loss": 0.4748,
+      "step": 2108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.389541930037516e-06,
+      "loss": 0.6274,
+      "step": 2109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38793284917155e-06,
+      "loss": 0.4229,
+      "step": 2110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386323119331095e-06,
+      "loss": 0.8992,
+      "step": 2111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.384712740824503e-06,
+      "loss": 0.8845,
+      "step": 2112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383101713960252e-06,
+      "loss": 1.0092,
+      "step": 2113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381490039046938e-06,
+      "loss": 0.7742,
+      "step": 2114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379877716393288e-06,
+      "loss": 0.8869,
+      "step": 2115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37826474630815e-06,
+      "loss": 0.8777,
+      "step": 2116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.376651129100494e-06,
+      "loss": 0.7964,
+      "step": 2117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.375036865079418e-06,
+      "loss": 0.9069,
+      "step": 2118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37342195455414e-06,
+      "loss": 0.7089,
+      "step": 2119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371806397834004e-06,
+      "loss": 0.6583,
+      "step": 2120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370190195228478e-06,
+      "loss": 1.2088,
+      "step": 2121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.368573347047153e-06,
+      "loss": 1.0329,
+      "step": 2122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366955853599745e-06,
+      "loss": 0.863,
+      "step": 2123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365337715196091e-06,
+      "loss": 0.9037,
+      "step": 2124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.363718932146155e-06,
+      "loss": 1.0681,
+      "step": 2125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.362099504760021e-06,
+      "loss": 0.7484,
+      "step": 2126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360479433347897e-06,
+      "loss": 0.8573,
+      "step": 2127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.358858718220116e-06,
+      "loss": 0.836,
+      "step": 2128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.357237359687135e-06,
+      "loss": 0.4995,
+      "step": 2129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355615358059532e-06,
+      "loss": 1.0181,
+      "step": 2130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353992713648007e-06,
+      "loss": 1.0689,
+      "step": 2131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35236942676339e-06,
+      "loss": 0.3828,
+      "step": 2132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.350745497716625e-06,
+      "loss": 0.8258,
+      "step": 2133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349120926818786e-06,
+      "loss": 0.7844,
+      "step": 2134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347495714381064e-06,
+      "loss": 0.5751,
+      "step": 2135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345869860714778e-06,
+      "loss": 0.8445,
+      "step": 2136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344243366131369e-06,
+      "loss": 0.6601,
+      "step": 2137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342616230942397e-06,
+      "loss": 0.879,
+      "step": 2138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.34098845545955e-06,
+      "loss": 0.8336,
+      "step": 2139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339360039994635e-06,
+      "loss": 0.818,
+      "step": 2140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337730984859582e-06,
+      "loss": 0.5758,
+      "step": 2141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.336101290366446e-06,
+      "loss": 0.6999,
+      "step": 2142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3344709568274e-06,
+      "loss": 0.6949,
+      "step": 2143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332839984554746e-06,
+      "loss": 0.8577,
+      "step": 2144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3312083738609e-06,
+      "loss": 0.9682,
+      "step": 2145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329576125058406e-06,
+      "loss": 0.8885,
+      "step": 2146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32794323845993e-06,
+      "loss": 0.7805,
+      "step": 2147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32630971437826e-06,
+      "loss": 0.6236,
+      "step": 2148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324675553126304e-06,
+      "loss": 0.3304,
+      "step": 2149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323040755017091e-06,
+      "loss": 1.034,
+      "step": 2150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.321405320363778e-06,
+      "loss": 0.4636,
+      "step": 2151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31976924947964e-06,
+      "loss": 0.5898,
+      "step": 2152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31813254267807e-06,
+      "loss": 0.8331,
+      "step": 2153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.316495200272592e-06,
+      "loss": 1.0163,
+      "step": 2154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314857222576845e-06,
+      "loss": 0.9853,
+      "step": 2155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.313218609904587e-06,
+      "loss": 1.0725,
+      "step": 2156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.311579362569708e-06,
+      "loss": 0.4731,
+      "step": 2157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309939480886212e-06,
+      "loss": 0.9183,
+      "step": 2158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308298965168223e-06,
+      "loss": 0.8417,
+      "step": 2159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306657815729995e-06,
+      "loss": 0.602,
+      "step": 2160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305016032885891e-06,
+      "loss": 0.4026,
+      "step": 2161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.303373616950408e-06,
+      "loss": 1.0711,
+      "step": 2162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301730568238156e-06,
+      "loss": 0.9748,
+      "step": 2163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.300086887063869e-06,
+      "loss": 0.8395,
+      "step": 2164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.298442573742402e-06,
+      "loss": 0.787,
+      "step": 2165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29679762858873e-06,
+      "loss": 1.1305,
+      "step": 2166
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.295152051917953e-06,
+      "loss": 0.6868,
+      "step": 2167
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.293505844045286e-06,
+      "loss": 1.076,
+      "step": 2168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.291859005286069e-06,
+      "loss": 0.8313,
+      "step": 2169
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290211535955761e-06,
+      "loss": 0.4046,
+      "step": 2170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288563436369944e-06,
+      "loss": 0.7004,
+      "step": 2171
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.286914706844316e-06,
+      "loss": 0.7259,
+      "step": 2172
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.285265347694702e-06,
+      "loss": 0.9831,
+      "step": 2173
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283615359237045e-06,
+      "loss": 1.0707,
+      "step": 2174
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.281964741787406e-06,
+      "loss": 0.8811,
+      "step": 2175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28031349566197e-06,
+      "loss": 0.7516,
+      "step": 2176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278661621177037e-06,
+      "loss": 0.8307,
+      "step": 2177
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.277009118649038e-06,
+      "loss": 0.7568,
+      "step": 2178
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.275355988394512e-06,
+      "loss": 0.9605,
+      "step": 2179
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.273702230730127e-06,
+      "loss": 0.6117,
+      "step": 2180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.272047845972667e-06,
+      "loss": 0.6654,
+      "step": 2181
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270392834439039e-06,
+      "loss": 0.3984,
+      "step": 2182
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.268737196446264e-06,
+      "loss": 0.8416,
+      "step": 2183
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.267080932311489e-06,
+      "loss": 0.9565,
+      "step": 2184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.265424042351982e-06,
+      "loss": 1.0398,
+      "step": 2185
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263766526885126e-06,
+      "loss": 1.0539,
+      "step": 2186
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.262108386228425e-06,
+      "loss": 0.8038,
+      "step": 2187
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.260449620699507e-06,
+      "loss": 1.0939,
+      "step": 2188
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.258790230616111e-06,
+      "loss": 0.9254,
+      "step": 2189
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.257130216296105e-06,
+      "loss": 1.0085,
+      "step": 2190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255469578057473e-06,
+      "loss": 0.5322,
+      "step": 2191
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253808316218315e-06,
+      "loss": 1.2291,
+      "step": 2192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.252146431096857e-06,
+      "loss": 1.0489,
+      "step": 2193
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250483923011437e-06,
+      "loss": 1.1284,
+      "step": 2194
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.248820792280518e-06,
+      "loss": 0.5993,
+      "step": 2195
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.247157039222683e-06,
+      "loss": 0.9624,
+      "step": 2196
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.245492664156627e-06,
+      "loss": 0.6237,
+      "step": 2197
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.243827667401174e-06,
+      "loss": 0.7984,
+      "step": 2198
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.242162049275257e-06,
+      "loss": 0.9218,
+      "step": 2199
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240495810097936e-06,
+      "loss": 0.5879,
+      "step": 2200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.238828950188387e-06,
+      "loss": 0.882,
+      "step": 2201
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237161469865904e-06,
+      "loss": 0.6583,
+      "step": 2202
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2354933694499e-06,
+      "loss": 1.0088,
+      "step": 2203
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23382464925991e-06,
+      "loss": 0.8312,
+      "step": 2204
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.232155309615583e-06,
+      "loss": 0.6544,
+      "step": 2205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.230485350836689e-06,
+      "loss": 0.465,
+      "step": 2206
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.228814773243116e-06,
+      "loss": 0.6651,
+      "step": 2207
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.227143577154874e-06,
+      "loss": 0.8406,
+      "step": 2208
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.225471762892086e-06,
+      "loss": 0.6695,
+      "step": 2209
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.223799330774995e-06,
+      "loss": 1.0433,
+      "step": 2210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.222126281123965e-06,
+      "loss": 1.0129,
+      "step": 2211
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.220452614259478e-06,
+      "loss": 0.2953,
+      "step": 2212
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218778330502129e-06,
+      "loss": 0.9371,
+      "step": 2213
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.217103430172638e-06,
+      "loss": 0.381,
+      "step": 2214
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.215427913591836e-06,
+      "loss": 1.0188,
+      "step": 2215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21375178108068e-06,
+      "loss": 1.0341,
+      "step": 2216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212075032960243e-06,
+      "loss": 0.5567,
+      "step": 2217
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.210397669551708e-06,
+      "loss": 1.1085,
+      "step": 2218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208719691176386e-06,
+      "loss": 0.8696,
+      "step": 2219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.207041098155701e-06,
+      "loss": 0.7773,
+      "step": 2220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205361890811194e-06,
+      "loss": 0.3637,
+      "step": 2221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.203682069464526e-06,
+      "loss": 0.7809,
+      "step": 2222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.202001634437474e-06,
+      "loss": 0.779,
+      "step": 2223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200320586051935e-06,
+      "loss": 0.8016,
+      "step": 2224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19863892462992e-06,
+      "loss": 0.9878,
+      "step": 2225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.196956650493562e-06,
+      "loss": 0.9163,
+      "step": 2226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.195273763965104e-06,
+      "loss": 0.7182,
+      "step": 2227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.193590265366913e-06,
+      "loss": 0.8046,
+      "step": 2228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191906155021471e-06,
+      "loss": 0.7953,
+      "step": 2229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19022143325138e-06,
+      "loss": 0.9184,
+      "step": 2230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.188536100379352e-06,
+      "loss": 0.9181,
+      "step": 2231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.186850156728223e-06,
+      "loss": 1.0212,
+      "step": 2232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185163602620943e-06,
+      "loss": 0.4653,
+      "step": 2233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.183476438380577e-06,
+      "loss": 0.7665,
+      "step": 2234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.181788664330315e-06,
+      "loss": 0.9342,
+      "step": 2235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.180100280793451e-06,
+      "loss": 0.6376,
+      "step": 2236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178411288093408e-06,
+      "loss": 0.7534,
+      "step": 2237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176721686553719e-06,
+      "loss": 1.0126,
+      "step": 2238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175031476498034e-06,
+      "loss": 0.8911,
+      "step": 2239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.173340658250121e-06,
+      "loss": 0.8443,
+      "step": 2240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.171649232133867e-06,
+      "loss": 0.757,
+      "step": 2241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.169957198473271e-06,
+      "loss": 1.0471,
+      "step": 2242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168264557592447e-06,
+      "loss": 0.8459,
+      "step": 2243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.166571309815632e-06,
+      "loss": 0.6674,
+      "step": 2244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.164877455467176e-06,
+      "loss": 0.8015,
+      "step": 2245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163182994871541e-06,
+      "loss": 0.7476,
+      "step": 2246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161487928353312e-06,
+      "loss": 0.9088,
+      "step": 2247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.159792256237185e-06,
+      "loss": 1.2052,
+      "step": 2248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.158095978847976e-06,
+      "loss": 0.8181,
+      "step": 2249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156399096510611e-06,
+      "loss": 0.3192,
+      "step": 2250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.154701609550143e-06,
+      "loss": 0.6443,
+      "step": 2251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.153003518291728e-06,
+      "loss": 0.8798,
+      "step": 2252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151304823060645e-06,
+      "loss": 0.8524,
+      "step": 2253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.149605524182287e-06,
+      "loss": 0.8663,
+      "step": 2254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14790562198216e-06,
+      "loss": 1.054,
+      "step": 2255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.146205116785893e-06,
+      "loss": 1.2488,
+      "step": 2256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.144504008919224e-06,
+      "loss": 0.7344,
+      "step": 2257
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.142802298708005e-06,
+      "loss": 0.5904,
+      "step": 2258
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141099986478212e-06,
+      "loss": 1.0461,
+      "step": 2259
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139397072555928e-06,
+      "loss": 0.9949,
+      "step": 2260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137693557267354e-06,
+      "loss": 0.7774,
+      "step": 2261
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.135989440938808e-06,
+      "loss": 0.8871,
+      "step": 2262
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.134284723896719e-06,
+      "loss": 0.9716,
+      "step": 2263
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.132579406467635e-06,
+      "loss": 1.1758,
+      "step": 2264
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.130873488978217e-06,
+      "loss": 0.6541,
+      "step": 2265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.129166971755244e-06,
+      "loss": 0.6434,
+      "step": 2266
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.127459855125604e-06,
+      "loss": 0.8837,
+      "step": 2267
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125752139416303e-06,
+      "loss": 1.1183,
+      "step": 2268
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.124043824954466e-06,
+      "loss": 0.723,
+      "step": 2269
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.122334912067327e-06,
+      "loss": 0.7597,
+      "step": 2270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.120625401082233e-06,
+      "loss": 1.1584,
+      "step": 2271
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.118915292326653e-06,
+      "loss": 0.97,
+      "step": 2272
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117204586128165e-06,
+      "loss": 0.7012,
+      "step": 2273
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.115493282814463e-06,
+      "loss": 0.5896,
+      "step": 2274
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.113781382713355e-06,
+      "loss": 0.9313,
+      "step": 2275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112068886152763e-06,
+      "loss": 0.9335,
+      "step": 2276
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110355793460725e-06,
+      "loss": 1.306,
+      "step": 2277
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.108642104965389e-06,
+      "loss": 1.0178,
+      "step": 2278
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106927820995025e-06,
+      "loss": 0.8885,
+      "step": 2279
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.105212941878009e-06,
+      "loss": 0.8237,
+      "step": 2280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103497467942833e-06,
+      "loss": 0.9274,
+      "step": 2281
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101781399518106e-06,
+      "loss": 1.0978,
+      "step": 2282
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10006473693255e-06,
+      "loss": 1.015,
+      "step": 2283
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098347480514998e-06,
+      "loss": 1.1044,
+      "step": 2284
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0966296305944e-06,
+      "loss": 0.6117,
+      "step": 2285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.094911187499816e-06,
+      "loss": 0.7735,
+      "step": 2286
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093192151560426e-06,
+      "loss": 0.649,
+      "step": 2287
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091472523105512e-06,
+      "loss": 0.857,
+      "step": 2288
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.089752302464485e-06,
+      "loss": 0.9738,
+      "step": 2289
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.088031489966856e-06,
+      "loss": 1.033,
+      "step": 2290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086310085942258e-06,
+      "loss": 0.9015,
+      "step": 2291
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.084588090720433e-06,
+      "loss": 1.0897,
+      "step": 2292
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082865504631236e-06,
+      "loss": 0.7892,
+      "step": 2293
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081142328004638e-06,
+      "loss": 0.9493,
+      "step": 2294
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07941856117072e-06,
+      "loss": 1.14,
+      "step": 2295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.077694204459678e-06,
+      "loss": 0.7973,
+      "step": 2296
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.075969258201823e-06,
+      "loss": 0.761,
+      "step": 2297
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.074243722727575e-06,
+      "loss": 0.9044,
+      "step": 2298
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072517598367468e-06,
+      "loss": 0.8064,
+      "step": 2299
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.070790885452151e-06,
+      "loss": 0.6171,
+      "step": 2300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.069063584312381e-06,
+      "loss": 0.7812,
+      "step": 2301
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.067335695279033e-06,
+      "loss": 0.6675,
+      "step": 2302
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06560721868309e-06,
+      "loss": 0.9727,
+      "step": 2303
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063878154855653e-06,
+      "loss": 1.2007,
+      "step": 2304
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.062148504127932e-06,
+      "loss": 0.9678,
+      "step": 2305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060418266831244e-06,
+      "loss": 0.8546,
+      "step": 2306
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.058687443297033e-06,
+      "loss": 1.1094,
+      "step": 2307
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05695603385684e-06,
+      "loss": 0.849,
+      "step": 2308
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055224038842327e-06,
+      "loss": 0.8465,
+      "step": 2309
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.053491458585266e-06,
+      "loss": 0.5216,
+      "step": 2310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.051758293417539e-06,
+      "loss": 0.5565,
+      "step": 2311
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050024543671143e-06,
+      "loss": 0.4998,
+      "step": 2312
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.048290209678186e-06,
+      "loss": 0.8697,
+      "step": 2313
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04655529177089e-06,
+      "loss": 0.7953,
+      "step": 2314
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.044819790281584e-06,
+      "loss": 0.8423,
+      "step": 2315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.043083705542712e-06,
+      "loss": 0.8502,
+      "step": 2316
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.041347037886829e-06,
+      "loss": 0.9558,
+      "step": 2317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.039609787646601e-06,
+      "loss": 0.6379,
+      "step": 2318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03787195515481e-06,
+      "loss": 0.8809,
+      "step": 2319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.036133540744342e-06,
+      "loss": 1.0275,
+      "step": 2320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0343945447482e-06,
+      "loss": 0.8588,
+      "step": 2321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0326549674995e-06,
+      "loss": 0.8277,
+      "step": 2322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.030914809331461e-06,
+      "loss": 0.9344,
+      "step": 2323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02917407057742e-06,
+      "loss": 0.9572,
+      "step": 2324
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027432751570824e-06,
+      "loss": 0.76,
+      "step": 2325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.025690852645234e-06,
+      "loss": 1.2334,
+      "step": 2326
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.023948374134313e-06,
+      "loss": 0.6002,
+      "step": 2327
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.022205316371847e-06,
+      "loss": 0.7717,
+      "step": 2328
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.020461679691725e-06,
+      "loss": 0.8959,
+      "step": 2329
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.018717464427945e-06,
+      "loss": 0.7609,
+      "step": 2330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016972670914624e-06,
+      "loss": 0.7195,
+      "step": 2331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.015227299485984e-06,
+      "loss": 0.6346,
+      "step": 2332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013481350476357e-06,
+      "loss": 1.0307,
+      "step": 2333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.011734824220193e-06,
+      "loss": 0.4814,
+      "step": 2334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.009987721052042e-06,
+      "loss": 0.4872,
+      "step": 2335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.008240041306575e-06,
+      "loss": 0.9093,
+      "step": 2336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006491785318563e-06,
+      "loss": 1.0962,
+      "step": 2337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.004742953422894e-06,
+      "loss": 0.8188,
+      "step": 2338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.002993545954566e-06,
+      "loss": 0.8071,
+      "step": 2339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.001243563248688e-06,
+      "loss": 1.0902,
+      "step": 2340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999493005640475e-06,
+      "loss": 0.7293,
+      "step": 2341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.997741873465254e-06,
+      "loss": 0.4376,
+      "step": 2342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.995990167058465e-06,
+      "loss": 1.0516,
+      "step": 2343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.994237886755655e-06,
+      "loss": 1.3369,
+      "step": 2344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99248503289248e-06,
+      "loss": 0.9476,
+      "step": 2345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990731605804709e-06,
+      "loss": 0.8671,
+      "step": 2346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988977605828218e-06,
+      "loss": 0.9275,
+      "step": 2347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987223033298995e-06,
+      "loss": 0.4565,
+      "step": 2348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.985467888553135e-06,
+      "loss": 0.6864,
+      "step": 2349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.983712171926847e-06,
+      "loss": 0.5281,
+      "step": 2350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981955883756443e-06,
+      "loss": 0.7738,
+      "step": 2351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980199024378353e-06,
+      "loss": 0.8605,
+      "step": 2352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.978441594129106e-06,
+      "loss": 0.8393,
+      "step": 2353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976683593345351e-06,
+      "loss": 0.8693,
+      "step": 2354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97492502236384e-06,
+      "loss": 1.086,
+      "step": 2355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.973165881521435e-06,
+      "loss": 0.5593,
+      "step": 2356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.971406171155105e-06,
+      "loss": 0.863,
+      "step": 2357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.969645891601936e-06,
+      "loss": 0.8428,
+      "step": 2358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.967885043199115e-06,
+      "loss": 0.9084,
+      "step": 2359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966123626283941e-06,
+      "loss": 0.8226,
+      "step": 2360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.964361641193824e-06,
+      "loss": 0.9244,
+      "step": 2361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.962599088266276e-06,
+      "loss": 0.9465,
+      "step": 2362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960835967838925e-06,
+      "loss": 0.99,
+      "step": 2363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.959072280249505e-06,
+      "loss": 0.65,
+      "step": 2364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95730802583586e-06,
+      "loss": 1.0677,
+      "step": 2365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955543204935937e-06,
+      "loss": 0.8741,
+      "step": 2366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9537778178878e-06,
+      "loss": 0.9691,
+      "step": 2367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.952011865029614e-06,
+      "loss": 0.5407,
+      "step": 2368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.950245346699659e-06,
+      "loss": 1.0827,
+      "step": 2369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.948478263236317e-06,
+      "loss": 0.7674,
+      "step": 2370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946710614978082e-06,
+      "loss": 1.1371,
+      "step": 2371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944942402263557e-06,
+      "loss": 0.8323,
+      "step": 2372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943173625431449e-06,
+      "loss": 0.528,
+      "step": 2373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941404284820577e-06,
+      "loss": 0.9299,
+      "step": 2374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.939634380769866e-06,
+      "loss": 0.6512,
+      "step": 2375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93786391361835e-06,
+      "loss": 0.8349,
+      "step": 2376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93609288370517e-06,
+      "loss": 0.878,
+      "step": 2377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.934321291369573e-06,
+      "loss": 0.4977,
+      "step": 2378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93254913695092e-06,
+      "loss": 1.2494,
+      "step": 2379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.930776420788673e-06,
+      "loss": 1.0891,
+      "step": 2380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.929003143222404e-06,
+      "loss": 0.9781,
+      "step": 2381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927229304591793e-06,
+      "loss": 0.7094,
+      "step": 2382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.925454905236626e-06,
+      "loss": 0.4964,
+      "step": 2383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923679945496803e-06,
+      "loss": 0.8579,
+      "step": 2384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.921904425712319e-06,
+      "loss": 1.1035,
+      "step": 2385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920128346223286e-06,
+      "loss": 0.6689,
+      "step": 2386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91835170736992e-06,
+      "loss": 0.7377,
+      "step": 2387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.916574509492546e-06,
+      "loss": 0.9844,
+      "step": 2388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.914796752931593e-06,
+      "loss": 0.7904,
+      "step": 2389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9130184380276e-06,
+      "loss": 0.4243,
+      "step": 2390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911239565121209e-06,
+      "loss": 0.886,
+      "step": 2391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.909460134553176e-06,
+      "loss": 0.8183,
+      "step": 2392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907680146664354e-06,
+      "loss": 0.971,
+      "step": 2393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90589960179571e-06,
+      "loss": 1.1936,
+      "step": 2394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.904118500288318e-06,
+      "loss": 0.4702,
+      "step": 2395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902336842483353e-06,
+      "loss": 0.9111,
+      "step": 2396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.900554628722104e-06,
+      "loss": 0.6952,
+      "step": 2397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.898771859345959e-06,
+      "loss": 0.534,
+      "step": 2398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.896988534696415e-06,
+      "loss": 1.1249,
+      "step": 2399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.895204655115081e-06,
+      "loss": 0.749,
+      "step": 2400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.893420220943663e-06,
+      "loss": 0.6978,
+      "step": 2401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89163523252398e-06,
+      "loss": 0.3585,
+      "step": 2402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.889849690197956e-06,
+      "loss": 0.5773,
+      "step": 2403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888063594307615e-06,
+      "loss": 0.5587,
+      "step": 2404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.886276945195098e-06,
+      "loss": 0.8101,
+      "step": 2405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884489743202643e-06,
+      "loss": 0.8086,
+      "step": 2406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.882701988672598e-06,
+      "loss": 0.9297,
+      "step": 2407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.880913681947413e-06,
+      "loss": 0.8515,
+      "step": 2408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.879124823369653e-06,
+      "loss": 1.2342,
+      "step": 2409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.877335413281975e-06,
+      "loss": 0.9315,
+      "step": 2410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.875545452027153e-06,
+      "loss": 0.828,
+      "step": 2411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.873754939948062e-06,
+      "loss": 0.7767,
+      "step": 2412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.871963877387681e-06,
+      "loss": 0.8275,
+      "step": 2413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8701722646891e-06,
+      "loss": 1.3121,
+      "step": 2414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.868380102195507e-06,
+      "loss": 0.5045,
+      "step": 2415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866587390250205e-06,
+      "loss": 0.5971,
+      "step": 2416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.864794129196589e-06,
+      "loss": 0.7303,
+      "step": 2417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86300031937817e-06,
+      "loss": 0.6785,
+      "step": 2418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.861205961138562e-06,
+      "loss": 0.9238,
+      "step": 2419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.859411054821482e-06,
+      "loss": 0.6832,
+      "step": 2420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857615600770752e-06,
+      "loss": 0.7959,
+      "step": 2421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.855819599330299e-06,
+      "loss": 0.2052,
+      "step": 2422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.854023050844156e-06,
+      "loss": 1.1408,
+      "step": 2423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852225955656463e-06,
+      "loss": 0.927,
+      "step": 2424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.85042831411146e-06,
+      "loss": 0.7594,
+      "step": 2425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.848630126553492e-06,
+      "loss": 1.0472,
+      "step": 2426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.846831393327012e-06,
+      "loss": 1.0448,
+      "step": 2427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.845032114776577e-06,
+      "loss": 0.7653,
+      "step": 2428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.843232291246847e-06,
+      "loss": 0.4554,
+      "step": 2429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841431923082583e-06,
+      "loss": 0.6989,
+      "step": 2430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.839631010628658e-06,
+      "loss": 0.5166,
+      "step": 2431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.837829554230043e-06,
+      "loss": 0.7614,
+      "step": 2432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.836027554231817e-06,
+      "loss": 0.9076,
+      "step": 2433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83422501097916e-06,
+      "loss": 0.6857,
+      "step": 2434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83242192481736e-06,
+      "loss": 0.8581,
+      "step": 2435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.830618296091805e-06,
+      "loss": 0.6539,
+      "step": 2436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82881412514799e-06,
+      "loss": 0.4778,
+      "step": 2437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82700941233151e-06,
+      "loss": 0.5334,
+      "step": 2438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.825204157988067e-06,
+      "loss": 1.1374,
+      "step": 2439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.823398362463469e-06,
+      "loss": 0.9359,
+      "step": 2440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821592026103619e-06,
+      "loss": 0.4437,
+      "step": 2441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.819785149254534e-06,
+      "loss": 0.7404,
+      "step": 2442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.817977732262327e-06,
+      "loss": 0.5975,
+      "step": 2443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.816169775473219e-06,
+      "loss": 0.6884,
+      "step": 2444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.814361279233533e-06,
+      "loss": 1.0008,
+      "step": 2445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812552243889692e-06,
+      "loss": 0.5507,
+      "step": 2446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.810742669788228e-06,
+      "loss": 0.9067,
+      "step": 2447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.808932557275773e-06,
+      "loss": 1.0639,
+      "step": 2448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.807121906699062e-06,
+      "loss": 1.0209,
+      "step": 2449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805310718404933e-06,
+      "loss": 0.818,
+      "step": 2450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.803498992740331e-06,
+      "loss": 0.9066,
+      "step": 2451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801686730052295e-06,
+      "loss": 0.6935,
+      "step": 2452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.799873930687979e-06,
+      "loss": 0.922,
+      "step": 2453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798060594994626e-06,
+      "loss": 0.5744,
+      "step": 2454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.796246723319595e-06,
+      "loss": 0.9379,
+      "step": 2455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.794432316010339e-06,
+      "loss": 1.0205,
+      "step": 2456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792617373414418e-06,
+      "loss": 0.8691,
+      "step": 2457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79080189587949e-06,
+      "loss": 1.167,
+      "step": 2458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78898588375332e-06,
+      "loss": 0.8885,
+      "step": 2459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.787169337383773e-06,
+      "loss": 0.5394,
+      "step": 2460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.785352257118817e-06,
+      "loss": 0.7716,
+      "step": 2461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.783534643306523e-06,
+      "loss": 0.9786,
+      "step": 2462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781716496295065e-06,
+      "loss": 0.9691,
+      "step": 2463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.779897816432713e-06,
+      "loss": 0.6761,
+      "step": 2464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778078604067847e-06,
+      "loss": 0.6548,
+      "step": 2465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.776258859548947e-06,
+      "loss": 0.7372,
+      "step": 2466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.774438583224592e-06,
+      "loss": 0.81,
+      "step": 2467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.772617775443463e-06,
+      "loss": 1.0021,
+      "step": 2468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.770796436554346e-06,
+      "loss": 0.8569,
+      "step": 2469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.768974566906126e-06,
+      "loss": 0.7474,
+      "step": 2470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767152166847794e-06,
+      "loss": 0.8239,
+      "step": 2471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765329236728434e-06,
+      "loss": 1.2278,
+      "step": 2472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76350577689724e-06,
+      "loss": 0.8577,
+      "step": 2473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761681787703504e-06,
+      "loss": 0.902,
+      "step": 2474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.75985726949662e-06,
+      "loss": 0.9577,
+      "step": 2475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.758032222626081e-06,
+      "loss": 0.6362,
+      "step": 2476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.756206647441488e-06,
+      "loss": 0.9313,
+      "step": 2477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.754380544292533e-06,
+      "loss": 0.8487,
+      "step": 2478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.752553913529019e-06,
+      "loss": 0.9198,
+      "step": 2479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750726755500842e-06,
+      "loss": 0.5551,
+      "step": 2480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748899070558006e-06,
+      "loss": 0.647,
+      "step": 2481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74707085905061e-06,
+      "loss": 0.884,
+      "step": 2482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.745242121328859e-06,
+      "loss": 0.7639,
+      "step": 2483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743412857743053e-06,
+      "loss": 0.8066,
+      "step": 2484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.741583068643602e-06,
+      "loss": 0.714,
+      "step": 2485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.739752754381003e-06,
+      "loss": 0.9532,
+      "step": 2486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.737921915305866e-06,
+      "loss": 0.8487,
+      "step": 2487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736090551768898e-06,
+      "loss": 0.5976,
+      "step": 2488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.734258664120902e-06,
+      "loss": 0.6655,
+      "step": 2489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732426252712783e-06,
+      "loss": 1.1717,
+      "step": 2490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730593317895554e-06,
+      "loss": 1.0821,
+      "step": 2491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728759860020321e-06,
+      "loss": 0.4379,
+      "step": 2492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.726925879438287e-06,
+      "loss": 0.3459,
+      "step": 2493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.725091376500763e-06,
+      "loss": 0.7085,
+      "step": 2494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.723256351559157e-06,
+      "loss": 0.3645,
+      "step": 2495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.721420804964975e-06,
+      "loss": 0.8781,
+      "step": 2496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719584737069827e-06,
+      "loss": 0.4104,
+      "step": 2497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.717748148225416e-06,
+      "loss": 1.1543,
+      "step": 2498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.715911038783555e-06,
+      "loss": 0.8099,
+      "step": 2499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714073409096145e-06,
+      "loss": 0.6924,
+      "step": 2500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.712235259515197e-06,
+      "loss": 0.8619,
+      "step": 2501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.710396590392817e-06,
+      "loss": 0.8966,
+      "step": 2502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708557402081209e-06,
+      "loss": 0.7763,
+      "step": 2503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706717694932678e-06,
+      "loss": 0.6114,
+      "step": 2504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70487746929963e-06,
+      "loss": 0.8121,
+      "step": 2505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.703036725534567e-06,
+      "loss": 0.9493,
+      "step": 2506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.701195463990096e-06,
+      "loss": 0.737,
+      "step": 2507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699353685018915e-06,
+      "loss": 1.1884,
+      "step": 2508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.697511388973827e-06,
+      "loss": 1.1393,
+      "step": 2509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.695668576207733e-06,
+      "loss": 0.6678,
+      "step": 2510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.693825247073631e-06,
+      "loss": 0.8501,
+      "step": 2511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.691981401924621e-06,
+      "loss": 0.6171,
+      "step": 2512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.690137041113899e-06,
+      "loss": 1.0098,
+      "step": 2513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.688292164994763e-06,
+      "loss": 0.9628,
+      "step": 2514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.686446773920604e-06,
+      "loss": 0.8842,
+      "step": 2515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68460086824492e-06,
+      "loss": 0.7808,
+      "step": 2516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6827544483213e-06,
+      "loss": 0.8397,
+      "step": 2517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.680907514503434e-06,
+      "loss": 0.6548,
+      "step": 2518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.679060067145113e-06,
+      "loss": 0.9935,
+      "step": 2519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.677212106600222e-06,
+      "loss": 0.7823,
+      "step": 2520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67536363322275e-06,
+      "loss": 1.0281,
+      "step": 2521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.673514647366777e-06,
+      "loss": 0.787,
+      "step": 2522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.671665149386485e-06,
+      "loss": 1.0023,
+      "step": 2523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.669815139636158e-06,
+      "loss": 0.9943,
+      "step": 2524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66796461847017e-06,
+      "loss": 0.7057,
+      "step": 2525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666113586242999e-06,
+      "loss": 0.4671,
+      "step": 2526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.664262043309218e-06,
+      "loss": 0.9138,
+      "step": 2527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6624099900235e-06,
+      "loss": 0.8685,
+      "step": 2528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.660557426740613e-06,
+      "loss": 0.6848,
+      "step": 2529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.658704353815426e-06,
+      "loss": 0.5913,
+      "step": 2530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.656850771602903e-06,
+      "loss": 0.9041,
+      "step": 2531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654996680458104e-06,
+      "loss": 0.837,
+      "step": 2532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.653142080736193e-06,
+      "loss": 0.4967,
+      "step": 2533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.651286972792426e-06,
+      "loss": 0.7167,
+      "step": 2534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.649431356982157e-06,
+      "loss": 0.8823,
+      "step": 2535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.647575233660836e-06,
+      "loss": 0.5443,
+      "step": 2536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.645718603184016e-06,
+      "loss": 0.6845,
+      "step": 2537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.643861465907342e-06,
+      "loss": 0.9037,
+      "step": 2538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642003822186556e-06,
+      "loss": 0.8165,
+      "step": 2539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6401456723775e-06,
+      "loss": 0.8545,
+      "step": 2540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.638287016836108e-06,
+      "loss": 0.7431,
+      "step": 2541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63642785591842e-06,
+      "loss": 0.6664,
+      "step": 2542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.634568189980563e-06,
+      "loss": 0.6616,
+      "step": 2543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.632708019378765e-06,
+      "loss": 1.2977,
+      "step": 2544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.630847344469352e-06,
+      "loss": 0.682,
+      "step": 2545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628986165608743e-06,
+      "loss": 1.1008,
+      "step": 2546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.627124483153456e-06,
+      "loss": 1.0164,
+      "step": 2547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.625262297460108e-06,
+      "loss": 0.4552,
+      "step": 2548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.623399608885407e-06,
+      "loss": 0.8543,
+      "step": 2549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.621536417786159e-06,
+      "loss": 0.8109,
+      "step": 2550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.619672724519268e-06,
+      "loss": 0.9695,
+      "step": 2551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.617808529441734e-06,
+      "loss": 0.4402,
+      "step": 2552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.61594383291065e-06,
+      "loss": 0.8655,
+      "step": 2553
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.614078635283211e-06,
+      "loss": 0.6727,
+      "step": 2554
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.612212936916701e-06,
+      "loss": 0.9005,
+      "step": 2555
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.610346738168506e-06,
+      "loss": 0.4621,
+      "step": 2556
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6084800393961025e-06,
+      "loss": 0.783,
+      "step": 2557
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.606612840957066e-06,
+      "loss": 1.0007,
+      "step": 2558
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.604745143209069e-06,
+      "loss": 0.7145,
+      "step": 2559
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.602876946509877e-06,
+      "loss": 0.903,
+      "step": 2560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.601008251217352e-06,
+      "loss": 0.7345,
+      "step": 2561
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.59913905768945e-06,
+      "loss": 1.1443,
+      "step": 2562
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.597269366284225e-06,
+      "loss": 0.7547,
+      "step": 2563
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.595399177359825e-06,
+      "loss": 0.5381,
+      "step": 2564
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.593528491274493e-06,
+      "loss": 0.3182,
+      "step": 2565
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.591657308386568e-06,
+      "loss": 1.0978,
+      "step": 2566
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5897856290544846e-06,
+      "loss": 0.5256,
+      "step": 2567
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.587913453636772e-06,
+      "loss": 0.7005,
+      "step": 2568
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.586040782492052e-06,
+      "loss": 1.0946,
+      "step": 2569
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584167615979046e-06,
+      "loss": 1.0476,
+      "step": 2570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582293954456567e-06,
+      "loss": 0.9912,
+      "step": 2571
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.580419798283523e-06,
+      "loss": 1.0002,
+      "step": 2572
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5785451478189165e-06,
+      "loss": 0.8162,
+      "step": 2573
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.576670003421849e-06,
+      "loss": 0.7124,
+      "step": 2574
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.574794365451512e-06,
+      "loss": 1.036,
+      "step": 2575
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5729182342671904e-06,
+      "loss": 0.6182,
+      "step": 2576
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5710416102282694e-06,
+      "loss": 0.7129,
+      "step": 2577
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569164493694223e-06,
+      "loss": 0.8013,
+      "step": 2578
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.567286885024622e-06,
+      "loss": 0.3673,
+      "step": 2579
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.565408784579131e-06,
+      "loss": 0.5596,
+      "step": 2580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.56353019271751e-06,
+      "loss": 0.2085,
+      "step": 2581
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.561651109799611e-06,
+      "loss": 0.7815,
+      "step": 2582
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.559771536185383e-06,
+      "loss": 0.8591,
+      "step": 2583
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.557891472234864e-06,
+      "loss": 0.8272,
+      "step": 2584
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556010918308192e-06,
+      "loss": 0.5903,
+      "step": 2585
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.554129874765594e-06,
+      "loss": 1.1911,
+      "step": 2586
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.552248341967396e-06,
+      "loss": 0.9241,
+      "step": 2587
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.55036632027401e-06,
+      "loss": 0.8844,
+      "step": 2588
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.548483810045949e-06,
+      "loss": 0.8981,
+      "step": 2589
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.546600811643816e-06,
+      "loss": 1.0494,
+      "step": 2590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544717325428308e-06,
+      "loss": 0.857,
+      "step": 2591
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.542833351760215e-06,
+      "loss": 0.7401,
+      "step": 2592
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540948891000422e-06,
+      "loss": 0.8319,
+      "step": 2593
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.539063943509907e-06,
+      "loss": 0.8038,
+      "step": 2594
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.537178509649738e-06,
+      "loss": 0.9795,
+      "step": 2595
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535292589781083e-06,
+      "loss": 0.6795,
+      "step": 2596
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.533406184265195e-06,
+      "loss": 0.7463,
+      "step": 2597
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5315192934634255e-06,
+      "loss": 1.0753,
+      "step": 2598
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529631917737217e-06,
+      "loss": 0.5792,
+      "step": 2599
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.527744057448104e-06,
+      "loss": 0.9768,
+      "step": 2600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.525855712957717e-06,
+      "loss": 1.1389,
+      "step": 2601
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.523966884627776e-06,
+      "loss": 0.7847,
+      "step": 2602
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.522077572820096e-06,
+      "loss": 0.9554,
+      "step": 2603
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.520187777896584e-06,
+      "loss": 0.6305,
+      "step": 2604
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.518297500219236e-06,
+      "loss": 0.9916,
+      "step": 2605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.516406740150146e-06,
+      "loss": 0.6869,
+      "step": 2606
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.514515498051499e-06,
+      "loss": 1.0207,
+      "step": 2607
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.51262377428557e-06,
+      "loss": 0.582,
+      "step": 2608
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.510731569214727e-06,
+      "loss": 0.809,
+      "step": 2609
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.50883888320143e-06,
+      "loss": 0.9182,
+      "step": 2610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.506945716608236e-06,
+      "loss": 0.672,
+      "step": 2611
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.505052069797785e-06,
+      "loss": 0.6458,
+      "step": 2612
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.503157943132817e-06,
+      "loss": 0.9922,
+      "step": 2613
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.501263336976161e-06,
+      "loss": 0.6685,
+      "step": 2614
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.499368251690737e-06,
+      "loss": 0.7626,
+      "step": 2615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.497472687639555e-06,
+      "loss": 1.2893,
+      "step": 2616
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4955766451857236e-06,
+      "loss": 0.563,
+      "step": 2617
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.493680124692436e-06,
+      "loss": 0.6878,
+      "step": 2618
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.491783126522981e-06,
+      "loss": 0.4609,
+      "step": 2619
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.489885651040736e-06,
+      "loss": 0.8575,
+      "step": 2620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.487987698609173e-06,
+      "loss": 0.7299,
+      "step": 2621
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.486089269591852e-06,
+      "loss": 0.6812,
+      "step": 2622
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.484190364352427e-06,
+      "loss": 0.9247,
+      "step": 2623
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.482290983254643e-06,
+      "loss": 0.7015,
+      "step": 2624
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.480391126662334e-06,
+      "loss": 0.9188,
+      "step": 2625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.478490794939427e-06,
+      "loss": 0.9593,
+      "step": 2626
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.476589988449939e-06,
+      "loss": 0.6676,
+      "step": 2627
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.47468870755798e-06,
+      "loss": 0.5742,
+      "step": 2628
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472786952627746e-06,
+      "loss": 1.1076,
+      "step": 2629
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.470884724023531e-06,
+      "loss": 0.9519,
+      "step": 2630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468982022109714e-06,
+      "loss": 0.9421,
+      "step": 2631
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4670788472507676e-06,
+      "loss": 0.9554,
+      "step": 2632
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.465175199811252e-06,
+      "loss": 0.9596,
+      "step": 2633
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4632710801558205e-06,
+      "loss": 0.7796,
+      "step": 2634
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4613664886492175e-06,
+      "loss": 0.7613,
+      "step": 2635
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.459461425656276e-06,
+      "loss": 0.4086,
+      "step": 2636
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.457555891541918e-06,
+      "loss": 0.5072,
+      "step": 2637
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.455649886671159e-06,
+      "loss": 0.8672,
+      "step": 2638
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453743411409104e-06,
+      "loss": 0.6916,
+      "step": 2639
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451836466120946e-06,
+      "loss": 0.7015,
+      "step": 2640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.449929051171969e-06,
+      "loss": 0.8517,
+      "step": 2641
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.448021166927548e-06,
+      "loss": 0.322,
+      "step": 2642
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4461128137531465e-06,
+      "loss": 0.6197,
+      "step": 2643
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.444203992014321e-06,
+      "loss": 0.6817,
+      "step": 2644
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442294702076712e-06,
+      "loss": 1.0395,
+      "step": 2645
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.440384944306056e-06,
+      "loss": 0.5642,
+      "step": 2646
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.438474719068174e-06,
+      "loss": 0.8792,
+      "step": 2647
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.436564026728978e-06,
+      "loss": 0.8718,
+      "step": 2648
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.434652867654473e-06,
+      "loss": 0.631,
+      "step": 2649
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.432741242210748e-06,
+      "loss": 0.9065,
+      "step": 2650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.430829150763985e-06,
+      "loss": 1.1446,
+      "step": 2651
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.428916593680454e-06,
+      "loss": 0.8041,
+      "step": 2652
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.427003571326514e-06,
+      "loss": 0.9924,
+      "step": 2653
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425090084068613e-06,
+      "loss": 1.1018,
+      "step": 2654
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.423176132273291e-06,
+      "loss": 0.927,
+      "step": 2655
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.421261716307171e-06,
+      "loss": 0.9881,
+      "step": 2656
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.419346836536969e-06,
+      "loss": 0.9513,
+      "step": 2657
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.417431493329491e-06,
+      "loss": 1.0877,
+      "step": 2658
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4155156870516294e-06,
+      "loss": 0.6587,
+      "step": 2659
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.413599418070366e-06,
+      "loss": 0.8091,
+      "step": 2660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411682686752771e-06,
+      "loss": 0.577,
+      "step": 2661
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.409765493466003e-06,
+      "loss": 0.6898,
+      "step": 2662
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4078478385773086e-06,
+      "loss": 0.7353,
+      "step": 2663
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.405929722454026e-06,
+      "loss": 0.8892,
+      "step": 2664
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.404011145463577e-06,
+      "loss": 0.7549,
+      "step": 2665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402092107973476e-06,
+      "loss": 0.2947,
+      "step": 2666
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.400172610351322e-06,
+      "loss": 1.1053,
+      "step": 2667
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.398252652964803e-06,
+      "loss": 0.765,
+      "step": 2668
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3963322361817e-06,
+      "loss": 0.9348,
+      "step": 2669
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3944113603698735e-06,
+      "loss": 0.7867,
+      "step": 2670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.392490025897279e-06,
+      "loss": 0.6302,
+      "step": 2671
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.390568233131956e-06,
+      "loss": 0.8762,
+      "step": 2672
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3886459824420335e-06,
+      "loss": 0.6035,
+      "step": 2673
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.386723274195726e-06,
+      "loss": 0.6971,
+      "step": 2674
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3848001087613394e-06,
+      "loss": 1.1443,
+      "step": 2675
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.382876486507264e-06,
+      "loss": 0.9417,
+      "step": 2676
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.38095240780198e-06,
+      "loss": 0.7243,
+      "step": 2677
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.379027873014052e-06,
+      "loss": 0.7395,
+      "step": 2678
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.377102882512135e-06,
+      "loss": 0.9631,
+      "step": 2679
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.375177436664968e-06,
+      "loss": 0.9782,
+      "step": 2680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373251535841383e-06,
+      "loss": 0.8866,
+      "step": 2681
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.371325180410291e-06,
+      "loss": 0.9083,
+      "step": 2682
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.369398370740698e-06,
+      "loss": 1.1778,
+      "step": 2683
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367471107201691e-06,
+      "loss": 0.7672,
+      "step": 2684
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365543390162447e-06,
+      "loss": 0.3258,
+      "step": 2685
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.363615219992231e-06,
+      "loss": 0.716,
+      "step": 2686
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.36168659706039e-06,
+      "loss": 0.9525,
+      "step": 2687
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.359757521736364e-06,
+      "loss": 0.9556,
+      "step": 2688
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357827994389675e-06,
+      "loss": 0.7423,
+      "step": 2689
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.355898015389932e-06,
+      "loss": 0.5309,
+      "step": 2690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.353967585106833e-06,
+      "loss": 0.8001,
+      "step": 2691
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3520367039101605e-06,
+      "loss": 0.8955,
+      "step": 2692
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.350105372169785e-06,
+      "loss": 0.5888,
+      "step": 2693
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.34817359025566e-06,
+      "loss": 0.7843,
+      "step": 2694
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346241358537829e-06,
+      "loss": 0.6942,
+      "step": 2695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.344308677386419e-06,
+      "loss": 1.2421,
+      "step": 2696
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.342375547171646e-06,
+      "loss": 0.9557,
+      "step": 2697
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.340441968263808e-06,
+      "loss": 1.1325,
+      "step": 2698
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3385079410332935e-06,
+      "loss": 1.3182,
+      "step": 2699
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.336573465850573e-06,
+      "loss": 1.3077,
+      "step": 2700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.334638543086203e-06,
+      "loss": 0.8022,
+      "step": 2701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.33270317311083e-06,
+      "loss": 0.9641,
+      "step": 2702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.330767356295181e-06,
+      "loss": 0.6215,
+      "step": 2703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328831093010073e-06,
+      "loss": 0.8918,
+      "step": 2704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.326894383626404e-06,
+      "loss": 0.5842,
+      "step": 2705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.324957228515161e-06,
+      "loss": 0.8387,
+      "step": 2706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.323019628047414e-06,
+      "loss": 0.7059,
+      "step": 2707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.321081582594322e-06,
+      "loss": 0.9242,
+      "step": 2708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.319143092527125e-06,
+      "loss": 0.9506,
+      "step": 2709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.31720415821715e-06,
+      "loss": 0.8705,
+      "step": 2710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.315264780035808e-06,
+      "loss": 0.6902,
+      "step": 2711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.313324958354598e-06,
+      "loss": 0.9619,
+      "step": 2712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.311384693545101e-06,
+      "loss": 0.75,
+      "step": 2713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309443985978983e-06,
+      "loss": 0.8767,
+      "step": 2714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.307502836027998e-06,
+      "loss": 0.8257,
+      "step": 2715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.305561244063981e-06,
+      "loss": 0.9838,
+      "step": 2716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.303619210458853e-06,
+      "loss": 0.8129,
+      "step": 2717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3016767355846195e-06,
+      "loss": 0.6152,
+      "step": 2718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.299733819813371e-06,
+      "loss": 0.9125,
+      "step": 2719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.297790463517283e-06,
+      "loss": 1.0273,
+      "step": 2720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.295846667068613e-06,
+      "loss": 1.1055,
+      "step": 2721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.293902430839705e-06,
+      "loss": 1.1345,
+      "step": 2722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.291957755202986e-06,
+      "loss": 0.716,
+      "step": 2723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.290012640530969e-06,
+      "loss": 1.068,
+      "step": 2724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.28806708719625e-06,
+      "loss": 1.0819,
+      "step": 2725
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.286121095571508e-06,
+      "loss": 0.4008,
+      "step": 2726
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.284174666029506e-06,
+      "loss": 0.4474,
+      "step": 2727
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.282227798943095e-06,
+      "loss": 0.9994,
+      "step": 2728
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.280280494685202e-06,
+      "loss": 0.694,
+      "step": 2729
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2783327536288465e-06,
+      "loss": 0.9812,
+      "step": 2730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.276384576147124e-06,
+      "loss": 0.8569,
+      "step": 2731
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27443596261322e-06,
+      "loss": 0.7573,
+      "step": 2732
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2724869134004e-06,
+      "loss": 0.5524,
+      "step": 2733
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.270537428882012e-06,
+      "loss": 0.9313,
+      "step": 2734
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.268587509431488e-06,
+      "loss": 0.5829,
+      "step": 2735
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2666371554223495e-06,
+      "loss": 0.8859,
+      "step": 2736
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.26468636722819e-06,
+      "loss": 0.6958,
+      "step": 2737
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262735145222696e-06,
+      "loss": 0.794,
+      "step": 2738
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.260783489779632e-06,
+      "loss": 0.6766,
+      "step": 2739
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.258831401272844e-06,
+      "loss": 0.7842,
+      "step": 2740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.256878880076269e-06,
+      "loss": 1.2402,
+      "step": 2741
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254925926563919e-06,
+      "loss": 0.6367,
+      "step": 2742
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.252972541109892e-06,
+      "loss": 0.8667,
+      "step": 2743
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.251018724088367e-06,
+      "loss": 0.5872,
+      "step": 2744
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249064475873608e-06,
+      "loss": 0.5848,
+      "step": 2745
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.247109796839959e-06,
+      "loss": 0.6509,
+      "step": 2746
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.245154687361851e-06,
+      "loss": 0.9666,
+      "step": 2747
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.243199147813791e-06,
+      "loss": 0.4671,
+      "step": 2748
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241243178570374e-06,
+      "loss": 0.735,
+      "step": 2749
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.239286780006274e-06,
+      "loss": 1.0749,
+      "step": 2750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.237329952496249e-06,
+      "loss": 0.8412,
+      "step": 2751
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.235372696415138e-06,
+      "loss": 0.9934,
+      "step": 2752
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.233415012137863e-06,
+      "loss": 1.0816,
+      "step": 2753
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.231456900039428e-06,
+      "loss": 0.7171,
+      "step": 2754
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.229498360494918e-06,
+      "loss": 0.792,
+      "step": 2755
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227539393879499e-06,
+      "loss": 0.857,
+      "step": 2756
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.225580000568423e-06,
+      "loss": 1.074,
+      "step": 2757
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.22362018093702e-06,
+      "loss": 0.8056,
+      "step": 2758
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.221659935360703e-06,
+      "loss": 0.7977,
+      "step": 2759
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2196992642149655e-06,
+      "loss": 0.5392,
+      "step": 2760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2177381678753836e-06,
+      "loss": 0.7179,
+      "step": 2761
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.215776646717614e-06,
+      "loss": 0.8331,
+      "step": 2762
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2138147011173955e-06,
+      "loss": 0.9585,
+      "step": 2763
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.211852331450548e-06,
+      "loss": 0.5979,
+      "step": 2764
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.209889538092976e-06,
+      "loss": 0.9312,
+      "step": 2765
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.207926321420655e-06,
+      "loss": 0.8825,
+      "step": 2766
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.205962681809655e-06,
+      "loss": 0.5155,
+      "step": 2767
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.203998619636116e-06,
+      "loss": 0.8706,
+      "step": 2768
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.202034135276265e-06,
+      "loss": 1.1201,
+      "step": 2769
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.200069229106409e-06,
+      "loss": 0.6816,
+      "step": 2770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1981039015029345e-06,
+      "loss": 1.0996,
+      "step": 2771
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.196138152842309e-06,
+      "loss": 0.8878,
+      "step": 2772
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.19417198350108e-06,
+      "loss": 0.6538,
+      "step": 2773
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1922053938558776e-06,
+      "loss": 0.9184,
+      "step": 2774
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.190238384283413e-06,
+      "loss": 0.9236,
+      "step": 2775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.188270955160471e-06,
+      "loss": 0.7904,
+      "step": 2776
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.186303106863928e-06,
+      "loss": 1.0668,
+      "step": 2777
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.18433483977073e-06,
+      "loss": 1.0096,
+      "step": 2778
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.18236615425791e-06,
+      "loss": 0.8139,
+      "step": 2779
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.180397050702579e-06,
+      "loss": 0.7201,
+      "step": 2780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.178427529481929e-06,
+      "loss": 0.4179,
+      "step": 2781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.176457590973227e-06,
+      "loss": 0.6205,
+      "step": 2782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1744872355538275e-06,
+      "loss": 1.1261,
+      "step": 2783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.172516463601161e-06,
+      "loss": 1.0092,
+      "step": 2784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.170545275492737e-06,
+      "loss": 0.6991,
+      "step": 2785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.168573671606148e-06,
+      "loss": 0.7558,
+      "step": 2786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.166601652319062e-06,
+      "loss": 0.8717,
+      "step": 2787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164629218009229e-06,
+      "loss": 1.0353,
+      "step": 2788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.162656369054478e-06,
+      "loss": 0.6578,
+      "step": 2789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.160683105832717e-06,
+      "loss": 0.843,
+      "step": 2790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.158709428721936e-06,
+      "loss": 0.7813,
+      "step": 2791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1567353381002e-06,
+      "loss": 0.7552,
+      "step": 2792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.154760834345657e-06,
+      "loss": 0.4936,
+      "step": 2793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1527859178365314e-06,
+      "loss": 0.6177,
+      "step": 2794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.150810588951127e-06,
+      "loss": 1.0622,
+      "step": 2795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.148834848067827e-06,
+      "loss": 0.8462,
+      "step": 2796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.146858695565096e-06,
+      "loss": 0.6864,
+      "step": 2797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.144882131821476e-06,
+      "loss": 0.4606,
+      "step": 2798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.142905157215585e-06,
+      "loss": 0.8486,
+      "step": 2799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1409277721261216e-06,
+      "loss": 0.7898,
+      "step": 2800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1389499769318645e-06,
+      "loss": 0.7431,
+      "step": 2801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13697177201167e-06,
+      "loss": 0.7765,
+      "step": 2802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134993157744472e-06,
+      "loss": 0.8311,
+      "step": 2803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.133014134509282e-06,
+      "loss": 0.8291,
+      "step": 2804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.131034702685191e-06,
+      "loss": 0.8131,
+      "step": 2805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.129054862651373e-06,
+      "loss": 0.7313,
+      "step": 2806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.127074614787069e-06,
+      "loss": 0.8985,
+      "step": 2807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.12509395947161e-06,
+      "loss": 0.9318,
+      "step": 2808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123112897084398e-06,
+      "loss": 1.1155,
+      "step": 2809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.121131428004915e-06,
+      "loss": 0.6959,
+      "step": 2810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.119149552612721e-06,
+      "loss": 0.3756,
+      "step": 2811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.117167271287453e-06,
+      "loss": 0.7633,
+      "step": 2812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.115184584408826e-06,
+      "loss": 1.0277,
+      "step": 2813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.113201492356634e-06,
+      "loss": 0.536,
+      "step": 2814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.111217995510746e-06,
+      "loss": 0.7163,
+      "step": 2815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.109234094251114e-06,
+      "loss": 0.61,
+      "step": 2816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.107249788957759e-06,
+      "loss": 0.7748,
+      "step": 2817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.105265080010786e-06,
+      "loss": 0.4246,
+      "step": 2818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103279967790377e-06,
+      "loss": 0.9378,
+      "step": 2819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.101294452676786e-06,
+      "loss": 1.0356,
+      "step": 2820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.099308535050351e-06,
+      "loss": 0.4472,
+      "step": 2821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.097322215291481e-06,
+      "loss": 0.7106,
+      "step": 2822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.095335493780668e-06,
+      "loss": 0.7313,
+      "step": 2823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.093348370898475e-06,
+      "loss": 0.706,
+      "step": 2824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.091360847025547e-06,
+      "loss": 1.308,
+      "step": 2825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.089372922542603e-06,
+      "loss": 0.6018,
+      "step": 2826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.087384597830438e-06,
+      "loss": 0.9013,
+      "step": 2827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.085395873269927e-06,
+      "loss": 1.041,
+      "step": 2828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.083406749242016e-06,
+      "loss": 0.9875,
+      "step": 2829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.081417226127736e-06,
+      "loss": 0.6977,
+      "step": 2830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0794273043081865e-06,
+      "loss": 0.9712,
+      "step": 2831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.077436984164548e-06,
+      "loss": 0.5617,
+      "step": 2832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075446266078072e-06,
+      "loss": 0.4221,
+      "step": 2833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.073455150430094e-06,
+      "loss": 0.6186,
+      "step": 2834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07146363760202e-06,
+      "loss": 0.5932,
+      "step": 2835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.069471727975334e-06,
+      "loss": 0.5774,
+      "step": 2836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.067479421931596e-06,
+      "loss": 0.9953,
+      "step": 2837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0654867198524406e-06,
+      "loss": 0.9896,
+      "step": 2838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.063493622119578e-06,
+      "loss": 0.5455,
+      "step": 2839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.061500129114799e-06,
+      "loss": 0.409,
+      "step": 2840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.059506241219964e-06,
+      "loss": 0.9892,
+      "step": 2841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.057511958817013e-06,
+      "loss": 0.6781,
+      "step": 2842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055517282287959e-06,
+      "loss": 0.614,
+      "step": 2843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0535222120148925e-06,
+      "loss": 0.9569,
+      "step": 2844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.051526748379978e-06,
+      "loss": 1.0973,
+      "step": 2845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.049530891765457e-06,
+      "loss": 0.736,
+      "step": 2846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0475346425536436e-06,
+      "loss": 0.8052,
+      "step": 2847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.04553800112693e-06,
+      "loss": 0.5763,
+      "step": 2848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.043540967867782e-06,
+      "loss": 0.5665,
+      "step": 2849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.041543543158741e-06,
+      "loss": 0.9351,
+      "step": 2850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.039545727382422e-06,
+      "loss": 0.9498,
+      "step": 2851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.037547520921516e-06,
+      "loss": 0.6905,
+      "step": 2852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0355489241587906e-06,
+      "loss": 0.5453,
+      "step": 2853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.033549937477085e-06,
+      "loss": 0.6252,
+      "step": 2854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.031550561259315e-06,
+      "loss": 0.6835,
+      "step": 2855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.029550795888471e-06,
+      "loss": 0.5308,
+      "step": 2856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.027550641747617e-06,
+      "loss": 0.6438,
+      "step": 2857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.025550099219891e-06,
+      "loss": 0.6053,
+      "step": 2858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0235491686885084e-06,
+      "loss": 0.6229,
+      "step": 2859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.021547850536755e-06,
+      "loss": 1.0572,
+      "step": 2860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.019546145147994e-06,
+      "loss": 0.8737,
+      "step": 2861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.017544052905661e-06,
+      "loss": 0.517,
+      "step": 2862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.015541574193264e-06,
+      "loss": 1.015,
+      "step": 2863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.013538709394391e-06,
+      "loss": 0.7192,
+      "step": 2864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.011535458892698e-06,
+      "loss": 0.9085,
+      "step": 2865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009531823071918e-06,
+      "loss": 0.6101,
+      "step": 2866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.007527802315857e-06,
+      "loss": 0.961,
+      "step": 2867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0055233970083926e-06,
+      "loss": 0.7326,
+      "step": 2868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.003518607533479e-06,
+      "loss": 0.6933,
+      "step": 2869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.001513434275146e-06,
+      "loss": 0.6876,
+      "step": 2870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99950787761749e-06,
+      "loss": 0.8943,
+      "step": 2871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997501937944687e-06,
+      "loss": 0.7076,
+      "step": 2872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.995495615640983e-06,
+      "loss": 1.0431,
+      "step": 2873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993488911090699e-06,
+      "loss": 0.5804,
+      "step": 2874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.991481824678229e-06,
+      "loss": 0.7267,
+      "step": 2875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.98947435678804e-06,
+      "loss": 1.3653,
+      "step": 2876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.98746650780467e-06,
+      "loss": 0.3568,
+      "step": 2877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.985458278112734e-06,
+      "loss": 0.9418,
+      "step": 2878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.983449668096918e-06,
+      "loss": 0.6656,
+      "step": 2879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.98144067814198e-06,
+      "loss": 0.7775,
+      "step": 2880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.979431308632751e-06,
+      "loss": 0.7298,
+      "step": 2881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9774215599541355e-06,
+      "loss": 0.962,
+      "step": 2882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.975411432491111e-06,
+      "loss": 1.0623,
+      "step": 2883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9734009266287265e-06,
+      "loss": 1.1694,
+      "step": 2884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.971390042752103e-06,
+      "loss": 0.3023,
+      "step": 2885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.969378781246436e-06,
+      "loss": 0.5654,
+      "step": 2886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.967367142496992e-06,
+      "loss": 0.9047,
+      "step": 2887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.965355126889109e-06,
+      "loss": 1.0814,
+      "step": 2888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.963342734808198e-06,
+      "loss": 0.6507,
+      "step": 2889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.961329966639745e-06,
+      "loss": 1.1519,
+      "step": 2890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9593168227693006e-06,
+      "loss": 0.9636,
+      "step": 2891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.957303303582495e-06,
+      "loss": 0.9942,
+      "step": 2892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.955289409465027e-06,
+      "loss": 0.7093,
+      "step": 2893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.953275140802666e-06,
+      "loss": 0.7023,
+      "step": 2894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.951260497981258e-06,
+      "loss": 0.9691,
+      "step": 2895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.949245481386714e-06,
+      "loss": 1.1034,
+      "step": 2896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.947230091405021e-06,
+      "loss": 0.7096,
+      "step": 2897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.945214328422235e-06,
+      "loss": 0.7748,
+      "step": 2898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9431981928244876e-06,
+      "loss": 0.9069,
+      "step": 2899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.941181684997977e-06,
+      "loss": 0.3353,
+      "step": 2900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.939164805328977e-06,
+      "loss": 1.1715,
+      "step": 2901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9371475542038295e-06,
+      "loss": 0.9279,
+      "step": 2902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.935129932008946e-06,
+      "loss": 0.8506,
+      "step": 2903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.933111939130814e-06,
+      "loss": 0.5229,
+      "step": 2904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.931093575955989e-06,
+      "loss": 0.7778,
+      "step": 2905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9290748428710974e-06,
+      "loss": 0.9346,
+      "step": 2906
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.927055740262838e-06,
+      "loss": 0.7705,
+      "step": 2907
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92503626851798e-06,
+      "loss": 0.8921,
+      "step": 2908
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.923016428023361e-06,
+      "loss": 0.8847,
+      "step": 2909
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.920996219165891e-06,
+      "loss": 0.6661,
+      "step": 2910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.918975642332552e-06,
+      "loss": 0.4428,
+      "step": 2911
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.916954697910394e-06,
+      "loss": 1.0514,
+      "step": 2912
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.914933386286538e-06,
+      "loss": 0.8324,
+      "step": 2913
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.912911707848177e-06,
+      "loss": 0.6828,
+      "step": 2914
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.910889662982572e-06,
+      "loss": 1.1023,
+      "step": 2915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.908867252077056e-06,
+      "loss": 0.7278,
+      "step": 2916
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.90684447551903e-06,
+      "loss": 0.5032,
+      "step": 2917
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904821333695968e-06,
+      "loss": 0.8698,
+      "step": 2918
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9027978269954125e-06,
+      "loss": 0.5137,
+      "step": 2919
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.900773955804973e-06,
+      "loss": 0.3708,
+      "step": 2920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.898749720512335e-06,
+      "loss": 1.0948,
+      "step": 2921
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.896725121505248e-06,
+      "loss": 0.4317,
+      "step": 2922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.894700159171535e-06,
+      "loss": 1.1579,
+      "step": 2923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.892674833899085e-06,
+      "loss": 0.7301,
+      "step": 2924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.890649146075861e-06,
+      "loss": 0.7194,
+      "step": 2925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888623096089892e-06,
+      "loss": 1.2272,
+      "step": 2926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.886596684329276e-06,
+      "loss": 0.971,
+      "step": 2927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884569911182184e-06,
+      "loss": 0.8809,
+      "step": 2928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.882542777036852e-06,
+      "loss": 0.9468,
+      "step": 2929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.880515282281589e-06,
+      "loss": 1.2775,
+      "step": 2930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.87848742730477e-06,
+      "loss": 1.0901,
+      "step": 2931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.87645921249484e-06,
+      "loss": 0.5172,
+      "step": 2932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8744306382403144e-06,
+      "loss": 0.6996,
+      "step": 2933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.872401704929774e-06,
+      "loss": 0.6429,
+      "step": 2934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8703724129518725e-06,
+      "loss": 0.8958,
+      "step": 2935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.868342762695329e-06,
+      "loss": 0.7567,
+      "step": 2936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.866312754548934e-06,
+      "loss": 1.0298,
+      "step": 2937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.864282388901544e-06,
+      "loss": 0.6903,
+      "step": 2938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.862251666142086e-06,
+      "loss": 0.5526,
+      "step": 2939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8602205866595514e-06,
+      "loss": 0.5545,
+      "step": 2940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.858189150843008e-06,
+      "loss": 0.5882,
+      "step": 2941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.856157359081583e-06,
+      "loss": 0.3667,
+      "step": 2942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.854125211764479e-06,
+      "loss": 0.7168,
+      "step": 2943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.85209270928096e-06,
+      "loss": 0.6402,
+      "step": 2944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.850059852020363e-06,
+      "loss": 0.6124,
+      "step": 2945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848026640372092e-06,
+      "loss": 0.8163,
+      "step": 2946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.845993074725616e-06,
+      "loss": 0.7611,
+      "step": 2947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.843959155470476e-06,
+      "loss": 0.9348,
+      "step": 2948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.84192488299628e-06,
+      "loss": 0.661,
+      "step": 2949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.839890257692699e-06,
+      "loss": 0.482,
+      "step": 2950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.837855279949479e-06,
+      "loss": 0.8853,
+      "step": 2951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.835819950156426e-06,
+      "loss": 0.5693,
+      "step": 2952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.833784268703418e-06,
+      "loss": 0.584,
+      "step": 2953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8317482359804024e-06,
+      "loss": 0.6096,
+      "step": 2954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.829711852377386e-06,
+      "loss": 0.7017,
+      "step": 2955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82767511828445e-06,
+      "loss": 0.7521,
+      "step": 2956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.825638034091741e-06,
+      "loss": 0.7266,
+      "step": 2957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.823600600189473e-06,
+      "loss": 0.8415,
+      "step": 2958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.821562816967921e-06,
+      "loss": 0.8012,
+      "step": 2959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.819524684817439e-06,
+      "loss": 0.802,
+      "step": 2960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.817486204128436e-06,
+      "loss": 0.8989,
+      "step": 2961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815447375291393e-06,
+      "loss": 0.6391,
+      "step": 2962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.813408198696858e-06,
+      "loss": 0.7193,
+      "step": 2963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.811368674735446e-06,
+      "loss": 0.906,
+      "step": 2964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809328803797837e-06,
+      "loss": 0.6121,
+      "step": 2965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8072885862747766e-06,
+      "loss": 0.9084,
+      "step": 2966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.805248022557078e-06,
+      "loss": 0.9487,
+      "step": 2967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.803207113035622e-06,
+      "loss": 0.847,
+      "step": 2968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.801165858101355e-06,
+      "loss": 0.4964,
+      "step": 2969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.799124258145288e-06,
+      "loss": 0.839,
+      "step": 2970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.797082313558498e-06,
+      "loss": 0.5937,
+      "step": 2971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.795040024732131e-06,
+      "loss": 1.0331,
+      "step": 2972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.792997392057396e-06,
+      "loss": 0.7871,
+      "step": 2973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.790954415925568e-06,
+      "loss": 0.7223,
+      "step": 2974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.788911096727992e-06,
+      "loss": 1.127,
+      "step": 2975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.786867434856071e-06,
+      "loss": 0.8837,
+      "step": 2976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.784823430701282e-06,
+      "loss": 0.8154,
+      "step": 2977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.782779084655161e-06,
+      "loss": 0.7237,
+      "step": 2978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.780734397109314e-06,
+      "loss": 0.5935,
+      "step": 2979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7786893684554085e-06,
+      "loss": 0.9551,
+      "step": 2980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.776643999085181e-06,
+      "loss": 0.698,
+      "step": 2981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.774598289390431e-06,
+      "loss": 0.7166,
+      "step": 2982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.772552239763023e-06,
+      "loss": 0.8159,
+      "step": 2983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.770505850594889e-06,
+      "loss": 0.9307,
+      "step": 2984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7684591222780236e-06,
+      "loss": 0.7135,
+      "step": 2985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.766412055204488e-06,
+      "loss": 0.6454,
+      "step": 2986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.764364649766406e-06,
+      "loss": 1.0969,
+      "step": 2987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76231690635597e-06,
+      "loss": 0.9494,
+      "step": 2988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.760268825365432e-06,
+      "loss": 0.9033,
+      "step": 2989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.758220407187115e-06,
+      "loss": 1.061,
+      "step": 2990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.756171652213399e-06,
+      "loss": 1.0282,
+      "step": 2991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7541225608367355e-06,
+      "loss": 0.7711,
+      "step": 2992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.752073133449638e-06,
+      "loss": 0.7323,
+      "step": 2993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.750023370444683e-06,
+      "loss": 0.7079,
+      "step": 2994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.747973272214512e-06,
+      "loss": 0.8845,
+      "step": 2995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.745922839151832e-06,
+      "loss": 0.932,
+      "step": 2996
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.743872071649411e-06,
+      "loss": 0.6638,
+      "step": 2997
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7418209701000854e-06,
+      "loss": 0.4806,
+      "step": 2998
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.739769534896753e-06,
+      "loss": 1.0005,
+      "step": 2999
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.737717766432375e-06,
+      "loss": 0.6533,
+      "step": 3000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.735665665099978e-06,
+      "loss": 1.0929,
+      "step": 3001
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.73361323129265e-06,
+      "loss": 1.187,
+      "step": 3002
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.731560465403546e-06,
+      "loss": 0.9403,
+      "step": 3003
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.729507367825884e-06,
+      "loss": 1.0092,
+      "step": 3004
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.727453938952942e-06,
+      "loss": 1.1094,
+      "step": 3005
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725400179178066e-06,
+      "loss": 0.7868,
+      "step": 3006
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.723346088894661e-06,
+      "loss": 1.1221,
+      "step": 3007
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7212916684962006e-06,
+      "loss": 0.9119,
+      "step": 3008
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.719236918376216e-06,
+      "loss": 0.9619,
+      "step": 3009
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.717181838928307e-06,
+      "loss": 0.561,
+      "step": 3010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.715126430546132e-06,
+      "loss": 0.638,
+      "step": 3011
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.713070693623413e-06,
+      "loss": 1.1047,
+      "step": 3012
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.711014628553939e-06,
+      "loss": 0.5145,
+      "step": 3013
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.708958235731558e-06,
+      "loss": 0.6961,
+      "step": 3014
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.706901515550179e-06,
+      "loss": 0.4295,
+      "step": 3015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7048444684037794e-06,
+      "loss": 0.9347,
+      "step": 3016
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.702787094686396e-06,
+      "loss": 0.9622,
+      "step": 3017
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.700729394792128e-06,
+      "loss": 0.5817,
+      "step": 3018
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.698671369115137e-06,
+      "loss": 0.9059,
+      "step": 3019
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.696613018049648e-06,
+      "loss": 0.977,
+      "step": 3020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.694554341989948e-06,
+      "loss": 0.6572,
+      "step": 3021
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.692495341330385e-06,
+      "loss": 0.8386,
+      "step": 3022
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.690436016465372e-06,
+      "loss": 0.7598,
+      "step": 3023
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.688376367789382e-06,
+      "loss": 0.749,
+      "step": 3024
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.686316395696948e-06,
+      "loss": 0.6182,
+      "step": 3025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6842561005826715e-06,
+      "loss": 1.0182,
+      "step": 3026
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6821954828412076e-06,
+      "loss": 0.5893,
+      "step": 3027
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68013454286728e-06,
+      "loss": 0.5788,
+      "step": 3028
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.67807328105567e-06,
+      "loss": 0.5767,
+      "step": 3029
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.676011697801223e-06,
+      "loss": 0.9308,
+      "step": 3030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.673949793498845e-06,
+      "loss": 0.9491,
+      "step": 3031
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.671887568543502e-06,
+      "loss": 0.7494,
+      "step": 3032
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.669825023330227e-06,
+      "loss": 0.9569,
+      "step": 3033
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667762158254104e-06,
+      "loss": 0.9323,
+      "step": 3034
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.665698973710289e-06,
+      "loss": 0.7539,
+      "step": 3035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663635470093993e-06,
+      "loss": 1.006,
+      "step": 3036
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.661571647800491e-06,
+      "loss": 0.8027,
+      "step": 3037
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.659507507225117e-06,
+      "loss": 0.999,
+      "step": 3038
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.657443048763267e-06,
+      "loss": 0.6163,
+      "step": 3039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.655378272810398e-06,
+      "loss": 0.8918,
+      "step": 3040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.653313179762026e-06,
+      "loss": 0.2958,
+      "step": 3041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.651247770013732e-06,
+      "loss": 0.465,
+      "step": 3042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.649182043961153e-06,
+      "loss": 0.6276,
+      "step": 3043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.64711600199999e-06,
+      "loss": 0.7857,
+      "step": 3044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6450496445260015e-06,
+      "loss": 0.9333,
+      "step": 3045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.642982971935008e-06,
+      "loss": 0.7599,
+      "step": 3046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.640915984622891e-06,
+      "loss": 0.5405,
+      "step": 3047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638848682985592e-06,
+      "loss": 1.0621,
+      "step": 3048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.636781067419113e-06,
+      "loss": 0.8337,
+      "step": 3049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6347131383195154e-06,
+      "loss": 0.9671,
+      "step": 3050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.632644896082918e-06,
+      "loss": 0.8296,
+      "step": 3051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.630576341105506e-06,
+      "loss": 1.1258,
+      "step": 3052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.628507473783518e-06,
+      "loss": 1.0634,
+      "step": 3053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.626438294513255e-06,
+      "loss": 0.6282,
+      "step": 3054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6243688036910815e-06,
+      "loss": 0.9479,
+      "step": 3055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.622299001713416e-06,
+      "loss": 0.992,
+      "step": 3056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.620228888976739e-06,
+      "loss": 1.2065,
+      "step": 3057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.618158465877589e-06,
+      "loss": 0.5198,
+      "step": 3058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.616087732812567e-06,
+      "loss": 0.58,
+      "step": 3059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.614016690178332e-06,
+      "loss": 0.9268,
+      "step": 3060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.611945338371601e-06,
+      "loss": 1.0038,
+      "step": 3061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.609873677789151e-06,
+      "loss": 0.8604,
+      "step": 3062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.607801708827819e-06,
+      "loss": 0.9578,
+      "step": 3063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6057294318845e-06,
+      "loss": 0.6524,
+      "step": 3064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.603656847356149e-06,
+      "loss": 0.7377,
+      "step": 3065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601583955639778e-06,
+      "loss": 1.0771,
+      "step": 3066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599510757132462e-06,
+      "loss": 0.9882,
+      "step": 3067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.597437252231328e-06,
+      "loss": 0.8029,
+      "step": 3068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.595363441333568e-06,
+      "loss": 0.8285,
+      "step": 3069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.59328932483643e-06,
+      "loss": 0.7621,
+      "step": 3070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.591214903137221e-06,
+      "loss": 1.1808,
+      "step": 3071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5891401766333056e-06,
+      "loss": 0.5839,
+      "step": 3072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.587065145722107e-06,
+      "loss": 0.9117,
+      "step": 3073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.584989810801108e-06,
+      "loss": 0.949,
+      "step": 3074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.582914172267848e-06,
+      "loss": 0.8074,
+      "step": 3075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.580838230519928e-06,
+      "loss": 0.6327,
+      "step": 3076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.578761985955001e-06,
+      "loss": 0.8457,
+      "step": 3077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.576685438970783e-06,
+      "loss": 0.8667,
+      "step": 3078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.574608589965046e-06,
+      "loss": 0.9586,
+      "step": 3079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.572531439335621e-06,
+      "loss": 1.081,
+      "step": 3080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.570453987480395e-06,
+      "loss": 0.5584,
+      "step": 3081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5683762347973134e-06,
+      "loss": 0.9533,
+      "step": 3082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.566298181684381e-06,
+      "loss": 0.61,
+      "step": 3083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5642198285396575e-06,
+      "loss": 0.93,
+      "step": 3084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562141175761261e-06,
+      "loss": 0.3524,
+      "step": 3085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5600622237473685e-06,
+      "loss": 0.8309,
+      "step": 3086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.557982972896212e-06,
+      "loss": 0.5532,
+      "step": 3087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.555903423606082e-06,
+      "loss": 0.8665,
+      "step": 3088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.553823576275327e-06,
+      "loss": 0.6764,
+      "step": 3089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551743431302348e-06,
+      "loss": 0.7791,
+      "step": 3090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.549662989085611e-06,
+      "loss": 1.0089,
+      "step": 3091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.547582250023633e-06,
+      "loss": 0.7986,
+      "step": 3092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545501214514989e-06,
+      "loss": 0.8272,
+      "step": 3093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.543419882958309e-06,
+      "loss": 0.6117,
+      "step": 3094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.541338255752286e-06,
+      "loss": 0.7163,
+      "step": 3095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.539256333295663e-06,
+      "loss": 0.9205,
+      "step": 3096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.537174115987242e-06,
+      "loss": 0.3127,
+      "step": 3097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5350916042258825e-06,
+      "loss": 1.2769,
+      "step": 3098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.533008798410499e-06,
+      "loss": 0.8461,
+      "step": 3099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.530925698940064e-06,
+      "loss": 0.6628,
+      "step": 3100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.528842306213603e-06,
+      "loss": 0.2516,
+      "step": 3101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.526758620630198e-06,
+      "loss": 0.8156,
+      "step": 3102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.524674642588994e-06,
+      "loss": 0.7267,
+      "step": 3103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.522590372489184e-06,
+      "loss": 0.5268,
+      "step": 3104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.520505810730018e-06,
+      "loss": 0.8225,
+      "step": 3105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.518420957710807e-06,
+      "loss": 0.9344,
+      "step": 3106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.516335813830912e-06,
+      "loss": 1.1682,
+      "step": 3107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.514250379489754e-06,
+      "loss": 0.801,
+      "step": 3108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.512164655086806e-06,
+      "loss": 0.8547,
+      "step": 3109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5100786410216e-06,
+      "loss": 1.0137,
+      "step": 3110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.507992337693719e-06,
+      "loss": 1.093,
+      "step": 3111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.505905745502807e-06,
+      "loss": 0.8768,
+      "step": 3112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50381886484856e-06,
+      "loss": 1.051,
+      "step": 3113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50173169613073e-06,
+      "loss": 0.8512,
+      "step": 3114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.499644239749122e-06,
+      "loss": 0.8696,
+      "step": 3115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.497556496103601e-06,
+      "loss": 1.0665,
+      "step": 3116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.495468465594081e-06,
+      "loss": 0.9944,
+      "step": 3117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.493380148620537e-06,
+      "loss": 0.7222,
+      "step": 3118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.491291545582994e-06,
+      "loss": 0.7581,
+      "step": 3119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.489202656881535e-06,
+      "loss": 0.6637,
+      "step": 3120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.487113482916296e-06,
+      "loss": 0.49,
+      "step": 3121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.485024024087468e-06,
+      "loss": 0.4891,
+      "step": 3122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.482934280795296e-06,
+      "loss": 0.7033,
+      "step": 3123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.480844253440082e-06,
+      "loss": 0.9751,
+      "step": 3124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.478753942422178e-06,
+      "loss": 0.8714,
+      "step": 3125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.476663348141996e-06,
+      "loss": 0.8859,
+      "step": 3126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.474572470999995e-06,
+      "loss": 1.3305,
+      "step": 3127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472481311396698e-06,
+      "loss": 0.2093,
+      "step": 3128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470389869732672e-06,
+      "loss": 0.6753,
+      "step": 3129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.468298146408544e-06,
+      "loss": 0.7916,
+      "step": 3130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466206141824993e-06,
+      "loss": 0.78,
+      "step": 3131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.464113856382752e-06,
+      "loss": 0.9735,
+      "step": 3132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.46202129048261e-06,
+      "loss": 0.7911,
+      "step": 3133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.459928444525406e-06,
+      "loss": 0.7888,
+      "step": 3134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.457835318912036e-06,
+      "loss": 0.8962,
+      "step": 3135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.455741914043446e-06,
+      "loss": 1.1529,
+      "step": 3136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.453648230320639e-06,
+      "loss": 0.9637,
+      "step": 3137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.451554268144671e-06,
+      "loss": 0.8621,
+      "step": 3138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.449460027916648e-06,
+      "loss": 0.794,
+      "step": 3139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.447365510037732e-06,
+      "loss": 0.7758,
+      "step": 3140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4452707149091385e-06,
+      "loss": 0.9647,
+      "step": 3141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.443175642932136e-06,
+      "loss": 1.0138,
+      "step": 3142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.441080294508044e-06,
+      "loss": 0.6818,
+      "step": 3143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.438984670038238e-06,
+      "loss": 0.7878,
+      "step": 3144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.436888769924142e-06,
+      "loss": 0.7589,
+      "step": 3145
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.434792594567238e-06,
+      "loss": 1.255,
+      "step": 3146
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.432696144369057e-06,
+      "loss": 0.569,
+      "step": 3147
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.430599419731186e-06,
+      "loss": 0.7756,
+      "step": 3148
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.42850242105526e-06,
+      "loss": 1.0288,
+      "step": 3149
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.42640514874297e-06,
+      "loss": 0.7838,
+      "step": 3150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.424307603196059e-06,
+      "loss": 1.1106,
+      "step": 3151
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.422209784816319e-06,
+      "loss": 0.8297,
+      "step": 3152
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4201116940056e-06,
+      "loss": 1.1144,
+      "step": 3153
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4180133311658e-06,
+      "loss": 0.8044,
+      "step": 3154
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.415914696698871e-06,
+      "loss": 0.8216,
+      "step": 3155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.413815791006816e-06,
+      "loss": 0.9165,
+      "step": 3156
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4117166144916885e-06,
+      "loss": 0.8195,
+      "step": 3157
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.409617167555597e-06,
+      "loss": 0.5033,
+      "step": 3158
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.407517450600702e-06,
+      "loss": 0.7343,
+      "step": 3159
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.405417464029211e-06,
+      "loss": 0.9366,
+      "step": 3160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.403317208243391e-06,
+      "loss": 0.5403,
+      "step": 3161
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40121668364555e-06,
+      "loss": 1.0722,
+      "step": 3162
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.399115890638058e-06,
+      "loss": 0.6672,
+      "step": 3163
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.397014829623329e-06,
+      "loss": 1.1353,
+      "step": 3164
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.394913501003835e-06,
+      "loss": 0.9819,
+      "step": 3165
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3928119051820895e-06,
+      "loss": 0.9913,
+      "step": 3166
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3907100425606675e-06,
+      "loss": 0.4429,
+      "step": 3167
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.388607913542189e-06,
+      "loss": 0.777,
+      "step": 3168
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.386505518529327e-06,
+      "loss": 1.1879,
+      "step": 3169
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.384402857924805e-06,
+      "loss": 0.7156,
+      "step": 3170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.382299932131399e-06,
+      "loss": 0.6837,
+      "step": 3171
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.380196741551933e-06,
+      "loss": 0.8444,
+      "step": 3172
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.378093286589282e-06,
+      "loss": 0.8267,
+      "step": 3173
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.375989567646375e-06,
+      "loss": 0.7613,
+      "step": 3174
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.373885585126187e-06,
+      "loss": 0.9121,
+      "step": 3175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.371781339431749e-06,
+      "loss": 0.6402,
+      "step": 3176
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.369676830966135e-06,
+      "loss": 0.7297,
+      "step": 3177
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3675720601324765e-06,
+      "loss": 0.9624,
+      "step": 3178
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3654670273339514e-06,
+      "loss": 0.9615,
+      "step": 3179
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3633617329737885e-06,
+      "loss": 1.3253,
+      "step": 3180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361256177455266e-06,
+      "loss": 0.4633,
+      "step": 3181
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3591503611817155e-06,
+      "loss": 1.0118,
+      "step": 3182
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.357044284556514e-06,
+      "loss": 0.5311,
+      "step": 3183
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.35493794798309e-06,
+      "loss": 0.4012,
+      "step": 3184
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3528313518649234e-06,
+      "loss": 0.4455,
+      "step": 3185
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.350724496605541e-06,
+      "loss": 0.9016,
+      "step": 3186
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.348617382608524e-06,
+      "loss": 0.8896,
+      "step": 3187
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.346510010277498e-06,
+      "loss": 0.6345,
+      "step": 3188
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.344402380016139e-06,
+      "loss": 0.6566,
+      "step": 3189
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.342294492228176e-06,
+      "loss": 0.7165,
+      "step": 3190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.340186347317383e-06,
+      "loss": 0.8914,
+      "step": 3191
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.338077945687585e-06,
+      "loss": 1.013,
+      "step": 3192
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.335969287742657e-06,
+      "loss": 0.6881,
+      "step": 3193
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.333860373886522e-06,
+      "loss": 1.104,
+      "step": 3194
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3317512045231535e-06,
+      "loss": 1.0786,
+      "step": 3195
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329641780056572e-06,
+      "loss": 0.6617,
+      "step": 3196
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.327532100890847e-06,
+      "loss": 0.4505,
+      "step": 3197
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.325422167430097e-06,
+      "loss": 0.6722,
+      "step": 3198
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.323311980078493e-06,
+      "loss": 1.1223,
+      "step": 3199
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.321201539240247e-06,
+      "loss": 1.0697,
+      "step": 3200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.319090845319628e-06,
+      "loss": 0.455,
+      "step": 3201
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.316979898720945e-06,
+      "loss": 0.7725,
+      "step": 3202
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.314868699848564e-06,
+      "loss": 0.8546,
+      "step": 3203
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.312757249106892e-06,
+      "loss": 0.8992,
+      "step": 3204
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.310645546900389e-06,
+      "loss": 0.7863,
+      "step": 3205
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.308533593633561e-06,
+      "loss": 0.9218,
+      "step": 3206
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.306421389710962e-06,
+      "loss": 0.4303,
+      "step": 3207
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.304308935537196e-06,
+      "loss": 0.6732,
+      "step": 3208
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.302196231516913e-06,
+      "loss": 0.7557,
+      "step": 3209
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3000832780548095e-06,
+      "loss": 0.6763,
+      "step": 3210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.297970075555634e-06,
+      "loss": 0.8033,
+      "step": 3211
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.29585662442418e-06,
+      "loss": 0.876,
+      "step": 3212
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.293742925065287e-06,
+      "loss": 0.7612,
+      "step": 3213
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.291628977883848e-06,
+      "loss": 0.9172,
+      "step": 3214
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.289514783284796e-06,
+      "loss": 1.0026,
+      "step": 3215
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.287400341673118e-06,
+      "loss": 0.6161,
+      "step": 3216
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.285285653453842e-06,
+      "loss": 0.708,
+      "step": 3217
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.283170719032048e-06,
+      "loss": 0.8488,
+      "step": 3218
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.281055538812861e-06,
+      "loss": 0.7009,
+      "step": 3219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.278940113201456e-06,
+      "loss": 1.2647,
+      "step": 3220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.276824442603051e-06,
+      "loss": 0.8993,
+      "step": 3221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.274708527422911e-06,
+      "loss": 0.5427,
+      "step": 3222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2725923680663525e-06,
+      "loss": 0.8551,
+      "step": 3223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.270475964938734e-06,
+      "loss": 0.8614,
+      "step": 3224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.268359318445464e-06,
+      "loss": 0.6384,
+      "step": 3225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.266242428991995e-06,
+      "loss": 0.7809,
+      "step": 3226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.264125296983826e-06,
+      "loss": 1.0398,
+      "step": 3227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.262007922826505e-06,
+      "loss": 0.5944,
+      "step": 3228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.259890306925627e-06,
+      "loss": 0.8025,
+      "step": 3229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.257772449686828e-06,
+      "loss": 0.5372,
+      "step": 3230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.255654351515793e-06,
+      "loss": 1.1244,
+      "step": 3231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.253536012818257e-06,
+      "loss": 0.416,
+      "step": 3232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251417433999995e-06,
+      "loss": 0.7026,
+      "step": 3233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.249298615466833e-06,
+      "loss": 0.9512,
+      "step": 3234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.24717955762464e-06,
+      "loss": 1.0774,
+      "step": 3235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2450602608793294e-06,
+      "loss": 0.9656,
+      "step": 3236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.242940725636865e-06,
+      "loss": 1.0549,
+      "step": 3237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.240820952303252e-06,
+      "loss": 0.5537,
+      "step": 3238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.238700941284545e-06,
+      "loss": 1.0096,
+      "step": 3239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23658069298684e-06,
+      "loss": 1.1171,
+      "step": 3240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.234460207816282e-06,
+      "loss": 0.919,
+      "step": 3241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2323394861790584e-06,
+      "loss": 1.1238,
+      "step": 3242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.230218528481406e-06,
+      "loss": 0.959,
+      "step": 3243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.228097335129602e-06,
+      "loss": 0.7687,
+      "step": 3244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2259759065299705e-06,
+      "loss": 0.7148,
+      "step": 3245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.223854243088883e-06,
+      "loss": 0.8018,
+      "step": 3246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.221732345212754e-06,
+      "loss": 0.5829,
+      "step": 3247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2196102133080415e-06,
+      "loss": 0.7972,
+      "step": 3248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2174878477812505e-06,
+      "loss": 1.0112,
+      "step": 3249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.215365249038932e-06,
+      "loss": 0.5183,
+      "step": 3250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.213242417487676e-06,
+      "loss": 0.7876,
+      "step": 3251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2111193535341245e-06,
+      "loss": 0.9611,
+      "step": 3252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.208996057584957e-06,
+      "loss": 0.7133,
+      "step": 3253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.206872530046904e-06,
+      "loss": 0.7122,
+      "step": 3254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.204748771326736e-06,
+      "loss": 1.1182,
+      "step": 3255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.202624781831269e-06,
+      "loss": 1.0443,
+      "step": 3256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.200500561967363e-06,
+      "loss": 0.8452,
+      "step": 3257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.198376112141922e-06,
+      "loss": 0.8332,
+      "step": 3258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.196251432761895e-06,
+      "loss": 1.0539,
+      "step": 3259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.194126524234274e-06,
+      "loss": 0.8953,
+      "step": 3260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.192001386966095e-06,
+      "loss": 0.795,
+      "step": 3261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.189876021364439e-06,
+      "loss": 0.5388,
+      "step": 3262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.18775042783643e-06,
+      "loss": 0.3344,
+      "step": 3263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.185624606789233e-06,
+      "loss": 0.91,
+      "step": 3264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.183498558630062e-06,
+      "loss": 1.0729,
+      "step": 3265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.181372283766169e-06,
+      "loss": 0.5767,
+      "step": 3266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.179245782604854e-06,
+      "loss": 0.6654,
+      "step": 3267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.177119055553456e-06,
+      "loss": 0.7875,
+      "step": 3268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1749921030193615e-06,
+      "loss": 1.1068,
+      "step": 3269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.172864925409999e-06,
+      "loss": 0.7586,
+      "step": 3270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.170737523132838e-06,
+      "loss": 0.8196,
+      "step": 3271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.168609896595392e-06,
+      "loss": 0.6193,
+      "step": 3272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.166482046205219e-06,
+      "loss": 0.8485,
+      "step": 3273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.164353972369917e-06,
+      "loss": 0.8852,
+      "step": 3274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.162225675497131e-06,
+      "loss": 0.9641,
+      "step": 3275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.160097155994546e-06,
+      "loss": 0.7216,
+      "step": 3276
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.157968414269887e-06,
+      "loss": 0.5526,
+      "step": 3277
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.155839450730929e-06,
+      "loss": 0.6979,
+      "step": 3278
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153710265785479e-06,
+      "loss": 0.7846,
+      "step": 3279
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.151580859841399e-06,
+      "loss": 1.056,
+      "step": 3280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1494512333065825e-06,
+      "loss": 0.5526,
+      "step": 3281
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1473213865889715e-06,
+      "loss": 0.7666,
+      "step": 3282
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.145191320096546e-06,
+      "loss": 0.8472,
+      "step": 3283
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.143061034237332e-06,
+      "loss": 0.7966,
+      "step": 3284
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.140930529419396e-06,
+      "loss": 0.6942,
+      "step": 3285
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.138799806050845e-06,
+      "loss": 0.7978,
+      "step": 3286
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.13666886453983e-06,
+      "loss": 0.941,
+      "step": 3287
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.134537705294542e-06,
+      "loss": 0.5885,
+      "step": 3288
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1324063287232145e-06,
+      "loss": 0.8364,
+      "step": 3289
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.130274735234125e-06,
+      "loss": 0.7203,
+      "step": 3290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.128142925235588e-06,
+      "loss": 1.004,
+      "step": 3291
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1260108991359625e-06,
+      "loss": 0.95,
+      "step": 3292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.123878657343648e-06,
+      "loss": 0.3509,
+      "step": 3293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.121746200267086e-06,
+      "loss": 1.0815,
+      "step": 3294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.119613528314757e-06,
+      "loss": 0.9154,
+      "step": 3295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.117480641895186e-06,
+      "loss": 0.4318,
+      "step": 3296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.115347541416936e-06,
+      "loss": 0.871,
+      "step": 3297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.113214227288614e-06,
+      "loss": 0.3554,
+      "step": 3298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.111080699918867e-06,
+      "loss": 0.7364,
+      "step": 3299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.10894695971638e-06,
+      "loss": 1.1867,
+      "step": 3300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.106813007089881e-06,
+      "loss": 0.9262,
+      "step": 3301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.104678842448141e-06,
+      "loss": 0.7612,
+      "step": 3302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.102544466199968e-06,
+      "loss": 0.811,
+      "step": 3303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.100409878754211e-06,
+      "loss": 0.8161,
+      "step": 3304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.098275080519761e-06,
+      "loss": 0.6159,
+      "step": 3305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0961400719055495e-06,
+      "loss": 0.9412,
+      "step": 3306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.094004853320545e-06,
+      "loss": 0.8821,
+      "step": 3307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.091869425173762e-06,
+      "loss": 0.988,
+      "step": 3308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.08973378787425e-06,
+      "loss": 0.5411,
+      "step": 3309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.087597941831099e-06,
+      "loss": 0.9777,
+      "step": 3310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.085461887453443e-06,
+      "loss": 0.9065,
+      "step": 3311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0833256251504525e-06,
+      "loss": 0.4354,
+      "step": 3312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.081189155331337e-06,
+      "loss": 0.9278,
+      "step": 3313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.07905247840535e-06,
+      "loss": 0.5815,
+      "step": 3314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.076915594781779e-06,
+      "loss": 0.7832,
+      "step": 3315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.074778504869956e-06,
+      "loss": 0.8032,
+      "step": 3316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.072641209079251e-06,
+      "loss": 0.4006,
+      "step": 3317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.070503707819072e-06,
+      "loss": 0.582,
+      "step": 3318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.068366001498868e-06,
+      "loss": 0.9183,
+      "step": 3319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.066228090528126e-06,
+      "loss": 1.1825,
+      "step": 3320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.064089975316374e-06,
+      "loss": 0.864,
+      "step": 3321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.061951656273178e-06,
+      "loss": 0.7392,
+      "step": 3322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.059813133808142e-06,
+      "loss": 1.0882,
+      "step": 3323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.057674408330911e-06,
+      "loss": 0.6762,
+      "step": 3324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0555354802511654e-06,
+      "loss": 0.7622,
+      "step": 3325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.053396349978632e-06,
+      "loss": 0.6258,
+      "step": 3326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.051257017923067e-06,
+      "loss": 0.7646,
+      "step": 3327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.049117484494272e-06,
+      "loss": 0.8723,
+      "step": 3328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.046977750102083e-06,
+      "loss": 0.944,
+      "step": 3329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.044837815156377e-06,
+      "loss": 1.0538,
+      "step": 3330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.042697680067068e-06,
+      "loss": 0.9444,
+      "step": 3331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.040557345244109e-06,
+      "loss": 0.9447,
+      "step": 3332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0384168110974915e-06,
+      "loss": 0.6053,
+      "step": 3333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.036276078037246e-06,
+      "loss": 0.705,
+      "step": 3334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.034135146473439e-06,
+      "loss": 0.6775,
+      "step": 3335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.031994016816175e-06,
+      "loss": 0.4988,
+      "step": 3336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.029852689475598e-06,
+      "loss": 0.6426,
+      "step": 3337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.027711164861891e-06,
+      "loss": 1.0553,
+      "step": 3338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.025569443385272e-06,
+      "loss": 0.6168,
+      "step": 3339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.023427525455997e-06,
+      "loss": 0.6746,
+      "step": 3340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.021285411484361e-06,
+      "loss": 0.8719,
+      "step": 3341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.019143101880697e-06,
+      "loss": 0.7868,
+      "step": 3342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.017000597055374e-06,
+      "loss": 0.6167,
+      "step": 3343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0148578974187975e-06,
+      "loss": 0.864,
+      "step": 3344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.012715003381413e-06,
+      "loss": 0.7512,
+      "step": 3345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.010571915353703e-06,
+      "loss": 0.8385,
+      "step": 3346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.008428633746184e-06,
+      "loss": 0.7788,
+      "step": 3347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0062851589694116e-06,
+      "loss": 0.9313,
+      "step": 3348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.00414149143398e-06,
+      "loss": 1.0337,
+      "step": 3349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.001997631550519e-06,
+      "loss": 0.8098,
+      "step": 3350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9998535797296945e-06,
+      "loss": 0.846,
+      "step": 3351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.997709336382207e-06,
+      "loss": 0.2425,
+      "step": 3352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9955649019188e-06,
+      "loss": 0.6629,
+      "step": 3353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9934202767502505e-06,
+      "loss": 1.0085,
+      "step": 3354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.991275461287369e-06,
+      "loss": 0.8398,
+      "step": 3355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.989130455941005e-06,
+      "loss": 1.1155,
+      "step": 3356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.986985261122047e-06,
+      "loss": 0.5931,
+      "step": 3357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.984839877241415e-06,
+      "loss": 0.6612,
+      "step": 3358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9826943047100686e-06,
+      "loss": 0.7199,
+      "step": 3359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.980548543939e-06,
+      "loss": 0.8117,
+      "step": 3360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.978402595339242e-06,
+      "loss": 0.853,
+      "step": 3361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.976256459321861e-06,
+      "loss": 0.9753,
+      "step": 3362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.974110136297958e-06,
+      "loss": 0.6869,
+      "step": 3363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.971963626678674e-06,
+      "loss": 0.7833,
+      "step": 3364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.96981693087518e-06,
+      "loss": 0.6218,
+      "step": 3365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.967670049298687e-06,
+      "loss": 0.7831,
+      "step": 3366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.965522982360441e-06,
+      "loss": 1.1306,
+      "step": 3367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.963375730471721e-06,
+      "loss": 0.9727,
+      "step": 3368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.961228294043845e-06,
+      "loss": 1.1044,
+      "step": 3369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.959080673488164e-06,
+      "loss": 0.8303,
+      "step": 3370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.956932869216065e-06,
+      "loss": 0.8284,
+      "step": 3371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9547848816389684e-06,
+      "loss": 0.6681,
+      "step": 3372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.952636711168333e-06,
+      "loss": 1.2332,
+      "step": 3373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.950488358215651e-06,
+      "loss": 0.7531,
+      "step": 3374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.948339823192448e-06,
+      "loss": 0.7794,
+      "step": 3375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.946191106510289e-06,
+      "loss": 0.9956,
+      "step": 3376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944042208580767e-06,
+      "loss": 1.2032,
+      "step": 3377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.941893129815517e-06,
+      "loss": 0.5506,
+      "step": 3378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.939743870626202e-06,
+      "loss": 0.7746,
+      "step": 3379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.937594431424524e-06,
+      "loss": 1.0048,
+      "step": 3380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93544481262222e-06,
+      "loss": 1.0553,
+      "step": 3381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.933295014631058e-06,
+      "loss": 0.774,
+      "step": 3382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.931145037862841e-06,
+      "loss": 0.5438,
+      "step": 3383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.928994882729408e-06,
+      "loss": 1.1266,
+      "step": 3384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.926844549642633e-06,
+      "loss": 0.9332,
+      "step": 3385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.92469403901442e-06,
+      "loss": 0.6441,
+      "step": 3386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.922543351256711e-06,
+      "loss": 0.9409,
+      "step": 3387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.920392486781479e-06,
+      "loss": 0.5556,
+      "step": 3388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.918241446000733e-06,
+      "loss": 0.7833,
+      "step": 3389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.916090229326515e-06,
+      "loss": 0.8215,
+      "step": 3390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9139388371709e-06,
+      "loss": 0.5959,
+      "step": 3391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.911787269946e-06,
+      "loss": 1.1724,
+      "step": 3392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.909635528063954e-06,
+      "loss": 0.9582,
+      "step": 3393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90748361193694e-06,
+      "loss": 0.5473,
+      "step": 3394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.905331521977169e-06,
+      "loss": 0.9348,
+      "step": 3395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.903179258596882e-06,
+      "loss": 0.9591,
+      "step": 3396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.901026822208356e-06,
+      "loss": 0.8309,
+      "step": 3397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8988742132238995e-06,
+      "loss": 0.7416,
+      "step": 3398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.896721432055856e-06,
+      "loss": 0.9058,
+      "step": 3399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8945684791166e-06,
+      "loss": 0.7437,
+      "step": 3400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.892415354818541e-06,
+      "loss": 0.8582,
+      "step": 3401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.890262059574119e-06,
+      "loss": 0.6532,
+      "step": 3402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.888108593795808e-06,
+      "loss": 0.9304,
+      "step": 3403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.885954957896115e-06,
+      "loss": 0.7833,
+      "step": 3404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.883801152287578e-06,
+      "loss": 0.6714,
+      "step": 3405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.88164717738277e-06,
+      "loss": 0.7905,
+      "step": 3406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.879493033594294e-06,
+      "loss": 0.9144,
+      "step": 3407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.877338721334785e-06,
+      "loss": 1.0118,
+      "step": 3408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.875184241016916e-06,
+      "loss": 0.8239,
+      "step": 3409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8730295930533855e-06,
+      "loss": 1.115,
+      "step": 3410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.870874777856924e-06,
+      "loss": 0.8523,
+      "step": 3411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.868719795840301e-06,
+      "loss": 1.01,
+      "step": 3412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.86656464741631e-06,
+      "loss": 0.6411,
+      "step": 3413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.864409332997782e-06,
+      "loss": 0.5623,
+      "step": 3414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.862253852997575e-06,
+      "loss": 0.7146,
+      "step": 3415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.860098207828584e-06,
+      "loss": 0.8184,
+      "step": 3416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.857942397903732e-06,
+      "loss": 0.6394,
+      "step": 3417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.855786423635974e-06,
+      "loss": 0.7904,
+      "step": 3418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.853630285438298e-06,
+      "loss": 0.8895,
+      "step": 3419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.851473983723721e-06,
+      "loss": 0.8926,
+      "step": 3420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8493175189052955e-06,
+      "loss": 0.8901,
+      "step": 3421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.847160891396098e-06,
+      "loss": 0.8813,
+      "step": 3422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8450041016092465e-06,
+      "loss": 1.2053,
+      "step": 3423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.842847149957879e-06,
+      "loss": 0.9839,
+      "step": 3424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.840690036855173e-06,
+      "loss": 0.3284,
+      "step": 3425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838532762714331e-06,
+      "loss": 0.8172,
+      "step": 3426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.836375327948591e-06,
+      "loss": 1.1563,
+      "step": 3427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.834217732971219e-06,
+      "loss": 0.9558,
+      "step": 3428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.832059978195514e-06,
+      "loss": 1.0051,
+      "step": 3429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.829902064034802e-06,
+      "loss": 0.8082,
+      "step": 3430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.827743990902442e-06,
+      "loss": 0.7338,
+      "step": 3431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.825585759211825e-06,
+      "loss": 0.6712,
+      "step": 3432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.823427369376368e-06,
+      "loss": 1.1173,
+      "step": 3433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.821268821809521e-06,
+      "loss": 0.8635,
+      "step": 3434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.819110116924766e-06,
+      "loss": 0.5591,
+      "step": 3435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.816951255135611e-06,
+      "loss": 1.1018,
+      "step": 3436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8147922368555965e-06,
+      "loss": 1.1171,
+      "step": 3437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.812633062498294e-06,
+      "loss": 1.0217,
+      "step": 3438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.810473732477302e-06,
+      "loss": 0.3328,
+      "step": 3439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8083142472062505e-06,
+      "loss": 1.0215,
+      "step": 3440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.806154607098799e-06,
+      "loss": 0.6072,
+      "step": 3441
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803994812568639e-06,
+      "loss": 1.0178,
+      "step": 3442
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.801834864029486e-06,
+      "loss": 0.6823,
+      "step": 3443
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.799674761895091e-06,
+      "loss": 0.7832,
+      "step": 3444
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.797514506579229e-06,
+      "loss": 0.6105,
+      "step": 3445
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.795354098495709e-06,
+      "loss": 0.995,
+      "step": 3446
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.793193538058367e-06,
+      "loss": 0.51,
+      "step": 3447
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.791032825681069e-06,
+      "loss": 0.9957,
+      "step": 3448
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.788871961777709e-06,
+      "loss": 0.8544,
+      "step": 3449
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.78671094676221e-06,
+      "loss": 0.8158,
+      "step": 3450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.784549781048525e-06,
+      "loss": 0.7633,
+      "step": 3451
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.782388465050637e-06,
+      "loss": 1.0915,
+      "step": 3452
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.780226999182555e-06,
+      "loss": 0.7828,
+      "step": 3453
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.778065383858316e-06,
+      "loss": 0.8414,
+      "step": 3454
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.775903619491991e-06,
+      "loss": 0.8438,
+      "step": 3455
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.773741706497676e-06,
+      "loss": 0.652,
+      "step": 3456
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.771579645289493e-06,
+      "loss": 1.1254,
+      "step": 3457
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.769417436281596e-06,
+      "loss": 0.7012,
+      "step": 3458
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.767255079888167e-06,
+      "loss": 1.1849,
+      "step": 3459
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.765092576523416e-06,
+      "loss": 0.9278,
+      "step": 3460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.762929926601578e-06,
+      "loss": 0.8588,
+      "step": 3461
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7607671305369214e-06,
+      "loss": 0.5677,
+      "step": 3462
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.75860418874374e-06,
+      "loss": 0.9011,
+      "step": 3463
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.756441101636354e-06,
+      "loss": 0.9353,
+      "step": 3464
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7542778696291144e-06,
+      "loss": 0.6528,
+      "step": 3465
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752114493136397e-06,
+      "loss": 0.6603,
+      "step": 3466
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.749950972572609e-06,
+      "loss": 0.9581,
+      "step": 3467
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.74778730835218e-06,
+      "loss": 0.737,
+      "step": 3468
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7456235008895724e-06,
+      "loss": 0.6693,
+      "step": 3469
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743459550599272e-06,
+      "loss": 1.2032,
+      "step": 3470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7412954578957944e-06,
+      "loss": 1.1323,
+      "step": 3471
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.739131223193683e-06,
+      "loss": 0.6414,
+      "step": 3472
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.736966846907505e-06,
+      "loss": 0.9097,
+      "step": 3473
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.734802329451857e-06,
+      "loss": 0.7939,
+      "step": 3474
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.732637671241365e-06,
+      "loss": 1.17,
+      "step": 3475
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.730472872690678e-06,
+      "loss": 0.6447,
+      "step": 3476
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.728307934214472e-06,
+      "loss": 0.6857,
+      "step": 3477
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.726142856227453e-06,
+      "loss": 0.8871,
+      "step": 3478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.72397763914435e-06,
+      "loss": 0.6591,
+      "step": 3479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.721812283379923e-06,
+      "loss": 0.9951,
+      "step": 3480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.719646789348955e-06,
+      "loss": 1.0548,
+      "step": 3481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.717481157466256e-06,
+      "loss": 0.2928,
+      "step": 3482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7153153881466626e-06,
+      "loss": 0.9013,
+      "step": 3483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.713149481805039e-06,
+      "loss": 0.8727,
+      "step": 3484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710983438856276e-06,
+      "loss": 0.5993,
+      "step": 3485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.708817259715287e-06,
+      "loss": 0.7407,
+      "step": 3486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7066509447970155e-06,
+      "loss": 0.9053,
+      "step": 3487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7044844945164276e-06,
+      "loss": 0.8915,
+      "step": 3488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.702317909288518e-06,
+      "loss": 0.4407,
+      "step": 3489
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.700151189528308e-06,
+      "loss": 0.7762,
+      "step": 3490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697984335650841e-06,
+      "loss": 0.6953,
+      "step": 3491
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.695817348071185e-06,
+      "loss": 0.9384,
+      "step": 3492
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6936502272044435e-06,
+      "loss": 0.9328,
+      "step": 3493
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.691482973465734e-06,
+      "loss": 0.8834,
+      "step": 3494
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.689315587270207e-06,
+      "loss": 0.6228,
+      "step": 3495
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6871480690330325e-06,
+      "loss": 0.6634,
+      "step": 3496
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.68498041916941e-06,
+      "loss": 1.1616,
+      "step": 3497
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.682812638094565e-06,
+      "loss": 1.0264,
+      "step": 3498
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.680644726223743e-06,
+      "loss": 0.9449,
+      "step": 3499
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.67847668397222e-06,
+      "loss": 0.9566,
+      "step": 3500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6763085117552944e-06,
+      "loss": 0.3592,
+      "step": 3501
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6741402099882885e-06,
+      "loss": 0.9992,
+      "step": 3502
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.671971779086551e-06,
+      "loss": 0.5012,
+      "step": 3503
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.669803219465456e-06,
+      "loss": 1.0512,
+      "step": 3504
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.667634531540402e-06,
+      "loss": 0.4749,
+      "step": 3505
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.665465715726809e-06,
+      "loss": 0.3078,
+      "step": 3506
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.663296772440125e-06,
+      "loss": 0.737,
+      "step": 3507
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.661127702095822e-06,
+      "loss": 0.9498,
+      "step": 3508
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.658958505109395e-06,
+      "loss": 0.9848,
+      "step": 3509
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6567891818963635e-06,
+      "loss": 1.1014,
+      "step": 3510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.654619732872273e-06,
+      "loss": 0.9489,
+      "step": 3511
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65245015845269e-06,
+      "loss": 0.9346,
+      "step": 3512
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6502804590532075e-06,
+      "loss": 0.7694,
+      "step": 3513
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.648110635089442e-06,
+      "loss": 0.371,
+      "step": 3514
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.645940686977033e-06,
+      "loss": 0.9551,
+      "step": 3515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.643770615131644e-06,
+      "loss": 0.747,
+      "step": 3516
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.641600419968962e-06,
+      "loss": 0.623,
+      "step": 3517
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6394301019047e-06,
+      "loss": 0.9546,
+      "step": 3518
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.63725966135459e-06,
+      "loss": 0.6513,
+      "step": 3519
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.635089098734394e-06,
+      "loss": 0.8315,
+      "step": 3520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.632918414459891e-06,
+      "loss": 0.8781,
+      "step": 3521
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630747608946884e-06,
+      "loss": 0.8412,
+      "step": 3522
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628576682611204e-06,
+      "loss": 0.3257,
+      "step": 3523
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.626405635868701e-06,
+      "loss": 0.7868,
+      "step": 3524
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.624234469135249e-06,
+      "loss": 0.8378,
+      "step": 3525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.622063182826747e-06,
+      "loss": 0.7194,
+      "step": 3526
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.619891777359112e-06,
+      "loss": 0.84,
+      "step": 3527
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6177202531482885e-06,
+      "loss": 0.8263,
+      "step": 3528
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.615548610610244e-06,
+      "loss": 0.7184,
+      "step": 3529
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.613376850160964e-06,
+      "loss": 0.7453,
+      "step": 3530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6112049722164605e-06,
+      "loss": 0.6489,
+      "step": 3531
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.609032977192767e-06,
+      "loss": 0.8841,
+      "step": 3532
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.60686086550594e-06,
+      "loss": 0.6709,
+      "step": 3533
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6046886375720565e-06,
+      "loss": 0.8535,
+      "step": 3534
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.602516293807219e-06,
+      "loss": 0.9338,
+      "step": 3535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.600343834627546e-06,
+      "loss": 0.6731,
+      "step": 3536
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.598171260449188e-06,
+      "loss": 0.8529,
+      "step": 3537
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.595998571688307e-06,
+      "loss": 0.9246,
+      "step": 3538
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.593825768761094e-06,
+      "loss": 0.7449,
+      "step": 3539
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5916528520837595e-06,
+      "loss": 0.8822,
+      "step": 3540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.589479822072535e-06,
+      "loss": 1.0241,
+      "step": 3541
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.587306679143677e-06,
+      "loss": 0.8908,
+      "step": 3542
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.585133423713458e-06,
+      "loss": 0.8316,
+      "step": 3543
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5829600561981774e-06,
+      "loss": 1.3939,
+      "step": 3544
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.580786577014153e-06,
+      "loss": 0.8299,
+      "step": 3545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.578612986577727e-06,
+      "loss": 0.9082,
+      "step": 3546
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.576439285305258e-06,
+      "loss": 0.6772,
+      "step": 3547
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.57426547361313e-06,
+      "loss": 0.343,
+      "step": 3548
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.572091551917747e-06,
+      "loss": 1.0833,
+      "step": 3549
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.569917520635534e-06,
+      "loss": 0.8828,
+      "step": 3550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.567743380182937e-06,
+      "loss": 0.6446,
+      "step": 3551
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5655691309764225e-06,
+      "loss": 1.1936,
+      "step": 3552
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.563394773432478e-06,
+      "loss": 0.7068,
+      "step": 3553
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.561220307967612e-06,
+      "loss": 0.8652,
+      "step": 3554
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.559045734998354e-06,
+      "loss": 0.9688,
+      "step": 3555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.556871054941254e-06,
+      "loss": 0.6164,
+      "step": 3556
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.554696268212881e-06,
+      "loss": 0.7263,
+      "step": 3557
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.552521375229827e-06,
+      "loss": 0.8169,
+      "step": 3558
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.550346376408701e-06,
+      "loss": 0.7714,
+      "step": 3559
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.548171272166137e-06,
+      "loss": 0.7213,
+      "step": 3560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.545996062918784e-06,
+      "loss": 0.9176,
+      "step": 3561
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.543820749083315e-06,
+      "loss": 0.7634,
+      "step": 3562
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.541645331076422e-06,
+      "loss": 0.9891,
+      "step": 3563
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.539469809314814e-06,
+      "loss": 0.801,
+      "step": 3564
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.537294184215225e-06,
+      "loss": 0.7629,
+      "step": 3565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.535118456194406e-06,
+      "loss": 0.7745,
+      "step": 3566
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.532942625669126e-06,
+      "loss": 0.8227,
+      "step": 3567
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5307666930561784e-06,
+      "loss": 1.0413,
+      "step": 3568
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.528590658772373e-06,
+      "loss": 0.6133,
+      "step": 3569
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.526414523234535e-06,
+      "loss": 0.8744,
+      "step": 3570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.524238286859519e-06,
+      "loss": 0.6667,
+      "step": 3571
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.52206195006419e-06,
+      "loss": 0.8305,
+      "step": 3572
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.519885513265439e-06,
+      "loss": 0.7708,
+      "step": 3573
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.517708976880167e-06,
+      "loss": 0.7597,
+      "step": 3574
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.515532341325303e-06,
+      "loss": 1.0884,
+      "step": 3575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.513355607017794e-06,
+      "loss": 0.6254,
+      "step": 3576
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.511178774374601e-06,
+      "loss": 0.7388,
+      "step": 3577
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.509001843812708e-06,
+      "loss": 0.7288,
+      "step": 3578
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5068248157491145e-06,
+      "loss": 0.8435,
+      "step": 3579
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.504647690600842e-06,
+      "loss": 0.7903,
+      "step": 3580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.502470468784927e-06,
+      "loss": 0.7256,
+      "step": 3581
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50029315071843e-06,
+      "loss": 0.6542,
+      "step": 3582
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.498115736818424e-06,
+      "loss": 0.7183,
+      "step": 3583
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.495938227502003e-06,
+      "loss": 0.582,
+      "step": 3584
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.493760623186279e-06,
+      "loss": 0.7126,
+      "step": 3585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491582924288384e-06,
+      "loss": 0.6359,
+      "step": 3586
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4894051312254645e-06,
+      "loss": 0.7399,
+      "step": 3587
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4872272444146894e-06,
+      "loss": 0.7021,
+      "step": 3588
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.485049264273241e-06,
+      "loss": 0.6393,
+      "step": 3589
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4828711912183234e-06,
+      "loss": 1.0527,
+      "step": 3590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.480693025667154e-06,
+      "loss": 0.8637,
+      "step": 3591
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478514768036974e-06,
+      "loss": 0.6813,
+      "step": 3592
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.476336418745036e-06,
+      "loss": 0.3798,
+      "step": 3593
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4741579782086164e-06,
+      "loss": 0.8465,
+      "step": 3594
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.471979446845001e-06,
+      "loss": 0.959,
+      "step": 3595
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.469800825071503e-06,
+      "loss": 0.8309,
+      "step": 3596
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.467622113305444e-06,
+      "loss": 0.6992,
+      "step": 3597
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.465443311964167e-06,
+      "loss": 0.8301,
+      "step": 3598
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.463264421465033e-06,
+      "loss": 0.565,
+      "step": 3599
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.461085442225419e-06,
+      "loss": 1.1884,
+      "step": 3600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.458906374662716e-06,
+      "loss": 0.4374,
+      "step": 3601
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.456727219194338e-06,
+      "loss": 0.9666,
+      "step": 3602
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.454547976237708e-06,
+      "loss": 0.7079,
+      "step": 3603
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4523686462102756e-06,
+      "loss": 0.4863,
+      "step": 3604
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.450189229529499e-06,
+      "loss": 0.8359,
+      "step": 3605
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.448009726612856e-06,
+      "loss": 0.5015,
+      "step": 3606
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44583013787784e-06,
+      "loss": 0.7807,
+      "step": 3607
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.443650463741963e-06,
+      "loss": 0.7708,
+      "step": 3608
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.441470704622751e-06,
+      "loss": 0.4798,
+      "step": 3609
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.439290860937746e-06,
+      "loss": 0.9762,
+      "step": 3610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.437110933104509e-06,
+      "loss": 0.68,
+      "step": 3611
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4349309215406155e-06,
+      "loss": 1.0761,
+      "step": 3612
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.432750826663655e-06,
+      "loss": 0.789,
+      "step": 3613
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.430570648891238e-06,
+      "loss": 1.1023,
+      "step": 3614
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.428390388640985e-06,
+      "loss": 0.9043,
+      "step": 3615
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.426210046330536e-06,
+      "loss": 0.7536,
+      "step": 3616
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4240296223775465e-06,
+      "loss": 0.736,
+      "step": 3617
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.421849117199687e-06,
+      "loss": 0.8027,
+      "step": 3618
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.419668531214641e-06,
+      "loss": 0.7295,
+      "step": 3619
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.417487864840114e-06,
+      "loss": 0.8803,
+      "step": 3620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.415307118493821e-06,
+      "loss": 1.13,
+      "step": 3621
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.413126292593494e-06,
+      "loss": 0.5747,
+      "step": 3622
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.41094538755688e-06,
+      "loss": 0.8393,
+      "step": 3623
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.408764403801742e-06,
+      "loss": 0.9711,
+      "step": 3624
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.406583341745858e-06,
+      "loss": 0.7665,
+      "step": 3625
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.404402201807022e-06,
+      "loss": 1.0699,
+      "step": 3626
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.402220984403039e-06,
+      "loss": 0.4573,
+      "step": 3627
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.400039689951734e-06,
+      "loss": 0.6449,
+      "step": 3628
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.397858318870942e-06,
+      "loss": 0.5112,
+      "step": 3629
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.395676871578515e-06,
+      "loss": 0.4881,
+      "step": 3630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3934953484923216e-06,
+      "loss": 0.8737,
+      "step": 3631
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.391313750030241e-06,
+      "loss": 0.6474,
+      "step": 3632
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.389132076610171e-06,
+      "loss": 0.6945,
+      "step": 3633
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.386950328650018e-06,
+      "loss": 0.8998,
+      "step": 3634
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.384768506567708e-06,
+      "loss": 0.88,
+      "step": 3635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.38258661078118e-06,
+      "loss": 0.5811,
+      "step": 3636
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.380404641708384e-06,
+      "loss": 0.9226,
+      "step": 3637
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37822259976729e-06,
+      "loss": 0.9016,
+      "step": 3638
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.376040485375875e-06,
+      "loss": 0.4145,
+      "step": 3639
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.373858298952135e-06,
+      "loss": 0.7696,
+      "step": 3640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.371676040914077e-06,
+      "loss": 1.1316,
+      "step": 3641
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.369493711679724e-06,
+      "loss": 0.8807,
+      "step": 3642
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.367311311667112e-06,
+      "loss": 1.1149,
+      "step": 3643
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3651288412942885e-06,
+      "loss": 0.7872,
+      "step": 3644
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.362946300979315e-06,
+      "loss": 0.538,
+      "step": 3645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.36076369114027e-06,
+      "loss": 0.8739,
+      "step": 3646
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.358581012195241e-06,
+      "loss": 0.9028,
+      "step": 3647
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.356398264562331e-06,
+      "loss": 0.6329,
+      "step": 3648
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.354215448659655e-06,
+      "loss": 0.861,
+      "step": 3649
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.352032564905344e-06,
+      "loss": 0.8971,
+      "step": 3650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.349849613717536e-06,
+      "loss": 0.9339,
+      "step": 3651
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347666595514388e-06,
+      "loss": 0.7463,
+      "step": 3652
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.345483510714067e-06,
+      "loss": 0.2396,
+      "step": 3653
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.343300359734753e-06,
+      "loss": 0.6536,
+      "step": 3654
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.341117142994639e-06,
+      "loss": 0.7873,
+      "step": 3655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.338933860911929e-06,
+      "loss": 0.6021,
+      "step": 3656
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.336750513904844e-06,
+      "loss": 0.5971,
+      "step": 3657
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.334567102391611e-06,
+      "loss": 0.9301,
+      "step": 3658
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.332383626790476e-06,
+      "loss": 0.67,
+      "step": 3659
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.330200087519691e-06,
+      "loss": 0.7506,
+      "step": 3660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.328016484997526e-06,
+      "loss": 1.1184,
+      "step": 3661
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.325832819642258e-06,
+      "loss": 0.8611,
+      "step": 3662
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.323649091872179e-06,
+      "loss": 0.7618,
+      "step": 3663
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.321465302105594e-06,
+      "loss": 0.2169,
+      "step": 3664
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.319281450760816e-06,
+      "loss": 1.0457,
+      "step": 3665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.317097538256173e-06,
+      "loss": 0.5284,
+      "step": 3666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.314913565010005e-06,
+      "loss": 0.9955,
+      "step": 3667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312729531440659e-06,
+      "loss": 0.9308,
+      "step": 3668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3105454379665e-06,
+      "loss": 0.6918,
+      "step": 3669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3083612850059e-06,
+      "loss": 0.9864,
+      "step": 3670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.306177072977245e-06,
+      "loss": 0.8139,
+      "step": 3671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.303992802298928e-06,
+      "loss": 0.7411,
+      "step": 3672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.30180847338936e-06,
+      "loss": 0.7235,
+      "step": 3673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.299624086666959e-06,
+      "loss": 0.8497,
+      "step": 3674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.297439642550153e-06,
+      "loss": 0.6534,
+      "step": 3675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2952551414573814e-06,
+      "loss": 1.1876,
+      "step": 3676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2930705838070995e-06,
+      "loss": 1.108,
+      "step": 3677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.290885970017766e-06,
+      "loss": 1.0977,
+      "step": 3678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.288701300507857e-06,
+      "loss": 0.6941,
+      "step": 3679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2865165756958535e-06,
+      "loss": 0.4605,
+      "step": 3680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.284331796000251e-06,
+      "loss": 0.5174,
+      "step": 3681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.282146961839555e-06,
+      "loss": 1.1223,
+      "step": 3682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.27996207363228e-06,
+      "loss": 0.6109,
+      "step": 3683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.27777713179695e-06,
+      "loss": 1.053,
+      "step": 3684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2755921367521045e-06,
+      "loss": 0.6208,
+      "step": 3685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.273407088916287e-06,
+      "loss": 0.9194,
+      "step": 3686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.271221988708054e-06,
+      "loss": 1.0007,
+      "step": 3687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.269036836545973e-06,
+      "loss": 0.6795,
+      "step": 3688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.26685163284862e-06,
+      "loss": 0.7268,
+      "step": 3689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.264666378034579e-06,
+      "loss": 0.8304,
+      "step": 3690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.262481072522448e-06,
+      "loss": 0.6145,
+      "step": 3691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2602957167308314e-06,
+      "loss": 0.7231,
+      "step": 3692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.258110311078345e-06,
+      "loss": 0.5715,
+      "step": 3693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.255924855983614e-06,
+      "loss": 0.8928,
+      "step": 3694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.253739351865272e-06,
+      "loss": 0.7482,
+      "step": 3695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251553799141961e-06,
+      "loss": 0.8112,
+      "step": 3696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.249368198232336e-06,
+      "loss": 0.6425,
+      "step": 3697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.247182549555059e-06,
+      "loss": 1.0517,
+      "step": 3698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.244996853528801e-06,
+      "loss": 0.728,
+      "step": 3699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.242811110572243e-06,
+      "loss": 0.8523,
+      "step": 3700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.240625321104072e-06,
+      "loss": 0.9215,
+      "step": 3701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.238439485542988e-06,
+      "loss": 0.6872,
+      "step": 3702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.236253604307698e-06,
+      "loss": 0.7821,
+      "step": 3703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.234067677816918e-06,
+      "loss": 0.8506,
+      "step": 3704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2318817064893735e-06,
+      "loss": 1.0116,
+      "step": 3705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.229695690743796e-06,
+      "loss": 0.7079,
+      "step": 3706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.227509630998927e-06,
+      "loss": 0.4809,
+      "step": 3707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.225323527673517e-06,
+      "loss": 0.7386,
+      "step": 3708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.223137381186325e-06,
+      "loss": 0.8663,
+      "step": 3709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2209511919561165e-06,
+      "loss": 0.4094,
+      "step": 3710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.218764960401668e-06,
+      "loss": 1.0919,
+      "step": 3711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.21657868694176e-06,
+      "loss": 0.7054,
+      "step": 3712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.214392371995185e-06,
+      "loss": 1.0175,
+      "step": 3713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.212206015980742e-06,
+      "loss": 1.0033,
+      "step": 3714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.210019619317237e-06,
+      "loss": 0.6785,
+      "step": 3715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.207833182423487e-06,
+      "loss": 0.7016,
+      "step": 3716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.205646705718311e-06,
+      "loss": 1.0352,
+      "step": 3717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.20346018962054e-06,
+      "loss": 1.1489,
+      "step": 3718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.201273634549013e-06,
+      "loss": 0.9756,
+      "step": 3719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.199087040922573e-06,
+      "loss": 0.6797,
+      "step": 3720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.196900409160074e-06,
+      "loss": 0.6611,
+      "step": 3721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.194713739680374e-06,
+      "loss": 0.9819,
+      "step": 3722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192527032902339e-06,
+      "loss": 1.0095,
+      "step": 3723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.190340289244846e-06,
+      "loss": 0.3411,
+      "step": 3724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.188153509126774e-06,
+      "loss": 0.6803,
+      "step": 3725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1859666929670125e-06,
+      "loss": 0.5654,
+      "step": 3726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.183779841184454e-06,
+      "loss": 1.0762,
+      "step": 3727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1815929541980026e-06,
+      "loss": 0.8409,
+      "step": 3728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.179406032426566e-06,
+      "loss": 0.7967,
+      "step": 3729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.177219076289059e-06,
+      "loss": 0.8525,
+      "step": 3730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.175032086204404e-06,
+      "loss": 0.7559,
+      "step": 3731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.172845062591527e-06,
+      "loss": 0.5524,
+      "step": 3732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.170658005869367e-06,
+      "loss": 0.6098,
+      "step": 3733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1684709164568605e-06,
+      "loss": 0.8975,
+      "step": 3734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1662837947729585e-06,
+      "loss": 0.9225,
+      "step": 3735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.164096641236611e-06,
+      "loss": 0.8653,
+      "step": 3736
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.161909456266781e-06,
+      "loss": 0.8748,
+      "step": 3737
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1597222402824306e-06,
+      "loss": 0.8772,
+      "step": 3738
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.157534993702535e-06,
+      "loss": 0.9933,
+      "step": 3739
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1553477169460675e-06,
+      "loss": 0.4857,
+      "step": 3740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.153160410432016e-06,
+      "loss": 0.6003,
+      "step": 3741
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.150973074579364e-06,
+      "loss": 1.0201,
+      "step": 3742
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.148785709807109e-06,
+      "loss": 0.9726,
+      "step": 3743
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.146598316534253e-06,
+      "loss": 0.9629,
+      "step": 3744
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.144410895179798e-06,
+      "loss": 0.633,
+      "step": 3745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.142223446162756e-06,
+      "loss": 1.0427,
+      "step": 3746
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.140035969902144e-06,
+      "loss": 0.9364,
+      "step": 3747
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.137848466816982e-06,
+      "loss": 0.7786,
+      "step": 3748
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.135660937326299e-06,
+      "loss": 1.0335,
+      "step": 3749
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.133473381849123e-06,
+      "loss": 0.7895,
+      "step": 3750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.131285800804493e-06,
+      "loss": 0.5338,
+      "step": 3751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.129098194611449e-06,
+      "loss": 0.5497,
+      "step": 3752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.12691056368904e-06,
+      "loss": 0.6647,
+      "step": 3753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.124722908456312e-06,
+      "loss": 1.1532,
+      "step": 3754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.122535229332326e-06,
+      "loss": 1.4328,
+      "step": 3755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.120347526736139e-06,
+      "loss": 0.8017,
+      "step": 3756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.118159801086816e-06,
+      "loss": 0.5732,
+      "step": 3757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.115972052803428e-06,
+      "loss": 0.858,
+      "step": 3758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1137842823050475e-06,
+      "loss": 0.6409,
+      "step": 3759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.111596490010751e-06,
+      "loss": 0.8336,
+      "step": 3760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.109408676339622e-06,
+      "loss": 1.0236,
+      "step": 3761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.107220841710746e-06,
+      "loss": 0.3464,
+      "step": 3762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.105032986543213e-06,
+      "loss": 0.8456,
+      "step": 3763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.102845111256116e-06,
+      "loss": 0.6408,
+      "step": 3764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.100657216268555e-06,
+      "loss": 0.7195,
+      "step": 3765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.09846930199963e-06,
+      "loss": 0.6214,
+      "step": 3766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.096281368868447e-06,
+      "loss": 0.7291,
+      "step": 3767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.094093417294113e-06,
+      "loss": 1.2286,
+      "step": 3768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.091905447695743e-06,
+      "loss": 0.791,
+      "step": 3769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089717460492451e-06,
+      "loss": 0.7281,
+      "step": 3770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0875294561033585e-06,
+      "loss": 0.822,
+      "step": 3771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.085341434947586e-06,
+      "loss": 0.9643,
+      "step": 3772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08315339744426e-06,
+      "loss": 1.0727,
+      "step": 3773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080965344012509e-06,
+      "loss": 0.9309,
+      "step": 3774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.078777275071464e-06,
+      "loss": 0.996,
+      "step": 3775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.076589191040262e-06,
+      "loss": 0.5355,
+      "step": 3776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.07440109233804e-06,
+      "loss": 1.0517,
+      "step": 3777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.072212979383939e-06,
+      "loss": 0.6756,
+      "step": 3778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.070024852597102e-06,
+      "loss": 0.8168,
+      "step": 3779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.067836712396673e-06,
+      "loss": 0.4368,
+      "step": 3780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.065648559201804e-06,
+      "loss": 0.9265,
+      "step": 3781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.063460393431644e-06,
+      "loss": 1.0097,
+      "step": 3782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0612722155053465e-06,
+      "loss": 0.618,
+      "step": 3783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059084025842068e-06,
+      "loss": 0.9527,
+      "step": 3784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.056895824860966e-06,
+      "loss": 0.7264,
+      "step": 3785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0547076129812014e-06,
+      "loss": 1.0744,
+      "step": 3786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052519390621935e-06,
+      "loss": 0.7821,
+      "step": 3787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050331158202333e-06,
+      "loss": 1.0326,
+      "step": 3788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.048142916141561e-06,
+      "loss": 1.0381,
+      "step": 3789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.045954664858785e-06,
+      "loss": 1.3327,
+      "step": 3790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.043766404773178e-06,
+      "loss": 0.993,
+      "step": 3791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04157813630391e-06,
+      "loss": 1.0348,
+      "step": 3792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.039389859870154e-06,
+      "loss": 0.9891,
+      "step": 3793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.037201575891084e-06,
+      "loss": 0.6811,
+      "step": 3794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.035013284785879e-06,
+      "loss": 0.762,
+      "step": 3795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032824986973714e-06,
+      "loss": 0.9088,
+      "step": 3796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.030636682873767e-06,
+      "loss": 0.6073,
+      "step": 3797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.028448372905218e-06,
+      "loss": 0.7651,
+      "step": 3798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.026260057487251e-06,
+      "loss": 0.5332,
+      "step": 3799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.024071737039047e-06,
+      "loss": 0.9817,
+      "step": 3800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.021883411979788e-06,
+      "loss": 0.676,
+      "step": 3801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.019695082728657e-06,
+      "loss": 0.639,
+      "step": 3802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.017506749704841e-06,
+      "loss": 0.8621,
+      "step": 3803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0153184133275255e-06,
+      "loss": 0.7049,
+      "step": 3804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.013130074015895e-06,
+      "loss": 1.1715,
+      "step": 3805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.010941732189137e-06,
+      "loss": 0.9074,
+      "step": 3806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.008753388266439e-06,
+      "loss": 0.7878,
+      "step": 3807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.006565042666989e-06,
+      "loss": 0.8847,
+      "step": 3808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.004376695809974e-06,
+      "loss": 1.1412,
+      "step": 3809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.002188348114581e-06,
+      "loss": 0.9439,
+      "step": 3810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5e-06,
+      "loss": 0.7864,
+      "step": 3811
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9978116518854195e-06,
+      "loss": 0.8361,
+      "step": 3812
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9956233041900275e-06,
+      "loss": 0.8365,
+      "step": 3813
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.993434957333011e-06,
+      "loss": 1.0795,
+      "step": 3814
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.991246611733561e-06,
+      "loss": 0.8858,
+      "step": 3815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.989058267810863e-06,
+      "loss": 0.8676,
+      "step": 3816
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9868699259841054e-06,
+      "loss": 1.196,
+      "step": 3817
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9846815866724745e-06,
+      "loss": 0.9208,
+      "step": 3818
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98249325029516e-06,
+      "loss": 0.9633,
+      "step": 3819
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.980304917271343e-06,
+      "loss": 0.8709,
+      "step": 3820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.978116588020214e-06,
+      "loss": 0.6565,
+      "step": 3821
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.975928262960954e-06,
+      "loss": 0.7539,
+      "step": 3822
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97373994251275e-06,
+      "loss": 0.7966,
+      "step": 3823
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.971551627094782e-06,
+      "loss": 0.559,
+      "step": 3824
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969363317126236e-06,
+      "loss": 0.6442,
+      "step": 3825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.967175013026289e-06,
+      "loss": 0.6192,
+      "step": 3826
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.964986715214124e-06,
+      "loss": 0.816,
+      "step": 3827
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.962798424108917e-06,
+      "loss": 0.6861,
+      "step": 3828
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.960610140129849e-06,
+      "loss": 0.7123,
+      "step": 3829
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.958421863696092e-06,
+      "loss": 0.7702,
+      "step": 3830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9562335952268245e-06,
+      "loss": 0.8942,
+      "step": 3831
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.954045335141216e-06,
+      "loss": 0.9235,
+      "step": 3832
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951857083858443e-06,
+      "loss": 0.8057,
+      "step": 3833
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949668841797669e-06,
+      "loss": 0.695,
+      "step": 3834
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.947480609378067e-06,
+      "loss": 1.142,
+      "step": 3835
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9452923870188e-06,
+      "loss": 0.5906,
+      "step": 3836
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.943104175139036e-06,
+      "loss": 0.9215,
+      "step": 3837
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.940915974157933e-06,
+      "loss": 0.9868,
+      "step": 3838
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9387277844946535e-06,
+      "loss": 0.9133,
+      "step": 3839
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.936539606568357e-06,
+      "loss": 0.7449,
+      "step": 3840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.934351440798196e-06,
+      "loss": 0.9497,
+      "step": 3841
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.932163287603328e-06,
+      "loss": 0.9263,
+      "step": 3842
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929975147402899e-06,
+      "loss": 0.7756,
+      "step": 3843
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.927787020616062e-06,
+      "loss": 0.7878,
+      "step": 3844
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.92559890766196e-06,
+      "loss": 0.7554,
+      "step": 3845
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.923410808959739e-06,
+      "loss": 0.819,
+      "step": 3846
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.921222724928536e-06,
+      "loss": 1.0097,
+      "step": 3847
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.919034655987493e-06,
+      "loss": 1.0542,
+      "step": 3848
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.916846602555741e-06,
+      "loss": 1.0413,
+      "step": 3849
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.914658565052416e-06,
+      "loss": 0.9146,
+      "step": 3850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9124705438966415e-06,
+      "loss": 1.2277,
+      "step": 3851
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9102825395075494e-06,
+      "loss": 0.7021,
+      "step": 3852
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908094552304259e-06,
+      "loss": 0.5711,
+      "step": 3853
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.905906582705888e-06,
+      "loss": 1.0256,
+      "step": 3854
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.903718631131556e-06,
+      "loss": 0.6985,
+      "step": 3855
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.901530698000372e-06,
+      "loss": 0.3381,
+      "step": 3856
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8993427837314465e-06,
+      "loss": 1.2123,
+      "step": 3857
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.897154888743885e-06,
+      "loss": 0.933,
+      "step": 3858
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.894967013456789e-06,
+      "loss": 0.3612,
+      "step": 3859
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.892779158289256e-06,
+      "loss": 0.555,
+      "step": 3860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.890591323660379e-06,
+      "loss": 0.8821,
+      "step": 3861
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88840350998925e-06,
+      "loss": 0.9332,
+      "step": 3862
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.886215717694954e-06,
+      "loss": 0.7352,
+      "step": 3863
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8840279471965735e-06,
+      "loss": 0.7515,
+      "step": 3864
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8818401989131844e-06,
+      "loss": 0.7196,
+      "step": 3865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8796524732638625e-06,
+      "loss": 0.8537,
+      "step": 3866
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.877464770667676e-06,
+      "loss": 0.8475,
+      "step": 3867
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.875277091543688e-06,
+      "loss": 0.6736,
+      "step": 3868
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.873089436310963e-06,
+      "loss": 0.7742,
+      "step": 3869
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.870901805388551e-06,
+      "loss": 1.0119,
+      "step": 3870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.868714199195509e-06,
+      "loss": 0.9564,
+      "step": 3871
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8665266181508774e-06,
+      "loss": 0.6536,
+      "step": 3872
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.864339062673703e-06,
+      "loss": 0.8542,
+      "step": 3873
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862151533183017e-06,
+      "loss": 0.7651,
+      "step": 3874
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.859964030097857e-06,
+      "loss": 0.9909,
+      "step": 3875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.857776553837243e-06,
+      "loss": 0.9014,
+      "step": 3876
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.855589104820203e-06,
+      "loss": 0.8984,
+      "step": 3877
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.853401683465748e-06,
+      "loss": 0.4964,
+      "step": 3878
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.851214290192892e-06,
+      "loss": 0.7356,
+      "step": 3879
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.849026925420637e-06,
+      "loss": 0.8624,
+      "step": 3880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8468395895679875e-06,
+      "loss": 1.14,
+      "step": 3881
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.844652283053933e-06,
+      "loss": 0.8575,
+      "step": 3882
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.842465006297468e-06,
+      "loss": 0.3037,
+      "step": 3883
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84027775971757e-06,
+      "loss": 0.9438,
+      "step": 3884
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.838090543733222e-06,
+      "loss": 1.2782,
+      "step": 3885
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.83590335876339e-06,
+      "loss": 0.5318,
+      "step": 3886
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.833716205227045e-06,
+      "loss": 0.8181,
+      "step": 3887
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.83152908354314e-06,
+      "loss": 0.6442,
+      "step": 3888
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.829341994130636e-06,
+      "loss": 1.0606,
+      "step": 3889
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.827154937408474e-06,
+      "loss": 0.8634,
+      "step": 3890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8249679137956e-06,
+      "loss": 1.0726,
+      "step": 3891
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.822780923710942e-06,
+      "loss": 0.561,
+      "step": 3892
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.820593967573436e-06,
+      "loss": 1.072,
+      "step": 3893
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.818407045801998e-06,
+      "loss": 0.9288,
+      "step": 3894
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8162201588155456e-06,
+      "loss": 1.026,
+      "step": 3895
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.814033307032989e-06,
+      "loss": 0.9227,
+      "step": 3896
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.811846490873225e-06,
+      "loss": 0.6632,
+      "step": 3897
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.809659710755155e-06,
+      "loss": 0.7497,
+      "step": 3898
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.80747296709766e-06,
+      "loss": 0.7368,
+      "step": 3899
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.805286260319628e-06,
+      "loss": 0.7747,
+      "step": 3900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8030995908399266e-06,
+      "loss": 0.7632,
+      "step": 3901
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.800912959077428e-06,
+      "loss": 0.5622,
+      "step": 3902
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.798726365450988e-06,
+      "loss": 0.8451,
+      "step": 3903
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7965398103794605e-06,
+      "loss": 0.6543,
+      "step": 3904
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.79435329428169e-06,
+      "loss": 0.8681,
+      "step": 3905
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.792166817576515e-06,
+      "loss": 0.8157,
+      "step": 3906
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789980380682763e-06,
+      "loss": 0.6495,
+      "step": 3907
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.78779398401926e-06,
+      "loss": 1.0976,
+      "step": 3908
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.785607628004816e-06,
+      "loss": 0.7823,
+      "step": 3909
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.783421313058242e-06,
+      "loss": 0.7262,
+      "step": 3910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7812350395983346e-06,
+      "loss": 0.806,
+      "step": 3911
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.779048808043886e-06,
+      "loss": 0.7232,
+      "step": 3912
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7768626188136765e-06,
+      "loss": 1.088,
+      "step": 3913
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.774676472326486e-06,
+      "loss": 0.8729,
+      "step": 3914
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.772490369001075e-06,
+      "loss": 0.8597,
+      "step": 3915
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7703043092562064e-06,
+      "loss": 0.6369,
+      "step": 3916
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.768118293510627e-06,
+      "loss": 0.6211,
+      "step": 3917
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.765932322183083e-06,
+      "loss": 1.0852,
+      "step": 3918
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7637463956923026e-06,
+      "loss": 1.0307,
+      "step": 3919
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.761560514457014e-06,
+      "loss": 1.1198,
+      "step": 3920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75937467889593e-06,
+      "loss": 0.952,
+      "step": 3921
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.757188889427761e-06,
+      "loss": 0.7838,
+      "step": 3922
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.755003146471201e-06,
+      "loss": 0.7893,
+      "step": 3923
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.752817450444941e-06,
+      "loss": 0.7677,
+      "step": 3924
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7506318017676655e-06,
+      "loss": 0.8162,
+      "step": 3925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.748446200858039e-06,
+      "loss": 0.6763,
+      "step": 3926
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.746260648134731e-06,
+      "loss": 0.7445,
+      "step": 3927
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.744075144016387e-06,
+      "loss": 0.6188,
+      "step": 3928
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.741889688921656e-06,
+      "loss": 0.7589,
+      "step": 3929
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7397042832691685e-06,
+      "loss": 0.4824,
+      "step": 3930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7375189274775535e-06,
+      "loss": 0.7951,
+      "step": 3931
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.735333621965421e-06,
+      "loss": 0.7359,
+      "step": 3932
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.733148367151382e-06,
+      "loss": 0.9763,
+      "step": 3933
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.730963163454027e-06,
+      "loss": 0.6222,
+      "step": 3934
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.728778011291947e-06,
+      "loss": 0.7871,
+      "step": 3935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.726592911083714e-06,
+      "loss": 0.8601,
+      "step": 3936
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.724407863247897e-06,
+      "loss": 0.5676,
+      "step": 3937
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7222228682030505e-06,
+      "loss": 0.905,
+      "step": 3938
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7200379263677235e-06,
+      "loss": 0.836,
+      "step": 3939
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.717853038160447e-06,
+      "loss": 0.7202,
+      "step": 3940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.715668203999752e-06,
+      "loss": 0.9369,
+      "step": 3941
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.713483424304148e-06,
+      "loss": 0.8876,
+      "step": 3942
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.711298699492146e-06,
+      "loss": 0.6457,
+      "step": 3943
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.709114029982235e-06,
+      "loss": 0.3996,
+      "step": 3944
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.706929416192904e-06,
+      "loss": 0.6615,
+      "step": 3945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.704744858542619e-06,
+      "loss": 0.9571,
+      "step": 3946
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7025603574498505e-06,
+      "loss": 0.8606,
+      "step": 3947
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.700375913333043e-06,
+      "loss": 0.7768,
+      "step": 3948
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6981915266106414e-06,
+      "loss": 1.1457,
+      "step": 3949
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.696007197701072e-06,
+      "loss": 0.9066,
+      "step": 3950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.693822927022757e-06,
+      "loss": 1.1689,
+      "step": 3951
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.691638714994101e-06,
+      "loss": 0.988,
+      "step": 3952
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.689454562033502e-06,
+      "loss": 1.2005,
+      "step": 3953
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6872704685593425e-06,
+      "loss": 0.7018,
+      "step": 3954
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.685086434989997e-06,
+      "loss": 0.5937,
+      "step": 3955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.682902461743828e-06,
+      "loss": 0.5718,
+      "step": 3956
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6807185492391845e-06,
+      "loss": 0.932,
+      "step": 3957
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.678534697894408e-06,
+      "loss": 0.6296,
+      "step": 3958
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6763509081278215e-06,
+      "loss": 0.9865,
+      "step": 3959
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6741671803577435e-06,
+      "loss": 0.8002,
+      "step": 3960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.671983515002476e-06,
+      "loss": 0.8228,
+      "step": 3961
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.66979991248031e-06,
+      "loss": 0.9523,
+      "step": 3962
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6676163732095255e-06,
+      "loss": 0.9753,
+      "step": 3963
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.66543289760839e-06,
+      "loss": 0.7742,
+      "step": 3964
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.663249486095158e-06,
+      "loss": 0.6151,
+      "step": 3965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.661066139088072e-06,
+      "loss": 0.6984,
+      "step": 3966
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.658882857005363e-06,
+      "loss": 0.9058,
+      "step": 3967
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6566996402652486e-06,
+      "loss": 0.7774,
+      "step": 3968
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.654516489285935e-06,
+      "loss": 0.3149,
+      "step": 3969
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.652333404485613e-06,
+      "loss": 0.8098,
+      "step": 3970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.650150386282465e-06,
+      "loss": 0.7828,
+      "step": 3971
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6479674350946576e-06,
+      "loss": 0.6134,
+      "step": 3972
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6457845513403455e-06,
+      "loss": 0.918,
+      "step": 3973
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.64360173543767e-06,
+      "loss": 0.5303,
+      "step": 3974
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.641418987804761e-06,
+      "loss": 0.8663,
+      "step": 3975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.639236308859731e-06,
+      "loss": 1.0866,
+      "step": 3976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637053699020687e-06,
+      "loss": 0.6619,
+      "step": 3977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.634871158705712e-06,
+      "loss": 0.5095,
+      "step": 3978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.63268868833289e-06,
+      "loss": 0.445,
+      "step": 3979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.630506288320276e-06,
+      "loss": 0.761,
+      "step": 3980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.628323959085924e-06,
+      "loss": 0.4526,
+      "step": 3981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.626141701047865e-06,
+      "loss": 0.962,
+      "step": 3982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6239595146241265e-06,
+      "loss": 0.801,
+      "step": 3983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.62177740023271e-06,
+      "loss": 0.858,
+      "step": 3984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.619595358291617e-06,
+      "loss": 1.0496,
+      "step": 3985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.61741338921882e-06,
+      "loss": 0.9199,
+      "step": 3986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.615231493432293e-06,
+      "loss": 0.9685,
+      "step": 3987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.613049671349982e-06,
+      "loss": 0.6002,
+      "step": 3988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.610867923389832e-06,
+      "loss": 0.9728,
+      "step": 3989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.608686249969759e-06,
+      "loss": 0.8182,
+      "step": 3990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60650465150768e-06,
+      "loss": 0.7307,
+      "step": 3991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6043231284214854e-06,
+      "loss": 0.6086,
+      "step": 3992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.602141681129061e-06,
+      "loss": 0.933,
+      "step": 3993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.599960310048268e-06,
+      "loss": 0.9952,
+      "step": 3994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5977790155969635e-06,
+      "loss": 0.6616,
+      "step": 3995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.59559779819298e-06,
+      "loss": 0.4855,
+      "step": 3996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.593416658254144e-06,
+      "loss": 0.7376,
+      "step": 3997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.59123559619826e-06,
+      "loss": 0.8218,
+      "step": 3998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.589054612443123e-06,
+      "loss": 0.8555,
+      "step": 3999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5868737074065086e-06,
+      "loss": 0.7919,
+      "step": 4000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.584692881506182e-06,
+      "loss": 1.1748,
+      "step": 4001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.582512135159887e-06,
+      "loss": 0.6677,
+      "step": 4002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.580331468785361e-06,
+      "loss": 0.8336,
+      "step": 4003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.578150882800315e-06,
+      "loss": 0.6752,
+      "step": 4004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.575970377622456e-06,
+      "loss": 1.0526,
+      "step": 4005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.573789953669465e-06,
+      "loss": 0.9538,
+      "step": 4006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5716096113590156e-06,
+      "loss": 0.7167,
+      "step": 4007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.569429351108764e-06,
+      "loss": 1.096,
+      "step": 4008
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5672491733363446e-06,
+      "loss": 0.6899,
+      "step": 4009
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565069078459386e-06,
+      "loss": 0.9017,
+      "step": 4010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.562889066895491e-06,
+      "loss": 0.9525,
+      "step": 4011
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.560709139062255e-06,
+      "loss": 0.6491,
+      "step": 4012
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5585292953772495e-06,
+      "loss": 0.6779,
+      "step": 4013
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5563495362580384e-06,
+      "loss": 0.9713,
+      "step": 4014
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.55416986212216e-06,
+      "loss": 0.6054,
+      "step": 4015
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.551990273387146e-06,
+      "loss": 0.338,
+      "step": 4016
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.549810770470501e-06,
+      "loss": 0.94,
+      "step": 4017
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.547631353789725e-06,
+      "loss": 0.9556,
+      "step": 4018
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.545452023762292e-06,
+      "loss": 0.8013,
+      "step": 4019
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.543272780805665e-06,
+      "loss": 0.617,
+      "step": 4020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.541093625337285e-06,
+      "loss": 0.8839,
+      "step": 4021
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.538914557774584e-06,
+      "loss": 0.9747,
+      "step": 4022
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.536735578534968e-06,
+      "loss": 0.778,
+      "step": 4023
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.534556688035835e-06,
+      "loss": 0.4121,
+      "step": 4024
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.532377886694558e-06,
+      "loss": 0.7797,
+      "step": 4025
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5301991749285e-06,
+      "loss": 1.0205,
+      "step": 4026
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.528020553154999e-06,
+      "loss": 1.1136,
+      "step": 4027
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.525842021791387e-06,
+      "loss": 0.8635,
+      "step": 4028
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523663581254965e-06,
+      "loss": 1.118,
+      "step": 4029
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.521485231963028e-06,
+      "loss": 0.6176,
+      "step": 4030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.519306974332847e-06,
+      "loss": 0.9789,
+      "step": 4031
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.517128808781679e-06,
+      "loss": 1.0632,
+      "step": 4032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.51495073572676e-06,
+      "loss": 0.806,
+      "step": 4033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5127727555853105e-06,
+      "loss": 0.42,
+      "step": 4034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.510594868774536e-06,
+      "loss": 0.9593,
+      "step": 4035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.508417075711616e-06,
+      "loss": 0.8548,
+      "step": 4036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.506239376813722e-06,
+      "loss": 1.1001,
+      "step": 4037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.504061772497998e-06,
+      "loss": 0.6048,
+      "step": 4038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.501884263181579e-06,
+      "loss": 0.8228,
+      "step": 4039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.499706849281571e-06,
+      "loss": 0.5677,
+      "step": 4040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.497529531215074e-06,
+      "loss": 1.0952,
+      "step": 4041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.495352309399159e-06,
+      "loss": 0.801,
+      "step": 4042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.493175184250887e-06,
+      "loss": 0.7799,
+      "step": 4043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.490998156187293e-06,
+      "loss": 0.9543,
+      "step": 4044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4888212256253995e-06,
+      "loss": 0.5551,
+      "step": 4045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.486644392982206e-06,
+      "loss": 0.8762,
+      "step": 4046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4844676586746975e-06,
+      "loss": 0.7614,
+      "step": 4047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.482291023119834e-06,
+      "loss": 0.8081,
+      "step": 4048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.480114486734565e-06,
+      "loss": 0.9406,
+      "step": 4049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.477938049935811e-06,
+      "loss": 0.8156,
+      "step": 4050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.475761713140483e-06,
+      "loss": 1.1099,
+      "step": 4051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4735854767654654e-06,
+      "loss": 0.603,
+      "step": 4052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.471409341227631e-06,
+      "loss": 1.0979,
+      "step": 4053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.469233306943822e-06,
+      "loss": 0.7087,
+      "step": 4054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.467057374330875e-06,
+      "loss": 0.737,
+      "step": 4055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.464881543805595e-06,
+      "loss": 0.5985,
+      "step": 4056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.462705815784777e-06,
+      "loss": 0.6986,
+      "step": 4057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.460530190685187e-06,
+      "loss": 0.8951,
+      "step": 4058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.458354668923581e-06,
+      "loss": 0.6515,
+      "step": 4059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.456179250916686e-06,
+      "loss": 0.5528,
+      "step": 4060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.454003937081218e-06,
+      "loss": 0.9712,
+      "step": 4061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.451828727833865e-06,
+      "loss": 0.9524,
+      "step": 4062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4496536235913e-06,
+      "loss": 0.7537,
+      "step": 4063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4474786247701755e-06,
+      "loss": 0.7542,
+      "step": 4064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.44530373178712e-06,
+      "loss": 0.5969,
+      "step": 4065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.443128945058748e-06,
+      "loss": 1.1533,
+      "step": 4066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.440954265001647e-06,
+      "loss": 0.899,
+      "step": 4067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.438779692032389e-06,
+      "loss": 0.6756,
+      "step": 4068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.436605226567523e-06,
+      "loss": 0.927,
+      "step": 4069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.434430869023579e-06,
+      "loss": 0.4814,
+      "step": 4070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.432256619817064e-06,
+      "loss": 0.5654,
+      "step": 4071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.430082479364467e-06,
+      "loss": 0.3933,
+      "step": 4072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.427908448082254e-06,
+      "loss": 0.3498,
+      "step": 4073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.425734526386871e-06,
+      "loss": 0.79,
+      "step": 4074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.423560714694743e-06,
+      "loss": 0.7498,
+      "step": 4075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.421387013422275e-06,
+      "loss": 0.6457,
+      "step": 4076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.419213422985848e-06,
+      "loss": 0.9406,
+      "step": 4077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.417039943801823e-06,
+      "loss": 0.8947,
+      "step": 4078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4148665762865425e-06,
+      "loss": 0.7133,
+      "step": 4079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.412693320856325e-06,
+      "loss": 0.7835,
+      "step": 4080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.410520177927466e-06,
+      "loss": 0.8677,
+      "step": 4081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.408347147916241e-06,
+      "loss": 0.9842,
+      "step": 4082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.406174231238907e-06,
+      "loss": 0.7213,
+      "step": 4083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.404001428311695e-06,
+      "loss": 0.5834,
+      "step": 4084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.401828739550814e-06,
+      "loss": 0.757,
+      "step": 4085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3996561653724545e-06,
+      "loss": 0.6346,
+      "step": 4086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.397483706192784e-06,
+      "loss": 1.1273,
+      "step": 4087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.395311362427944e-06,
+      "loss": 0.8496,
+      "step": 4088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.393139134494062e-06,
+      "loss": 0.7452,
+      "step": 4089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.390967022807233e-06,
+      "loss": 0.644,
+      "step": 4090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38879502778354e-06,
+      "loss": 0.498,
+      "step": 4091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.386623149839037e-06,
+      "loss": 0.6826,
+      "step": 4092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.384451389389758e-06,
+      "loss": 0.6035,
+      "step": 4093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.382279746851711e-06,
+      "loss": 0.5627,
+      "step": 4094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38010822264089e-06,
+      "loss": 0.9181,
+      "step": 4095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.377936817173254e-06,
+      "loss": 1.061,
+      "step": 4096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.375765530864752e-06,
+      "loss": 0.5089,
+      "step": 4097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3735943641313e-06,
+      "loss": 0.5282,
+      "step": 4098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.371423317388797e-06,
+      "loss": 1.1658,
+      "step": 4099
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369252391053116e-06,
+      "loss": 0.6543,
+      "step": 4100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.367081585540112e-06,
+      "loss": 0.7732,
+      "step": 4101
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.364910901265607e-06,
+      "loss": 0.3909,
+      "step": 4102
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3627403386454105e-06,
+      "loss": 0.3752,
+      "step": 4103
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.360569898095301e-06,
+      "loss": 0.7432,
+      "step": 4104
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.35839958003104e-06,
+      "loss": 0.6823,
+      "step": 4105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.356229384868358e-06,
+      "loss": 1.0142,
+      "step": 4106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3540593130229695e-06,
+      "loss": 0.9482,
+      "step": 4107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.351889364910559e-06,
+      "loss": 0.8201,
+      "step": 4108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.349719540946795e-06,
+      "loss": 0.9501,
+      "step": 4109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.347549841547311e-06,
+      "loss": 0.8683,
+      "step": 4110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.34538026712773e-06,
+      "loss": 0.9508,
+      "step": 4111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.343210818103637e-06,
+      "loss": 0.6127,
+      "step": 4112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.341041494890607e-06,
+      "loss": 0.696,
+      "step": 4113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3388722979041796e-06,
+      "loss": 0.9349,
+      "step": 4114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.336703227559877e-06,
+      "loss": 0.6817,
+      "step": 4115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.334534284273193e-06,
+      "loss": 1.379,
+      "step": 4116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.332365468459601e-06,
+      "loss": 1.1049,
+      "step": 4117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.330196780534545e-06,
+      "loss": 0.6394,
+      "step": 4118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.32802822091345e-06,
+      "loss": 0.6737,
+      "step": 4119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.325859790011714e-06,
+      "loss": 0.8586,
+      "step": 4120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.323691488244706e-06,
+      "loss": 0.8527,
+      "step": 4121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.321523316027781e-06,
+      "loss": 0.5788,
+      "step": 4122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.319355273776257e-06,
+      "loss": 0.6301,
+      "step": 4123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.317187361905437e-06,
+      "loss": 0.6318,
+      "step": 4124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.315019580830589e-06,
+      "loss": 0.6816,
+      "step": 4125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.312851930966969e-06,
+      "loss": 0.9504,
+      "step": 4126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.310684412729793e-06,
+      "loss": 0.8467,
+      "step": 4127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.308517026534267e-06,
+      "loss": 0.9488,
+      "step": 4128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.306349772795557e-06,
+      "loss": 0.6539,
+      "step": 4129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.304182651928816e-06,
+      "loss": 0.5395,
+      "step": 4130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302015664349161e-06,
+      "loss": 0.7616,
+      "step": 4131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2998488104716946e-06,
+      "loss": 0.9203,
+      "step": 4132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.297682090711483e-06,
+      "loss": 1.0033,
+      "step": 4133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.295515505483575e-06,
+      "loss": 0.6662,
+      "step": 4134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293349055202986e-06,
+      "loss": 0.6968,
+      "step": 4135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291182740284715e-06,
+      "loss": 0.8676,
+      "step": 4136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.289016561143726e-06,
+      "loss": 0.928,
+      "step": 4137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286850518194962e-06,
+      "loss": 0.4827,
+      "step": 4138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.284684611853338e-06,
+      "loss": 0.9232,
+      "step": 4139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.282518842533747e-06,
+      "loss": 0.9116,
+      "step": 4140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.280353210651047e-06,
+      "loss": 0.8136,
+      "step": 4141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.27818771662008e-06,
+      "loss": 0.6132,
+      "step": 4142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2760223608556515e-06,
+      "loss": 0.8257,
+      "step": 4143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.27385714377255e-06,
+      "loss": 0.696,
+      "step": 4144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2716920657855295e-06,
+      "loss": 0.8019,
+      "step": 4145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.269527127309323e-06,
+      "loss": 0.8976,
+      "step": 4146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267362328758636e-06,
+      "loss": 0.9698,
+      "step": 4147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.265197670548143e-06,
+      "loss": 0.6219,
+      "step": 4148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.263033153092497e-06,
+      "loss": 0.41,
+      "step": 4149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.260868776806318e-06,
+      "loss": 0.7397,
+      "step": 4150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258704542104207e-06,
+      "loss": 1.0121,
+      "step": 4151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.256540449400729e-06,
+      "loss": 0.5246,
+      "step": 4152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.25437649911043e-06,
+      "loss": 0.8126,
+      "step": 4153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.252212691647821e-06,
+      "loss": 0.7911,
+      "step": 4154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.250049027427393e-06,
+      "loss": 0.3895,
+      "step": 4155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.247885506863603e-06,
+      "loss": 0.7761,
+      "step": 4156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.245722130370887e-06,
+      "loss": 0.8935,
+      "step": 4157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.243558898363646e-06,
+      "loss": 1.0081,
+      "step": 4158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.241395811256262e-06,
+      "loss": 1.0147,
+      "step": 4159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239232869463079e-06,
+      "loss": 0.5605,
+      "step": 4160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.237070073398425e-06,
+      "loss": 0.796,
+      "step": 4161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.234907423476587e-06,
+      "loss": 0.546,
+      "step": 4162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.232744920111836e-06,
+      "loss": 0.8373,
+      "step": 4163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.230582563718405e-06,
+      "loss": 0.7359,
+      "step": 4164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.22842035471051e-06,
+      "loss": 1.048,
+      "step": 4165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.226258293502326e-06,
+      "loss": 0.5685,
+      "step": 4166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.224096380508011e-06,
+      "loss": 0.6833,
+      "step": 4167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.221934616141685e-06,
+      "loss": 0.6275,
+      "step": 4168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.219773000817449e-06,
+      "loss": 0.8704,
+      "step": 4169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.217611534949365e-06,
+      "loss": 0.8333,
+      "step": 4170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.215450218951477e-06,
+      "loss": 0.7689,
+      "step": 4171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.213289053237792e-06,
+      "loss": 1.148,
+      "step": 4172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.211128038222293e-06,
+      "loss": 0.8306,
+      "step": 4173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.208967174318933e-06,
+      "loss": 0.5112,
+      "step": 4174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.206806461941634e-06,
+      "loss": 0.8393,
+      "step": 4175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.204645901504292e-06,
+      "loss": 0.7283,
+      "step": 4176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.202485493420773e-06,
+      "loss": 0.7992,
+      "step": 4177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.200325238104911e-06,
+      "loss": 0.7913,
+      "step": 4178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.198165135970515e-06,
+      "loss": 0.4882,
+      "step": 4179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.196005187431362e-06,
+      "loss": 0.8092,
+      "step": 4180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1938453929012014e-06,
+      "loss": 1.0069,
+      "step": 4181
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.19168575279375e-06,
+      "loss": 0.7995,
+      "step": 4182
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1895262675227e-06,
+      "loss": 0.5468,
+      "step": 4183
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1873669375017075e-06,
+      "loss": 1.0831,
+      "step": 4184
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.185207763144404e-06,
+      "loss": 0.539,
+      "step": 4185
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1830487448643906e-06,
+      "loss": 1.0205,
+      "step": 4186
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.180889883075235e-06,
+      "loss": 0.6643,
+      "step": 4187
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.17873117819048e-06,
+      "loss": 0.8351,
+      "step": 4188
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.176572630623634e-06,
+      "loss": 0.7333,
+      "step": 4189
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.174414240788177e-06,
+      "loss": 1.3185,
+      "step": 4190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1722560090975585e-06,
+      "loss": 0.8872,
+      "step": 4191
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.170097935965199e-06,
+      "loss": 0.9698,
+      "step": 4192
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.167940021804487e-06,
+      "loss": 1.0101,
+      "step": 4193
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1657822670287815e-06,
+      "loss": 0.6606,
+      "step": 4194
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.163624672051411e-06,
+      "loss": 0.6499,
+      "step": 4195
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.161467237285671e-06,
+      "loss": 0.8776,
+      "step": 4196
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.15930996314483e-06,
+      "loss": 0.8843,
+      "step": 4197
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1571528500421226e-06,
+      "loss": 0.4941,
+      "step": 4198
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.154995898390756e-06,
+      "loss": 0.7059,
+      "step": 4199
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.152839108603903e-06,
+      "loss": 0.983,
+      "step": 4200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.150682481094707e-06,
+      "loss": 0.9164,
+      "step": 4201
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.148526016276279e-06,
+      "loss": 0.572,
+      "step": 4202
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.146369714561704e-06,
+      "loss": 1.1897,
+      "step": 4203
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.144213576364026e-06,
+      "loss": 0.6541,
+      "step": 4204
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.14205760209627e-06,
+      "loss": 0.6899,
+      "step": 4205
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.139901792171416e-06,
+      "loss": 0.9301,
+      "step": 4206
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.137746147002427e-06,
+      "loss": 0.7796,
+      "step": 4207
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.135590667002219e-06,
+      "loss": 0.8316,
+      "step": 4208
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.133435352583691e-06,
+      "loss": 1.1605,
+      "step": 4209
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1312802041597e-06,
+      "loss": 0.6581,
+      "step": 4210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.129125222143077e-06,
+      "loss": 0.9949,
+      "step": 4211
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.126970406946615e-06,
+      "loss": 0.8502,
+      "step": 4212
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.124815758983085e-06,
+      "loss": 0.735,
+      "step": 4213
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.122661278665215e-06,
+      "loss": 0.9793,
+      "step": 4214
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.120506966405708e-06,
+      "loss": 0.9067,
+      "step": 4215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.118352822617232e-06,
+      "loss": 0.6429,
+      "step": 4216
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.116198847712424e-06,
+      "loss": 0.771,
+      "step": 4217
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1140450421038865e-06,
+      "loss": 0.7145,
+      "step": 4218
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.111891406204195e-06,
+      "loss": 0.1177,
+      "step": 4219
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.109737940425883e-06,
+      "loss": 0.8883,
+      "step": 4220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.107584645181462e-06,
+      "loss": 0.6724,
+      "step": 4221
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.105431520883401e-06,
+      "loss": 0.7457,
+      "step": 4222
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.103278567944147e-06,
+      "loss": 1.0999,
+      "step": 4223
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.101125786776102e-06,
+      "loss": 0.8273,
+      "step": 4224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.098973177791647e-06,
+      "loss": 1.0668,
+      "step": 4225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.096820741403119e-06,
+      "loss": 0.4377,
+      "step": 4226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.094668478022834e-06,
+      "loss": 0.4419,
+      "step": 4227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.092516388063061e-06,
+      "loss": 0.462,
+      "step": 4228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.090364471936046e-06,
+      "loss": 0.5284,
+      "step": 4229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.088212730054001e-06,
+      "loss": 0.6829,
+      "step": 4230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.086061162829099e-06,
+      "loss": 1.1044,
+      "step": 4231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.083909770673486e-06,
+      "loss": 0.935,
+      "step": 4232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.081758553999267e-06,
+      "loss": 0.7949,
+      "step": 4233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.079607513218522e-06,
+      "loss": 0.9507,
+      "step": 4234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.07745664874329e-06,
+      "loss": 0.6485,
+      "step": 4235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0753059609855815e-06,
+      "loss": 0.984,
+      "step": 4236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.073155450357368e-06,
+      "loss": 0.6056,
+      "step": 4237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.071005117270593e-06,
+      "loss": 1.0272,
+      "step": 4238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.06885496213716e-06,
+      "loss": 0.8685,
+      "step": 4239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.066704985368945e-06,
+      "loss": 0.762,
+      "step": 4240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.06455518737778e-06,
+      "loss": 0.6941,
+      "step": 4241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.062405568575477e-06,
+      "loss": 0.4383,
+      "step": 4242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.060256129373799e-06,
+      "loss": 0.9685,
+      "step": 4243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.058106870184486e-06,
+      "loss": 0.9411,
+      "step": 4244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0559577914192335e-06,
+      "loss": 1.0631,
+      "step": 4245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.053808893489714e-06,
+      "loss": 0.7912,
+      "step": 4246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.051660176807553e-06,
+      "loss": 0.7407,
+      "step": 4247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.049511641784352e-06,
+      "loss": 0.7908,
+      "step": 4248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.047363288831668e-06,
+      "loss": 0.6072,
+      "step": 4249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.045215118361034e-06,
+      "loss": 0.6057,
+      "step": 4250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.043067130783938e-06,
+      "loss": 0.9364,
+      "step": 4251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.040919326511839e-06,
+      "loss": 0.8691,
+      "step": 4252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0387717059561566e-06,
+      "loss": 0.6506,
+      "step": 4253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.036624269528281e-06,
+      "loss": 0.8614,
+      "step": 4254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.034477017639561e-06,
+      "loss": 1.113,
+      "step": 4255
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.032329950701315e-06,
+      "loss": 0.9214,
+      "step": 4256
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.030183069124822e-06,
+      "loss": 0.9415,
+      "step": 4257
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.028036373321327e-06,
+      "loss": 0.9149,
+      "step": 4258
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.025889863702043e-06,
+      "loss": 0.8093,
+      "step": 4259
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.02374354067814e-06,
+      "loss": 0.9674,
+      "step": 4260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0215974046607595e-06,
+      "loss": 0.7541,
+      "step": 4261
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.019451456061e-06,
+      "loss": 0.6358,
+      "step": 4262
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.017305695289934e-06,
+      "loss": 0.7074,
+      "step": 4263
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.015160122758586e-06,
+      "loss": 0.8819,
+      "step": 4264
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.013014738877954e-06,
+      "loss": 0.7511,
+      "step": 4265
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.010869544058995e-06,
+      "loss": 0.8859,
+      "step": 4266
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.008724538712633e-06,
+      "loss": 0.8726,
+      "step": 4267
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.00657972324975e-06,
+      "loss": 0.6738,
+      "step": 4268
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.004435098081201e-06,
+      "loss": 0.7488,
+      "step": 4269
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.002290663617793e-06,
+      "loss": 0.6632,
+      "step": 4270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.000146420270309e-06,
+      "loss": 0.6817,
+      "step": 4271
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.998002368449482e-06,
+      "loss": 0.6874,
+      "step": 4272
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.995858508566022e-06,
+      "loss": 0.6772,
+      "step": 4273
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.993714841030589e-06,
+      "loss": 0.8686,
+      "step": 4274
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.991571366253819e-06,
+      "loss": 1.1084,
+      "step": 4275
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.989428084646299e-06,
+      "loss": 0.6917,
+      "step": 4276
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.987284996618589e-06,
+      "loss": 0.3858,
+      "step": 4277
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.985142102581203e-06,
+      "loss": 0.7417,
+      "step": 4278
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.982999402944629e-06,
+      "loss": 0.8399,
+      "step": 4279
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.980856898119304e-06,
+      "loss": 0.878,
+      "step": 4280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.978714588515641e-06,
+      "loss": 0.8249,
+      "step": 4281
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9765724745440044e-06,
+      "loss": 0.6363,
+      "step": 4282
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.974430556614731e-06,
+      "loss": 0.8424,
+      "step": 4283
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.972288835138111e-06,
+      "loss": 0.8277,
+      "step": 4284
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9701473105244025e-06,
+      "loss": 1.0419,
+      "step": 4285
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.968005983183827e-06,
+      "loss": 0.4953,
+      "step": 4286
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.965864853526563e-06,
+      "loss": 0.9508,
+      "step": 4287
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.963723921962755e-06,
+      "loss": 0.8765,
+      "step": 4288
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.96158318890251e-06,
+      "loss": 0.7112,
+      "step": 4289
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.959442654755893e-06,
+      "loss": 1.057,
+      "step": 4290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.957302319932934e-06,
+      "loss": 0.63,
+      "step": 4291
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.955162184843625e-06,
+      "loss": 0.8969,
+      "step": 4292
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.953022249897919e-06,
+      "loss": 0.4077,
+      "step": 4293
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9508825155057294e-06,
+      "loss": 0.5271,
+      "step": 4294
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.948742982076934e-06,
+      "loss": 1.122,
+      "step": 4295
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.94660365002137e-06,
+      "loss": 0.7551,
+      "step": 4296
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9444645197488345e-06,
+      "loss": 0.7091,
+      "step": 4297
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.942325591669091e-06,
+      "loss": 0.6474,
+      "step": 4298
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.940186866191859e-06,
+      "loss": 0.67,
+      "step": 4299
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9380483437268245e-06,
+      "loss": 0.9876,
+      "step": 4300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9359100246836266e-06,
+      "loss": 0.7984,
+      "step": 4301
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.933771909471876e-06,
+      "loss": 0.3869,
+      "step": 4302
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.931633998501133e-06,
+      "loss": 0.8425,
+      "step": 4303
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.92949629218093e-06,
+      "loss": 0.7795,
+      "step": 4304
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.927358790920751e-06,
+      "loss": 0.9768,
+      "step": 4305
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.925221495130046e-06,
+      "loss": 1.0298,
+      "step": 4306
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.923084405218223e-06,
+      "loss": 0.4003,
+      "step": 4307
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9209475215946535e-06,
+      "loss": 0.7718,
+      "step": 4308
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.918810844668664e-06,
+      "loss": 0.7986,
+      "step": 4309
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916674374849551e-06,
+      "loss": 0.5297,
+      "step": 4310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.914538112546558e-06,
+      "loss": 0.7497,
+      "step": 4311
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9124020581689006e-06,
+      "loss": 0.7331,
+      "step": 4312
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.910266212125752e-06,
+      "loss": 0.8643,
+      "step": 4313
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.908130574826238e-06,
+      "loss": 0.7941,
+      "step": 4314
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9059951466794555e-06,
+      "loss": 0.6527,
+      "step": 4315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.903859928094451e-06,
+      "loss": 0.7604,
+      "step": 4316
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9017249194802394e-06,
+      "loss": 0.7165,
+      "step": 4317
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8995901212457894e-06,
+      "loss": 0.8656,
+      "step": 4318
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8974555338000335e-06,
+      "loss": 0.7497,
+      "step": 4319
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.895321157551859e-06,
+      "loss": 0.8073,
+      "step": 4320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.89318699291012e-06,
+      "loss": 0.7805,
+      "step": 4321
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.891053040283621e-06,
+      "loss": 0.519,
+      "step": 4322
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.888919300081135e-06,
+      "loss": 0.5099,
+      "step": 4323
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.886785772711386e-06,
+      "loss": 0.8349,
+      "step": 4324
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.884652458583066e-06,
+      "loss": 1.1598,
+      "step": 4325
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.882519358104815e-06,
+      "loss": 0.5836,
+      "step": 4326
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.880386471685246e-06,
+      "loss": 0.8225,
+      "step": 4327
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.878253799732916e-06,
+      "loss": 0.646,
+      "step": 4328
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8761213426563546e-06,
+      "loss": 0.6801,
+      "step": 4329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.873989100864039e-06,
+      "loss": 0.7944,
+      "step": 4330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.871857074764414e-06,
+      "loss": 0.6021,
+      "step": 4331
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.869725264765876e-06,
+      "loss": 0.8331,
+      "step": 4332
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.867593671276787e-06,
+      "loss": 0.9696,
+      "step": 4333
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.86546229470546e-06,
+      "loss": 0.9959,
+      "step": 4334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.863331135460173e-06,
+      "loss": 0.7303,
+      "step": 4335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861200193949157e-06,
+      "loss": 0.6484,
+      "step": 4336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.859069470580607e-06,
+      "loss": 0.957,
+      "step": 4337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.856938965762669e-06,
+      "loss": 0.4218,
+      "step": 4338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.854808679903457e-06,
+      "loss": 0.9533,
+      "step": 4339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.85267861341103e-06,
+      "loss": 0.8398,
+      "step": 4340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8505487666934175e-06,
+      "loss": 0.9037,
+      "step": 4341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8484191401586025e-06,
+      "loss": 0.8538,
+      "step": 4342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.84628973421452e-06,
+      "loss": 0.6207,
+      "step": 4343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.844160549269074e-06,
+      "loss": 0.5694,
+      "step": 4344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842031585730113e-06,
+      "loss": 0.8794,
+      "step": 4345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.839902844005456e-06,
+      "loss": 0.8215,
+      "step": 4346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.837774324502869e-06,
+      "loss": 0.9859,
+      "step": 4347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835646027630084e-06,
+      "loss": 0.5736,
+      "step": 4348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.833517953794781e-06,
+      "loss": 0.8323,
+      "step": 4349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8313901034046095e-06,
+      "loss": 0.5421,
+      "step": 4350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8292624768671624e-06,
+      "loss": 0.5379,
+      "step": 4351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.827135074590003e-06,
+      "loss": 1.0124,
+      "step": 4352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8250078969806385e-06,
+      "loss": 0.5744,
+      "step": 4353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.822880944446545e-06,
+      "loss": 0.7428,
+      "step": 4354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.820754217395147e-06,
+      "loss": 0.951,
+      "step": 4355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.818627716233833e-06,
+      "loss": 0.3104,
+      "step": 4356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.816501441369939e-06,
+      "loss": 0.6569,
+      "step": 4357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.814375393210769e-06,
+      "loss": 0.928,
+      "step": 4358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8122495721635717e-06,
+      "loss": 0.5948,
+      "step": 4359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.810123978635563e-06,
+      "loss": 0.4224,
+      "step": 4360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.807998613033905e-06,
+      "loss": 0.8665,
+      "step": 4361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.805873475765728e-06,
+      "loss": 0.4917,
+      "step": 4362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.803748567238106e-06,
+      "loss": 0.8666,
+      "step": 4363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.80162388785808e-06,
+      "loss": 0.7109,
+      "step": 4364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7994994380326384e-06,
+      "loss": 0.7903,
+      "step": 4365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7973752181687336e-06,
+      "loss": 0.8954,
+      "step": 4366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.795251228673265e-06,
+      "loss": 0.9279,
+      "step": 4367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.793127469953096e-06,
+      "loss": 0.6487,
+      "step": 4368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.791003942415044e-06,
+      "loss": 0.4173,
+      "step": 4369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7888806464658767e-06,
+      "loss": 0.6951,
+      "step": 4370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7867575825123253e-06,
+      "loss": 0.9626,
+      "step": 4371
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.784634750961069e-06,
+      "loss": 1.1881,
+      "step": 4372
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7825121522187507e-06,
+      "loss": 0.6093,
+      "step": 4373
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7803897866919593e-06,
+      "loss": 0.6654,
+      "step": 4374
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7782676547872478e-06,
+      "loss": 1.0271,
+      "step": 4375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7761457569111183e-06,
+      "loss": 0.9265,
+      "step": 4376
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7740240934700307e-06,
+      "loss": 0.5502,
+      "step": 4377
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7719026648704e-06,
+      "loss": 0.4479,
+      "step": 4378
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7697814715185956e-06,
+      "loss": 0.9311,
+      "step": 4379
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.767660513820942e-06,
+      "loss": 0.8924,
+      "step": 4380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7655397921837195e-06,
+      "loss": 0.7201,
+      "step": 4381
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7634193070131613e-06,
+      "loss": 0.7205,
+      "step": 4382
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.761299058715456e-06,
+      "loss": 0.8217,
+      "step": 4383
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7591790476967492e-06,
+      "loss": 0.3079,
+      "step": 4384
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.757059274363137e-06,
+      "loss": 0.5929,
+      "step": 4385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7549397391206722e-06,
+      "loss": 0.6757,
+      "step": 4386
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7528204423753624e-06,
+      "loss": 0.567,
+      "step": 4387
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7507013845331686e-06,
+      "loss": 0.6363,
+      "step": 4388
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.748582566000006e-06,
+      "loss": 1.0363,
+      "step": 4389
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7464639871817444e-06,
+      "loss": 0.7601,
+      "step": 4390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7443456484842084e-06,
+      "loss": 0.932,
+      "step": 4391
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7422275503131745e-06,
+      "loss": 0.9,
+      "step": 4392
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7401096930743753e-06,
+      "loss": 0.5365,
+      "step": 4393
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7379920771734956e-06,
+      "loss": 1.0402,
+      "step": 4394
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7358747030161753e-06,
+      "loss": 0.9769,
+      "step": 4395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7337575710080067e-06,
+      "loss": 0.719,
+      "step": 4396
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7316406815545364e-06,
+      "loss": 0.4251,
+      "step": 4397
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7295240350612672e-06,
+      "loss": 0.961,
+      "step": 4398
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.727407631933647e-06,
+      "loss": 0.7088,
+      "step": 4399
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.72529147257709e-06,
+      "loss": 0.7738,
+      "step": 4400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7231755573969497e-06,
+      "loss": 0.712,
+      "step": 4401
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.721059886798545e-06,
+      "loss": 0.7918,
+      "step": 4402
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7189444611871383e-06,
+      "loss": 0.8783,
+      "step": 4403
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7168292809679533e-06,
+      "loss": 0.843,
+      "step": 4404
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7147143465461587e-06,
+      "loss": 0.8092,
+      "step": 4405
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.712599658326884e-06,
+      "loss": 0.3602,
+      "step": 4406
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7104852167152037e-06,
+      "loss": 0.5243,
+      "step": 4407
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.708371022116154e-06,
+      "loss": 1.017,
+      "step": 4408
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7062570749347126e-06,
+      "loss": 0.6368,
+      "step": 4409
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.704143375575823e-06,
+      "loss": 0.7613,
+      "step": 4410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.702029924444367e-06,
+      "loss": 0.4489,
+      "step": 4411
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.699916721945193e-06,
+      "loss": 0.5403,
+      "step": 4412
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6978037684830892e-06,
+      "loss": 0.8289,
+      "step": 4413
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6956910644628063e-06,
+      "loss": 0.7199,
+      "step": 4414
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.693578610289039e-06,
+      "loss": 0.506,
+      "step": 4415
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6914664063664417e-06,
+      "loss": 0.5551,
+      "step": 4416
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6893544530996124e-06,
+      "loss": 1.0143,
+      "step": 4417
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.68724275089311e-06,
+      "loss": 0.8892,
+      "step": 4418
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6851313001514378e-06,
+      "loss": 0.7982,
+      "step": 4419
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.683020101279057e-06,
+      "loss": 0.8246,
+      "step": 4420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.680909154680374e-06,
+      "loss": 0.9054,
+      "step": 4421
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.678798460759755e-06,
+      "loss": 0.7562,
+      "step": 4422
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.676688019921508e-06,
+      "loss": 0.4455,
+      "step": 4423
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6745778325699024e-06,
+      "loss": 0.7735,
+      "step": 4424
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6724678991091543e-06,
+      "loss": 0.5623,
+      "step": 4425
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6703582199434285e-06,
+      "loss": 0.8535,
+      "step": 4426
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6682487954768477e-06,
+      "loss": 0.493,
+      "step": 4427
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6661396261134773e-06,
+      "loss": 0.7001,
+      "step": 4428
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.664030712257344e-06,
+      "loss": 0.7699,
+      "step": 4429
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6619220543124156e-06,
+      "loss": 0.8608,
+      "step": 4430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.659813652682619e-06,
+      "loss": 0.4571,
+      "step": 4431
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.657705507771825e-06,
+      "loss": 0.9186,
+      "step": 4432
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6555976199838615e-06,
+      "loss": 0.3195,
+      "step": 4433
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.653489989722503e-06,
+      "loss": 0.9551,
+      "step": 4434
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6513826173914773e-06,
+      "loss": 0.1685,
+      "step": 4435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6492755033944587e-06,
+      "loss": 0.6551,
+      "step": 4436
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6471686481350787e-06,
+      "loss": 0.6943,
+      "step": 4437
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.645062052016911e-06,
+      "loss": 0.4036,
+      "step": 4438
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.642955715443489e-06,
+      "loss": 0.7213,
+      "step": 4439
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6408496388182857e-06,
+      "loss": 0.7226,
+      "step": 4440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6387438225447357e-06,
+      "loss": 0.3716,
+      "step": 4441
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6366382670262136e-06,
+      "loss": 0.7235,
+      "step": 4442
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6345329726660515e-06,
+      "loss": 0.7542,
+      "step": 4443
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6324279398675248e-06,
+      "loss": 0.6998,
+      "step": 4444
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6303231690338676e-06,
+      "loss": 0.7815,
+      "step": 4445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6282186605682534e-06,
+      "loss": 0.8849,
+      "step": 4446
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6261144148738147e-06,
+      "loss": 0.6432,
+      "step": 4447
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.624010432353626e-06,
+      "loss": 0.5916,
+      "step": 4448
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.62190671341072e-06,
+      "loss": 0.6247,
+      "step": 4449
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6198032584480692e-06,
+      "loss": 0.7107,
+      "step": 4450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6177000678686014e-06,
+      "loss": 0.8578,
+      "step": 4451
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6155971420751955e-06,
+      "loss": 0.7791,
+      "step": 4452
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.613494481470673e-06,
+      "loss": 0.916,
+      "step": 4453
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.611392086457812e-06,
+      "loss": 0.6928,
+      "step": 4454
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6092899574393338e-06,
+      "loss": 1.0211,
+      "step": 4455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.607188094817912e-06,
+      "loss": 0.6935,
+      "step": 4456
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6050864989961666e-06,
+      "loss": 0.6144,
+      "step": 4457
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6029851703766716e-06,
+      "loss": 0.7456,
+      "step": 4458
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6008841093619422e-06,
+      "loss": 1.0053,
+      "step": 4459
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5987833163544513e-06,
+      "loss": 0.8069,
+      "step": 4460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5966827917566106e-06,
+      "loss": 0.8474,
+      "step": 4461
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5945825359707897e-06,
+      "loss": 1.0607,
+      "step": 4462
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5924825493992986e-06,
+      "loss": 0.8589,
+      "step": 4463
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5903828324444047e-06,
+      "loss": 0.5726,
+      "step": 4464
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.588283385508313e-06,
+      "loss": 0.7337,
+      "step": 4465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5861842089931876e-06,
+      "loss": 1.0896,
+      "step": 4466
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5840853033011302e-06,
+      "loss": 0.5656,
+      "step": 4467
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.581986668834202e-06,
+      "loss": 0.7182,
+      "step": 4468
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.579888305994401e-06,
+      "loss": 1.029,
+      "step": 4469
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5777902151836834e-06,
+      "loss": 0.9508,
+      "step": 4470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.575692396803944e-06,
+      "loss": 0.7944,
+      "step": 4471
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.573594851257033e-06,
+      "loss": 0.5688,
+      "step": 4472
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5714975789447415e-06,
+      "loss": 0.6402,
+      "step": 4473
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5694005802688166e-06,
+      "loss": 0.9593,
+      "step": 4474
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5673038556309437e-06,
+      "loss": 1.0135,
+      "step": 4475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5652074054327646e-06,
+      "loss": 0.8808,
+      "step": 4476
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5631112300758595e-06,
+      "loss": 1.1677,
+      "step": 4477
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5610153299617656e-06,
+      "loss": 0.9603,
+      "step": 4478
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5589197054919577e-06,
+      "loss": 0.3041,
+      "step": 4479
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.556824357067866e-06,
+      "loss": 0.5761,
+      "step": 4480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.554729285090863e-06,
+      "loss": 0.7648,
+      "step": 4481
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.55263448996227e-06,
+      "loss": 0.3643,
+      "step": 4482
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5505399720833543e-06,
+      "loss": 0.595,
+      "step": 4483
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.548445731855331e-06,
+      "loss": 0.9822,
+      "step": 4484
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.546351769679362e-06,
+      "loss": 0.9005,
+      "step": 4485
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5442580859565556e-06,
+      "loss": 0.5548,
+      "step": 4486
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5421646810879655e-06,
+      "loss": 0.679,
+      "step": 4487
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.540071555474595e-06,
+      "loss": 0.8885,
+      "step": 4488
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.537978709517391e-06,
+      "loss": 0.6442,
+      "step": 4489
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5358861436172487e-06,
+      "loss": 0.9069,
+      "step": 4490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5337938581750088e-06,
+      "loss": 0.4118,
+      "step": 4491
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5317018535914575e-06,
+      "loss": 0.6223,
+      "step": 4492
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5296101302673293e-06,
+      "loss": 0.58,
+      "step": 4493
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.527518688603303e-06,
+      "loss": 0.7434,
+      "step": 4494
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.525427529000005e-06,
+      "loss": 0.7771,
+      "step": 4495
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5233366518580057e-06,
+      "loss": 0.5987,
+      "step": 4496
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.521246057577823e-06,
+      "loss": 0.9681,
+      "step": 4497
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5191557465599196e-06,
+      "loss": 0.5906,
+      "step": 4498
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5170657192047054e-06,
+      "loss": 0.9704,
+      "step": 4499
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5149759759125336e-06,
+      "loss": 0.7101,
+      "step": 4500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5128865170837055e-06,
+      "loss": 0.7738,
+      "step": 4501
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.510797343118466e-06,
+      "loss": 0.708,
+      "step": 4502
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.508708454417007e-06,
+      "loss": 0.9847,
+      "step": 4503
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.506619851379464e-06,
+      "loss": 1.1393,
+      "step": 4504
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.50453153440592e-06,
+      "loss": 0.7797,
+      "step": 4505
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.502443503896401e-06,
+      "loss": 1.198,
+      "step": 4506
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.500355760250878e-06,
+      "loss": 0.8351,
+      "step": 4507
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.498268303869272e-06,
+      "loss": 0.8997,
+      "step": 4508
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49618113515144e-06,
+      "loss": 0.9525,
+      "step": 4509
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4940942544971938e-06,
+      "loss": 0.8733,
+      "step": 4510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.492007662306281e-06,
+      "loss": 0.724,
+      "step": 4511
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.489921358978402e-06,
+      "loss": 1.0603,
+      "step": 4512
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.487835344913194e-06,
+      "loss": 0.8244,
+      "step": 4513
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4857496205102475e-06,
+      "loss": 0.6326,
+      "step": 4514
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4836641861690878e-06,
+      "loss": 0.6369,
+      "step": 4515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4815790422891942e-06,
+      "loss": 0.9585,
+      "step": 4516
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.479494189269982e-06,
+      "loss": 1.0275,
+      "step": 4517
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4774096275108183e-06,
+      "loss": 0.735,
+      "step": 4518
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.475325357411007e-06,
+      "loss": 0.5447,
+      "step": 4519
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4732413793698026e-06,
+      "loss": 0.8402,
+      "step": 4520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.471157693786399e-06,
+      "loss": 0.8917,
+      "step": 4521
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.469074301059939e-06,
+      "loss": 0.8483,
+      "step": 4522
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4669912015895012e-06,
+      "loss": 0.8238,
+      "step": 4523
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.464908395774119e-06,
+      "loss": 0.6734,
+      "step": 4524
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.462825884012759e-06,
+      "loss": 1.1029,
+      "step": 4525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.460743666704339e-06,
+      "loss": 0.9999,
+      "step": 4526
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4586617442477154e-06,
+      "loss": 0.9185,
+      "step": 4527
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.456580117041693e-06,
+      "loss": 0.8562,
+      "step": 4528
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.454498785485013e-06,
+      "loss": 0.8286,
+      "step": 4529
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.45241774997637e-06,
+      "loss": 1.0948,
+      "step": 4530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4503370109143895e-06,
+      "loss": 0.6595,
+      "step": 4531
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4482565686976537e-06,
+      "loss": 0.6516,
+      "step": 4532
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4461764237246755e-06,
+      "loss": 0.7236,
+      "step": 4533
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4440965763939204e-06,
+      "loss": 0.6971,
+      "step": 4534
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4420170271037894e-06,
+      "loss": 0.774,
+      "step": 4535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.439937776252632e-06,
+      "loss": 0.8248,
+      "step": 4536
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.43785882423874e-06,
+      "loss": 0.8668,
+      "step": 4537
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.435780171460343e-06,
+      "loss": 0.8996,
+      "step": 4538
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.43370181831562e-06,
+      "loss": 0.6995,
+      "step": 4539
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.431623765202686e-06,
+      "loss": 0.8799,
+      "step": 4540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.429546012519607e-06,
+      "loss": 0.5582,
+      "step": 4541
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4274685606643797e-06,
+      "loss": 0.9753,
+      "step": 4542
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.425391410034955e-06,
+      "loss": 0.6441,
+      "step": 4543
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4233145610292172e-06,
+      "loss": 0.9559,
+      "step": 4544
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4212380140450007e-06,
+      "loss": 0.6337,
+      "step": 4545
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.419161769480073e-06,
+      "loss": 0.9084,
+      "step": 4546
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4170858277321526e-06,
+      "loss": 0.7128,
+      "step": 4547
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.415010189198893e-06,
+      "loss": 0.6552,
+      "step": 4548
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.412934854277896e-06,
+      "loss": 0.8938,
+      "step": 4549
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.410859823366697e-06,
+      "loss": 0.7383,
+      "step": 4550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4087850968627823e-06,
+      "loss": 0.8133,
+      "step": 4551
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4067106751635716e-06,
+      "loss": 0.229,
+      "step": 4552
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4046365586664343e-06,
+      "loss": 0.4885,
+      "step": 4553
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.402562747768674e-06,
+      "loss": 0.6879,
+      "step": 4554
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4004892428675418e-06,
+      "loss": 0.9153,
+      "step": 4555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3984160443602233e-06,
+      "loss": 0.6559,
+      "step": 4556
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3963431526438533e-06,
+      "loss": 0.9279,
+      "step": 4557
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.394270568115501e-06,
+      "loss": 0.782,
+      "step": 4558
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.392198291172183e-06,
+      "loss": 0.363,
+      "step": 4559
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3901263222108506e-06,
+      "loss": 1.0114,
+      "step": 4560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3880546616284014e-06,
+      "loss": 1.0058,
+      "step": 4561
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.385983309821669e-06,
+      "loss": 0.7372,
+      "step": 4562
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3839122671874323e-06,
+      "loss": 0.569,
+      "step": 4563
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.381841534122412e-06,
+      "loss": 0.8893,
+      "step": 4564
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3797711110232622e-06,
+      "loss": 0.728,
+      "step": 4565
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3777009982865854e-06,
+      "loss": 0.7432,
+      "step": 4566
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3756311963089184e-06,
+      "loss": 0.7381,
+      "step": 4567
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3735617054867455e-06,
+      "loss": 0.9246,
+      "step": 4568
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3714925262164834e-06,
+      "loss": 0.8492,
+      "step": 4569
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3694236588944964e-06,
+      "loss": 0.3454,
+      "step": 4570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.367355103917082e-06,
+      "loss": 0.297,
+      "step": 4571
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3652868616804867e-06,
+      "loss": 0.6395,
+      "step": 4572
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3632189325808866e-06,
+      "loss": 0.5219,
+      "step": 4573
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3611513170144085e-06,
+      "loss": 1.2709,
+      "step": 4574
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3590840153771087e-06,
+      "loss": 0.7116,
+      "step": 4575
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.357017028064994e-06,
+      "loss": 0.8284,
+      "step": 4576
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3549503554739997e-06,
+      "loss": 0.9094,
+      "step": 4577
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3528839980000126e-06,
+      "loss": 0.5502,
+      "step": 4578
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3508179560388476e-06,
+      "loss": 1.141,
+      "step": 4579
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.34875222998627e-06,
+      "loss": 0.9733,
+      "step": 4580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3466868202379745e-06,
+      "loss": 0.7851,
+      "step": 4581
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.344621727189605e-06,
+      "loss": 0.8428,
+      "step": 4582
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3425569512367345e-06,
+      "loss": 0.6256,
+      "step": 4583
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3404924927748855e-06,
+      "loss": 0.8915,
+      "step": 4584
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.33842835219951e-06,
+      "loss": 0.9194,
+      "step": 4585
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.336364529906009e-06,
+      "loss": 0.2334,
+      "step": 4586
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3343010262897125e-06,
+      "loss": 0.7637,
+      "step": 4587
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3322378417458985e-06,
+      "loss": 0.7442,
+      "step": 4588
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3301749766697754e-06,
+      "loss": 0.8055,
+      "step": 4589
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3281124314564993e-06,
+      "loss": 0.9961,
+      "step": 4590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3260502065011557e-06,
+      "loss": 0.9537,
+      "step": 4591
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.323988302198778e-06,
+      "loss": 0.4,
+      "step": 4592
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3219267189443313e-06,
+      "loss": 0.6779,
+      "step": 4593
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3198654571327217e-06,
+      "loss": 1.0971,
+      "step": 4594
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3178045171587937e-06,
+      "loss": 0.8293,
+      "step": 4595
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3157438994173306e-06,
+      "loss": 0.7335,
+      "step": 4596
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.313683604303053e-06,
+      "loss": 0.7116,
+      "step": 4597
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3116236322106198e-06,
+      "loss": 0.8806,
+      "step": 4598
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3095639835346293e-06,
+      "loss": 0.5848,
+      "step": 4599
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.307504658669616e-06,
+      "loss": 0.8428,
+      "step": 4600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3054456580100534e-06,
+      "loss": 0.5125,
+      "step": 4601
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3033869819503533e-06,
+      "loss": 0.3606,
+      "step": 4602
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.301328630884864e-06,
+      "loss": 0.6535,
+      "step": 4603
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2992706052078732e-06,
+      "loss": 0.7608,
+      "step": 4604
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2972129053136047e-06,
+      "loss": 0.732,
+      "step": 4605
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2951555315962214e-06,
+      "loss": 1.3449,
+      "step": 4606
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2930984844498226e-06,
+      "loss": 0.4454,
+      "step": 4607
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2910417642684444e-06,
+      "loss": 0.6669,
+      "step": 4608
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2889853714460625e-06,
+      "loss": 0.9195,
+      "step": 4609
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.286929306376588e-06,
+      "loss": 0.698,
+      "step": 4610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.28487356945387e-06,
+      "loss": 0.9643,
+      "step": 4611
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.282818161071695e-06,
+      "loss": 0.8932,
+      "step": 4612
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2807630816237845e-06,
+      "loss": 0.6707,
+      "step": 4613
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.278708331503801e-06,
+      "loss": 0.2686,
+      "step": 4614
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.27665391110534e-06,
+      "loss": 1.2234,
+      "step": 4615
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2745998208219364e-06,
+      "loss": 0.9457,
+      "step": 4616
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2725460610470596e-06,
+      "loss": 0.9311,
+      "step": 4617
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.270492632174118e-06,
+      "loss": 0.6018,
+      "step": 4618
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.268439534596454e-06,
+      "loss": 1.066,
+      "step": 4619
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2663867687073513e-06,
+      "loss": 0.4212,
+      "step": 4620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2643343349000234e-06,
+      "loss": 0.9939,
+      "step": 4621
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2622822335676268e-06,
+      "loss": 0.8158,
+      "step": 4622
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2602304651032474e-06,
+      "loss": 0.8331,
+      "step": 4623
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2581790298999154e-06,
+      "loss": 1.0519,
+      "step": 4624
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2561279283505888e-06,
+      "loss": 0.9624,
+      "step": 4625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2540771608481697e-06,
+      "loss": 1.0789,
+      "step": 4626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.252026727785488e-06,
+      "loss": 0.8394,
+      "step": 4627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2499766295553182e-06,
+      "loss": 0.785,
+      "step": 4628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2479268665503623e-06,
+      "loss": 0.7393,
+      "step": 4629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2458774391632653e-06,
+      "loss": 0.6932,
+      "step": 4630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2438283477866016e-06,
+      "loss": 1.0724,
+      "step": 4631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2417795928128883e-06,
+      "loss": 0.5601,
+      "step": 4632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2397311746345694e-06,
+      "loss": 0.8126,
+      "step": 4633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.237683093644033e-06,
+      "loss": 0.8878,
+      "step": 4634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2356353502335953e-06,
+      "loss": 0.6137,
+      "step": 4635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2335879447955144e-06,
+      "loss": 0.7633,
+      "step": 4636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2315408777219777e-06,
+      "loss": 1.1312,
+      "step": 4637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2294941494051134e-06,
+      "loss": 0.7949,
+      "step": 4638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.227447760236978e-06,
+      "loss": 0.8587,
+      "step": 4639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225401710609572e-06,
+      "loss": 0.6936,
+      "step": 4640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2233560009148205e-06,
+      "loss": 0.8176,
+      "step": 4641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.221310631544594e-06,
+      "loss": 0.7335,
+      "step": 4642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2192656028906875e-06,
+      "loss": 0.303,
+      "step": 4643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2172209153448407e-06,
+      "loss": 0.724,
+      "step": 4644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.215176569298719e-06,
+      "loss": 0.794,
+      "step": 4645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2131325651439288e-06,
+      "loss": 1.0474,
+      "step": 4646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.21108890327201e-06,
+      "loss": 0.5965,
+      "step": 4647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2090455840744316e-06,
+      "loss": 0.4412,
+      "step": 4648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.207002607942605e-06,
+      "loss": 1.147,
+      "step": 4649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2049599752678695e-06,
+      "loss": 0.613,
+      "step": 4650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2029176864415035e-06,
+      "loss": 1.1011,
+      "step": 4651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.200875741854713e-06,
+      "loss": 1.0664,
+      "step": 4652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1988341418986463e-06,
+      "loss": 0.6319,
+      "step": 4653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1967928869643778e-06,
+      "loss": 0.7804,
+      "step": 4654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.194751977442923e-06,
+      "loss": 0.8252,
+      "step": 4655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1927114137252247e-06,
+      "loss": 0.9402,
+      "step": 4656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.190671196202165e-06,
+      "loss": 0.6718,
+      "step": 4657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.188631325264554e-06,
+      "loss": 1.1024,
+      "step": 4658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1865918013031433e-06,
+      "loss": 0.6292,
+      "step": 4659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.184552624708608e-06,
+      "loss": 0.8596,
+      "step": 4660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.182513795871567e-06,
+      "loss": 0.6793,
+      "step": 4661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.180475315182563e-06,
+      "loss": 1.0315,
+      "step": 4662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1784371830320803e-06,
+      "loss": 1.0157,
+      "step": 4663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.176399399810529e-06,
+      "loss": 0.6284,
+      "step": 4664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1743619659082608e-06,
+      "loss": 0.8348,
+      "step": 4665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1723248817155506e-06,
+      "loss": 0.813,
+      "step": 4666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1702881476226165e-06,
+      "loss": 0.5529,
+      "step": 4667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1682517640195997e-06,
+      "loss": 0.6377,
+      "step": 4668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1662157312965834e-06,
+      "loss": 0.7436,
+      "step": 4669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1641800498435754e-06,
+      "loss": 1.1239,
+      "step": 4670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.162144720050524e-06,
+      "loss": 0.8731,
+      "step": 4671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1601097423073014e-06,
+      "loss": 0.6712,
+      "step": 4672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.158075117003723e-06,
+      "loss": 0.4079,
+      "step": 4673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.156040844529524e-06,
+      "loss": 0.6876,
+      "step": 4674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.154006925274384e-06,
+      "loss": 0.6656,
+      "step": 4675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1519733596279103e-06,
+      "loss": 0.8783,
+      "step": 4676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.149940147979638e-06,
+      "loss": 0.6011,
+      "step": 4677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.147907290719041e-06,
+      "loss": 0.8396,
+      "step": 4678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1458747882355216e-06,
+      "loss": 1.0898,
+      "step": 4679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1438426409184175e-06,
+      "loss": 1.2109,
+      "step": 4680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.141810849156992e-06,
+      "loss": 0.9154,
+      "step": 4681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.139779413340449e-06,
+      "loss": 0.7546,
+      "step": 4682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.137748333857915e-06,
+      "loss": 0.8074,
+      "step": 4683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1357176110984578e-06,
+      "loss": 0.8636,
+      "step": 4684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.133687245451067e-06,
+      "loss": 0.547,
+      "step": 4685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.131657237304672e-06,
+      "loss": 0.6755,
+      "step": 4686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1296275870481287e-06,
+      "loss": 0.6248,
+      "step": 4687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1275982950702277e-06,
+      "loss": 0.8447,
+      "step": 4688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1255693617596872e-06,
+      "loss": 0.6351,
+      "step": 4689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1235407875051616e-06,
+      "loss": 0.728,
+      "step": 4690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.121512572695231e-06,
+      "loss": 1.0226,
+      "step": 4691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.119484717718413e-06,
+      "loss": 0.7859,
+      "step": 4692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1174572229631487e-06,
+      "loss": 0.5713,
+      "step": 4693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.115430088817819e-06,
+      "loss": 0.9362,
+      "step": 4694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1134033156707256e-06,
+      "loss": 0.9344,
+      "step": 4695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1113769039101112e-06,
+      "loss": 0.76,
+      "step": 4696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1093508539241402e-06,
+      "loss": 0.7907,
+      "step": 4697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1073251661009175e-06,
+      "loss": 0.8465,
+      "step": 4698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1052998408284664e-06,
+      "loss": 0.9665,
+      "step": 4699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.103274878494754e-06,
+      "loss": 0.7837,
+      "step": 4700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1012502794876665e-06,
+      "loss": 0.8346,
+      "step": 4701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.099226044195027e-06,
+      "loss": 0.521,
+      "step": 4702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.097202173004589e-06,
+      "loss": 1.0825,
+      "step": 4703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0951786663040327e-06,
+      "loss": 0.9065,
+      "step": 4704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0931555244809706e-06,
+      "loss": 0.4278,
+      "step": 4705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.091132747922946e-06,
+      "loss": 0.8251,
+      "step": 4706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0891103370174293e-06,
+      "loss": 0.8829,
+      "step": 4707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0870882921518242e-06,
+      "loss": 0.7915,
+      "step": 4708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0850666137134633e-06,
+      "loss": 0.7044,
+      "step": 4709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.083045302089608e-06,
+      "loss": 0.7843,
+      "step": 4710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0810243576674492e-06,
+      "loss": 0.8981,
+      "step": 4711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.07900378083411e-06,
+      "loss": 1.0027,
+      "step": 4712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0769835719766406e-06,
+      "loss": 1.1513,
+      "step": 4713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0749637314820215e-06,
+      "loss": 0.8357,
+      "step": 4714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.072944259737163e-06,
+      "loss": 0.7976,
+      "step": 4715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0709251571289034e-06,
+      "loss": 0.9096,
+      "step": 4716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0689064240440126e-06,
+      "loss": 0.9688,
+      "step": 4717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0668880608691876e-06,
+      "loss": 0.7445,
+      "step": 4718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.064870067991056e-06,
+      "loss": 0.4218,
+      "step": 4719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.062852445796173e-06,
+      "loss": 0.5518,
+      "step": 4720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.060835194671024e-06,
+      "loss": 0.9897,
+      "step": 4721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0588183150020234e-06,
+      "loss": 0.6157,
+      "step": 4722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0568018071755133e-06,
+      "loss": 0.6546,
+      "step": 4723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0547856715777656e-06,
+      "loss": 0.6564,
+      "step": 4724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0527699085949813e-06,
+      "loss": 0.9605,
+      "step": 4725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.050754518613288e-06,
+      "loss": 0.706,
+      "step": 4726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0487395020187436e-06,
+      "loss": 0.54,
+      "step": 4727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.046724859197334e-06,
+      "loss": 0.7302,
+      "step": 4728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0447105905349745e-06,
+      "loss": 0.7175,
+      "step": 4729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0426966964175063e-06,
+      "loss": 0.9219,
+      "step": 4730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0406831772307e-06,
+      "loss": 0.6713,
+      "step": 4731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.038670033360257e-06,
+      "loss": 0.6691,
+      "step": 4732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0366572651918013e-06,
+      "loss": 0.7203,
+      "step": 4733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0346448731108925e-06,
+      "loss": 0.9643,
+      "step": 4734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0326328575030085e-06,
+      "loss": 0.9608,
+      "step": 4735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0306212187535653e-06,
+      "loss": 0.7207,
+      "step": 4736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.028609957247897e-06,
+      "loss": 0.7411,
+      "step": 4737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.026599073371275e-06,
+      "loss": 0.622,
+      "step": 4738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0245885675088893e-06,
+      "loss": 0.8937,
+      "step": 4739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0225784400458658e-06,
+      "loss": 0.7033,
+      "step": 4740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.02056869136725e-06,
+      "loss": 0.9558,
+      "step": 4741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0185593218580223e-06,
+      "loss": 0.9509,
+      "step": 4742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0165503319030832e-06,
+      "loss": 0.899,
+      "step": 4743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0145417218872676e-06,
+      "loss": 0.575,
+      "step": 4744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0125334921953308e-06,
+      "loss": 0.4234,
+      "step": 4745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0105256432119634e-06,
+      "loss": 0.8082,
+      "step": 4746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.008518175321773e-06,
+      "loss": 0.4522,
+      "step": 4747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0065110889093033e-06,
+      "loss": 1.1657,
+      "step": 4748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0045043843590187e-06,
+      "loss": 0.459,
+      "step": 4749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0024980620553155e-06,
+      "loss": 0.6995,
+      "step": 4750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.000492122382511e-06,
+      "loss": 0.9045,
+      "step": 4751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.998486565724857e-06,
+      "loss": 1.159,
+      "step": 4752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9964813924665215e-06,
+      "loss": 0.5637,
+      "step": 4753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.99447660299161e-06,
+      "loss": 0.9523,
+      "step": 4754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9924721976841455e-06,
+      "loss": 0.8881,
+      "step": 4755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.990468176928084e-06,
+      "loss": 0.8003,
+      "step": 4756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9884645411073033e-06,
+      "loss": 0.7949,
+      "step": 4757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.986461290605609e-06,
+      "loss": 0.7532,
+      "step": 4758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9844584258067364e-06,
+      "loss": 1.103,
+      "step": 4759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.98245594709434e-06,
+      "loss": 0.7375,
+      "step": 4760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9804538548520078e-06,
+      "loss": 0.3949,
+      "step": 4761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.978452149463245e-06,
+      "loss": 0.8641,
+      "step": 4762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9764508313114924e-06,
+      "loss": 0.9663,
+      "step": 4763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9744499007801085e-06,
+      "loss": 0.8066,
+      "step": 4764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.972449358252384e-06,
+      "loss": 0.724,
+      "step": 4765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.970449204111529e-06,
+      "loss": 1.1089,
+      "step": 4766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.968449438740686e-06,
+      "loss": 1.0067,
+      "step": 4767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.966450062522915e-06,
+      "loss": 0.7901,
+      "step": 4768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.964451075841211e-06,
+      "loss": 0.615,
+      "step": 4769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.962452479078485e-06,
+      "loss": 0.912,
+      "step": 4770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9604542726175806e-06,
+      "loss": 0.8372,
+      "step": 4771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.958456456841261e-06,
+      "loss": 0.9574,
+      "step": 4772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9564590321322206e-06,
+      "loss": 0.9591,
+      "step": 4773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.954461998873071e-06,
+      "loss": 0.5389,
+      "step": 4774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9524653574463585e-06,
+      "loss": 0.7326,
+      "step": 4775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.950469108234545e-06,
+      "loss": 1.1684,
+      "step": 4776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.948473251620024e-06,
+      "loss": 0.7971,
+      "step": 4777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9464777879851087e-06,
+      "loss": 0.8696,
+      "step": 4778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.944482717712043e-06,
+      "loss": 0.6821,
+      "step": 4779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9424880411829883e-06,
+      "loss": 0.5027,
+      "step": 4780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9404937587800374e-06,
+      "loss": 0.8261,
+      "step": 4781
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.938499870885202e-06,
+      "loss": 1.1677,
+      "step": 4782
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9365063778804237e-06,
+      "loss": 0.8299,
+      "step": 4783
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.934513280147562e-06,
+      "loss": 1.0148,
+      "step": 4784
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9325205780684053e-06,
+      "loss": 0.7041,
+      "step": 4785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9305282720246675e-06,
+      "loss": 0.8214,
+      "step": 4786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.92853636239798e-06,
+      "loss": 0.6097,
+      "step": 4787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.926544849569907e-06,
+      "loss": 0.6839,
+      "step": 4788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9245537339219277e-06,
+      "loss": 0.6077,
+      "step": 4789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9225630158354544e-06,
+      "loss": 0.6806,
+      "step": 4790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9205726956918135e-06,
+      "loss": 0.5442,
+      "step": 4791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.918582773872266e-06,
+      "loss": 0.919,
+      "step": 4792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.916593250757984e-06,
+      "loss": 0.6835,
+      "step": 4793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9146041267300763e-06,
+      "loss": 0.8628,
+      "step": 4794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9126154021695633e-06,
+      "loss": 0.8202,
+      "step": 4795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9106270774574e-06,
+      "loss": 0.6231,
+      "step": 4796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9086391529744544e-06,
+      "loss": 0.999,
+      "step": 4797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9066516291015277e-06,
+      "loss": 1.0614,
+      "step": 4798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9046645062193346e-06,
+      "loss": 0.8117,
+      "step": 4799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.902677784708522e-06,
+      "loss": 0.9561,
+      "step": 4800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9006914649496522e-06,
+      "loss": 1.0902,
+      "step": 4801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8987055473232174e-06,
+      "loss": 0.5432,
+      "step": 4802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8967200322096262e-06,
+      "loss": 0.7872,
+      "step": 4803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8947349199892173e-06,
+      "loss": 1.0713,
+      "step": 4804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.892750211042243e-06,
+      "loss": 0.8534,
+      "step": 4805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.89076590574889e-06,
+      "loss": 0.8825,
+      "step": 4806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8887820044892557e-06,
+      "loss": 0.8147,
+      "step": 4807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8867985076433696e-06,
+      "loss": 0.6258,
+      "step": 4808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.884815415591177e-06,
+      "loss": 0.758,
+      "step": 4809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.882832728712551e-06,
+      "loss": 0.5434,
+      "step": 4810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8808504473872823e-06,
+      "loss": 0.6862,
+      "step": 4811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.878868571995088e-06,
+      "loss": 0.8027,
+      "step": 4812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8768871029156044e-06,
+      "loss": 1.217,
+      "step": 4813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.874906040528391e-06,
+      "loss": 0.9744,
+      "step": 4814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8729253852129336e-06,
+      "loss": 0.4231,
+      "step": 4815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8709451373486295e-06,
+      "loss": 1.1473,
+      "step": 4816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8689652973148106e-06,
+      "loss": 1.0601,
+      "step": 4817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.86698586549072e-06,
+      "loss": 0.5458,
+      "step": 4818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.865006842255531e-06,
+      "loss": 0.8926,
+      "step": 4819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8630282279883307e-06,
+      "loss": 0.7267,
+      "step": 4820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8610500230681367e-06,
+      "loss": 0.9197,
+      "step": 4821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.859072227873878e-06,
+      "loss": 0.6855,
+      "step": 4822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8570948427844163e-06,
+      "loss": 0.4189,
+      "step": 4823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855117868178524e-06,
+      "loss": 0.4115,
+      "step": 4824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.853141304434904e-06,
+      "loss": 0.8542,
+      "step": 4825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8511651519321735e-06,
+      "loss": 1.1288,
+      "step": 4826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.849189411048876e-06,
+      "loss": 1.2013,
+      "step": 4827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8472140821634707e-06,
+      "loss": 0.7076,
+      "step": 4828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8452391656543456e-06,
+      "loss": 0.7777,
+      "step": 4829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.843264661899801e-06,
+      "loss": 0.8438,
+      "step": 4830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8412905712780657e-06,
+      "loss": 1.0514,
+      "step": 4831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8393168941672832e-06,
+      "loss": 0.7618,
+      "step": 4832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8373436309455243e-06,
+      "loss": 0.6801,
+      "step": 4833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.835370781990773e-06,
+      "loss": 0.9877,
+      "step": 4834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8333983476809403e-06,
+      "loss": 0.913,
+      "step": 4835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8314263283938533e-06,
+      "loss": 1.0282,
+      "step": 4836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8294547245072646e-06,
+      "loss": 0.6553,
+      "step": 4837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8274835363988403e-06,
+      "loss": 1.1042,
+      "step": 4838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.825512764446174e-06,
+      "loss": 0.8233,
+      "step": 4839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8235424090267744e-06,
+      "loss": 0.6135,
+      "step": 4840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8215724705180726e-06,
+      "loss": 0.8539,
+      "step": 4841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.819602949297422e-06,
+      "loss": 0.7793,
+      "step": 4842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.81763384574209e-06,
+      "loss": 0.7757,
+      "step": 4843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8156651602292713e-06,
+      "loss": 1.0794,
+      "step": 4844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.813696893136073e-06,
+      "loss": 0.8445,
+      "step": 4845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8117290448395295e-06,
+      "loss": 0.4959,
+      "step": 4846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8097616157165886e-06,
+      "loss": 1.0085,
+      "step": 4847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.807794606144123e-06,
+      "loss": 0.7569,
+      "step": 4848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8058280164989205e-06,
+      "loss": 0.8506,
+      "step": 4849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8038618471576927e-06,
+      "loss": 0.8841,
+      "step": 4850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8018960984970663e-06,
+      "loss": 0.8939,
+      "step": 4851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.799930770893592e-06,
+      "loss": 0.993,
+      "step": 4852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.797965864723735e-06,
+      "loss": 0.8381,
+      "step": 4853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7960013803638857e-06,
+      "loss": 0.6743,
+      "step": 4854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.794037318190347e-06,
+      "loss": 0.8617,
+      "step": 4855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7920736785793468e-06,
+      "loss": 0.8016,
+      "step": 4856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.790110461907026e-06,
+      "loss": 0.7181,
+      "step": 4857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7881476685494535e-06,
+      "loss": 0.6662,
+      "step": 4858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.786185298882606e-06,
+      "loss": 0.8329,
+      "step": 4859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.784223353282389e-06,
+      "loss": 0.6757,
+      "step": 4860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7822618321246185e-06,
+      "loss": 0.6515,
+      "step": 4861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7803007357850374e-06,
+      "loss": 1.0346,
+      "step": 4862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7783400646392988e-06,
+      "loss": 0.8389,
+      "step": 4863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.776379819062982e-06,
+      "loss": 1.0975,
+      "step": 4864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.774419999431578e-06,
+      "loss": 0.5689,
+      "step": 4865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.772460606120503e-06,
+      "loss": 0.8149,
+      "step": 4866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.770501639505084e-06,
+      "loss": 0.6736,
+      "step": 4867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7685430999605743e-06,
+      "loss": 0.8131,
+      "step": 4868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7665849878621387e-06,
+      "loss": 0.7227,
+      "step": 4869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7646273035848625e-06,
+      "loss": 1.0103,
+      "step": 4870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7626700475037526e-06,
+      "loss": 0.9853,
+      "step": 4871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.760713219993726e-06,
+      "loss": 0.8592,
+      "step": 4872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.758756821429627e-06,
+      "loss": 0.5175,
+      "step": 4873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.756800852186209e-06,
+      "loss": 1.0345,
+      "step": 4874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7548453126381503e-06,
+      "loss": 0.6935,
+      "step": 4875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.75289020316004e-06,
+      "loss": 0.388,
+      "step": 4876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7509355241263937e-06,
+      "loss": 0.7848,
+      "step": 4877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.748981275911633e-06,
+      "loss": 0.886,
+      "step": 4878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7470274588901096e-06,
+      "loss": 0.7669,
+      "step": 4879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7450740734360814e-06,
+      "loss": 0.8568,
+      "step": 4880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.743121119923732e-06,
+      "loss": 1.1535,
+      "step": 4881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7411685987271565e-06,
+      "loss": 0.7021,
+      "step": 4882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7392165102203716e-06,
+      "loss": 0.9612,
+      "step": 4883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7372648547773063e-06,
+      "loss": 0.5415,
+      "step": 4884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.735313632771812e-06,
+      "loss": 0.6933,
+      "step": 4885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7333628445776526e-06,
+      "loss": 0.9193,
+      "step": 4886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7314124905685137e-06,
+      "loss": 0.9142,
+      "step": 4887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.72946257111799e-06,
+      "loss": 0.7602,
+      "step": 4888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.727513086599603e-06,
+      "loss": 0.8625,
+      "step": 4889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7255640373867807e-06,
+      "loss": 0.671,
+      "step": 4890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7236154238528777e-06,
+      "loss": 1.0277,
+      "step": 4891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.721667246371155e-06,
+      "loss": 0.4252,
+      "step": 4892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7197195053147997e-06,
+      "loss": 0.8055,
+      "step": 4893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7177722010569074e-06,
+      "loss": 0.4726,
+      "step": 4894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7158253339704954e-06,
+      "loss": 0.8679,
+      "step": 4895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7138789044284937e-06,
+      "loss": 0.6142,
+      "step": 4896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7119329128037506e-06,
+      "loss": 0.6057,
+      "step": 4897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7099873594690317e-06,
+      "loss": 0.8708,
+      "step": 4898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7080422447970145e-06,
+      "loss": 1.1654,
+      "step": 4899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.706097569160297e-06,
+      "loss": 0.6376,
+      "step": 4900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7041533329313885e-06,
+      "loss": 0.8711,
+      "step": 4901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7022095364827194e-06,
+      "loss": 0.478,
+      "step": 4902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7002661801866292e-06,
+      "loss": 1.0687,
+      "step": 4903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.698323264415382e-06,
+      "loss": 0.2702,
+      "step": 4904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6963807895411476e-06,
+      "loss": 0.9223,
+      "step": 4905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.69443875593602e-06,
+      "loss": 0.5086,
+      "step": 4906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.692497163972002e-06,
+      "loss": 0.6761,
+      "step": 4907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6905560140210174e-06,
+      "loss": 0.793,
+      "step": 4908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6886153064548993e-06,
+      "loss": 0.9615,
+      "step": 4909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6866750416454036e-06,
+      "loss": 0.7862,
+      "step": 4910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6847352199641923e-06,
+      "loss": 0.7046,
+      "step": 4911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6827958417828524e-06,
+      "loss": 0.7284,
+      "step": 4912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.680856907472876e-06,
+      "loss": 0.7289,
+      "step": 4913
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6789184174056793e-06,
+      "loss": 0.8956,
+      "step": 4914
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6769803719525855e-06,
+      "loss": 0.7264,
+      "step": 4915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6750427714848404e-06,
+      "loss": 0.6068,
+      "step": 4916
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.673105616373597e-06,
+      "loss": 0.6852,
+      "step": 4917
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6711689069899287e-06,
+      "loss": 0.7195,
+      "step": 4918
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.669232643704819e-06,
+      "loss": 0.9591,
+      "step": 4919
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6672968268891716e-06,
+      "loss": 0.9935,
+      "step": 4920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.665361456913797e-06,
+      "loss": 1.0496,
+      "step": 4921
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.663426534149429e-06,
+      "loss": 0.7845,
+      "step": 4922
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.661492058966707e-06,
+      "loss": 1.03,
+      "step": 4923
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.659558031736191e-06,
+      "loss": 0.6831,
+      "step": 4924
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6576244528283545e-06,
+      "loss": 0.8024,
+      "step": 4925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6556913226135796e-06,
+      "loss": 0.59,
+      "step": 4926
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6537586414621718e-06,
+      "loss": 1.019,
+      "step": 4927
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6518264097443392e-06,
+      "loss": 0.4527,
+      "step": 4928
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6498946278302163e-06,
+      "loss": 0.875,
+      "step": 4929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.647963296089839e-06,
+      "loss": 0.6577,
+      "step": 4930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6460324148931672e-06,
+      "loss": 0.9067,
+      "step": 4931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.644101984610068e-06,
+      "loss": 1.0836,
+      "step": 4932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.642172005610326e-06,
+      "loss": 1.0172,
+      "step": 4933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.640242478263636e-06,
+      "loss": 0.6281,
+      "step": 4934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.63831340293961e-06,
+      "loss": 1.2077,
+      "step": 4935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.636384780007769e-06,
+      "loss": 1.0402,
+      "step": 4936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6344566098375534e-06,
+      "loss": 0.7481,
+      "step": 4937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6325288927983095e-06,
+      "loss": 0.7163,
+      "step": 4938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.630601629259304e-06,
+      "loss": 0.8718,
+      "step": 4939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6286748195897096e-06,
+      "loss": 0.2705,
+      "step": 4940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.626748464158619e-06,
+      "loss": 0.7108,
+      "step": 4941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6248225633350317e-06,
+      "loss": 0.903,
+      "step": 4942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.622897117487867e-06,
+      "loss": 0.8301,
+      "step": 4943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6209721269859483e-06,
+      "loss": 1.0481,
+      "step": 4944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.619047592198022e-06,
+      "loss": 0.7694,
+      "step": 4945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6171235134927364e-06,
+      "loss": 0.4827,
+      "step": 4946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6151998912386622e-06,
+      "loss": 0.6225,
+      "step": 4947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6132767258042745e-06,
+      "loss": 1.0192,
+      "step": 4948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6113540175579686e-06,
+      "loss": 0.8153,
+      "step": 4949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6094317668680446e-06,
+      "loss": 0.8136,
+      "step": 4950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6075099741027226e-06,
+      "loss": 1.3107,
+      "step": 4951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6055886396301265e-06,
+      "loss": 1.0268,
+      "step": 4952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.603667763818302e-06,
+      "loss": 0.7309,
+      "step": 4953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6017473470351967e-06,
+      "loss": 0.6981,
+      "step": 4954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5998273896486805e-06,
+      "loss": 0.7504,
+      "step": 4955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5979078920265257e-06,
+      "loss": 1.1553,
+      "step": 4956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.595988854536424e-06,
+      "loss": 0.7386,
+      "step": 4957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.594070277545975e-06,
+      "loss": 0.5917,
+      "step": 4958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5921521614226922e-06,
+      "loss": 0.6763,
+      "step": 4959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5902345065339977e-06,
+      "loss": 0.7492,
+      "step": 4960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5883173132472306e-06,
+      "loss": 0.4463,
+      "step": 4961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5864005819296346e-06,
+      "loss": 0.8123,
+      "step": 4962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5844843129483722e-06,
+      "loss": 0.6219,
+      "step": 4963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5825685066705095e-06,
+      "loss": 1.3213,
+      "step": 4964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.580653163463033e-06,
+      "loss": 0.5703,
+      "step": 4965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.578738283692831e-06,
+      "loss": 0.7238,
+      "step": 4966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5768238677267122e-06,
+      "loss": 0.6056,
+      "step": 4967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5749099159313875e-06,
+      "loss": 0.8322,
+      "step": 4968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.572996428673488e-06,
+      "loss": 0.5763,
+      "step": 4969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5710834063195474e-06,
+      "loss": 0.8662,
+      "step": 4970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5691708492360168e-06,
+      "loss": 0.9307,
+      "step": 4971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5672587577892528e-06,
+      "loss": 0.5468,
+      "step": 4972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5653471323455294e-06,
+      "loss": 0.8714,
+      "step": 4973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5634359732710224e-06,
+      "loss": 0.4509,
+      "step": 4974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5615252809318287e-06,
+      "loss": 1.0879,
+      "step": 4975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.559615055693946e-06,
+      "loss": 0.9443,
+      "step": 4976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5577052979232898e-06,
+      "loss": 1.0056,
+      "step": 4977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.55579600798568e-06,
+      "loss": 0.844,
+      "step": 4978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.553887186246855e-06,
+      "loss": 0.634,
+      "step": 4979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5519788330724536e-06,
+      "loss": 0.668,
+      "step": 4980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.550070948828034e-06,
+      "loss": 0.7998,
+      "step": 4981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.548163533879056e-06,
+      "loss": 0.7075,
+      "step": 4982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5462565885908984e-06,
+      "loss": 0.6679,
+      "step": 4983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.544350113328842e-06,
+      "loss": 0.3308,
+      "step": 4984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5424441084580842e-06,
+      "loss": 0.5687,
+      "step": 4985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.540538574343726e-06,
+      "loss": 0.7158,
+      "step": 4986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.538633511350784e-06,
+      "loss": 1.0716,
+      "step": 4987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5367289198441803e-06,
+      "loss": 0.6779,
+      "step": 4988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.53482480018875e-06,
+      "loss": 0.8116,
+      "step": 4989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5329211527492337e-06,
+      "loss": 0.7789,
+      "step": 4990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.531017977890288e-06,
+      "loss": 0.8538,
+      "step": 4991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5291152759764703e-06,
+      "loss": 0.4755,
+      "step": 4992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5272130473722563e-06,
+      "loss": 0.9043,
+      "step": 4993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.525311292442023e-06,
+      "loss": 0.6079,
+      "step": 4994
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5234100115500643e-06,
+      "loss": 0.64,
+      "step": 4995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.521509205060576e-06,
+      "loss": 0.9584,
+      "step": 4996
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5196088733376696e-06,
+      "loss": 0.7328,
+      "step": 4997
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.51770901674536e-06,
+      "loss": 0.3852,
+      "step": 4998
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5158096356475758e-06,
+      "loss": 0.9528,
+      "step": 4999
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.51391073040815e-06,
+      "loss": 0.7723,
+      "step": 5000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.51201230139083e-06,
+      "loss": 0.7235,
+      "step": 5001
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.510114348959266e-06,
+      "loss": 0.8183,
+      "step": 5002
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5082168734770224e-06,
+      "loss": 0.8058,
+      "step": 5003
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5063198753075655e-06,
+      "loss": 0.2985,
+      "step": 5004
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5044233548142794e-06,
+      "loss": 0.5885,
+      "step": 5005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5025273123604466e-06,
+      "loss": 0.6975,
+      "step": 5006
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.500631748309267e-06,
+      "loss": 0.6351,
+      "step": 5007
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4987366630238414e-06,
+      "loss": 0.9379,
+      "step": 5008
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4968420568671843e-06,
+      "loss": 0.8975,
+      "step": 5009
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4949479302022173e-06,
+      "loss": 0.799,
+      "step": 5010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.493054283391766e-06,
+      "loss": 0.8594,
+      "step": 5011
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4911611167985716e-06,
+      "loss": 0.9533,
+      "step": 5012
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.489268430785275e-06,
+      "loss": 1.0177,
+      "step": 5013
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.487376225714433e-06,
+      "loss": 0.7477,
+      "step": 5014
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4854845019485026e-06,
+      "loss": 0.7154,
+      "step": 5015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4835932598498557e-06,
+      "loss": 0.5272,
+      "step": 5016
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4817024997807655e-06,
+      "loss": 0.9213,
+      "step": 5017
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4798122221034194e-06,
+      "loss": 0.6973,
+      "step": 5018
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4779224271799053e-06,
+      "loss": 0.7412,
+      "step": 5019
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.476033115372226e-06,
+      "loss": 0.8257,
+      "step": 5020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4741442870422843e-06,
+      "loss": 0.4978,
+      "step": 5021
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4722559425518984e-06,
+      "loss": 1.037,
+      "step": 5022
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.470368082262785e-06,
+      "loss": 0.951,
+      "step": 5023
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4684807065365774e-06,
+      "loss": 0.5918,
+      "step": 5024
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.466593815734806e-06,
+      "loss": 0.9449,
+      "step": 5025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4647074102189193e-06,
+      "loss": 0.9362,
+      "step": 5026
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4628214903502623e-06,
+      "loss": 1.0669,
+      "step": 5027
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4609360564900953e-06,
+      "loss": 0.6896,
+      "step": 5028
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.459051108999579e-06,
+      "loss": 1.1039,
+      "step": 5029
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4571666482397875e-06,
+      "loss": 0.9566,
+      "step": 5030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.455282674571694e-06,
+      "loss": 0.5077,
+      "step": 5031
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4533991883561868e-06,
+      "loss": 0.6664,
+      "step": 5032
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4515161899540525e-06,
+      "loss": 1.0382,
+      "step": 5033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4496336797259925e-06,
+      "loss": 0.8386,
+      "step": 5034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.447751658032606e-06,
+      "loss": 0.7713,
+      "step": 5035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4458701252344055e-06,
+      "loss": 0.6839,
+      "step": 5036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.443989081691809e-06,
+      "loss": 0.8904,
+      "step": 5037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.442108527765136e-06,
+      "loss": 0.8228,
+      "step": 5038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.440228463814619e-06,
+      "loss": 0.962,
+      "step": 5039
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4383488902003886e-06,
+      "loss": 0.6272,
+      "step": 5040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.436469807282491e-06,
+      "loss": 0.4359,
+      "step": 5041
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.434591215420869e-06,
+      "loss": 0.8323,
+      "step": 5042
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4327131149753796e-06,
+      "loss": 0.686,
+      "step": 5043
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4308355063057775e-06,
+      "loss": 0.5436,
+      "step": 5044
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4289583897717322e-06,
+      "loss": 0.5922,
+      "step": 5045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4270817657328095e-06,
+      "loss": 0.8786,
+      "step": 5046
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.42520563454849e-06,
+      "loss": 0.5276,
+      "step": 5047
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4233299965781514e-06,
+      "loss": 0.6597,
+      "step": 5048
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4214548521810844e-06,
+      "loss": 0.8839,
+      "step": 5049
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4195802017164787e-06,
+      "loss": 1.1534,
+      "step": 5050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.417706045543436e-06,
+      "loss": 0.7107,
+      "step": 5051
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4158323840209556e-06,
+      "loss": 0.9452,
+      "step": 5052
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4139592175079503e-06,
+      "loss": 0.6895,
+      "step": 5053
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4120865463632296e-06,
+      "loss": 0.776,
+      "step": 5054
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.410214370945518e-06,
+      "loss": 0.7821,
+      "step": 5055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4083426916134334e-06,
+      "loss": 0.8647,
+      "step": 5056
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4064715087255097e-06,
+      "loss": 0.9365,
+      "step": 5057
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.404600822640177e-06,
+      "loss": 0.7886,
+      "step": 5058
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.402730633715778e-06,
+      "loss": 0.6041,
+      "step": 5059
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4008609423105517e-06,
+      "loss": 1.0359,
+      "step": 5060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3989917487826508e-06,
+      "loss": 0.4948,
+      "step": 5061
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.397123053490124e-06,
+      "loss": 1.0782,
+      "step": 5062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3952548567909305e-06,
+      "loss": 0.761,
+      "step": 5063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3933871590429348e-06,
+      "loss": 0.6576,
+      "step": 5064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.391519960603898e-06,
+      "loss": 0.6453,
+      "step": 5065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.389653261831496e-06,
+      "loss": 0.6882,
+      "step": 5066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.387787063083299e-06,
+      "loss": 0.7941,
+      "step": 5067
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3859213647167905e-06,
+      "loss": 1.0155,
+      "step": 5068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.38405616708935e-06,
+      "loss": 0.9052,
+      "step": 5069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3821914705582676e-06,
+      "loss": 0.7096,
+      "step": 5070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.380327275480732e-06,
+      "loss": 0.3088,
+      "step": 5071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3784635822138424e-06,
+      "loss": 0.8552,
+      "step": 5072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.376600391114594e-06,
+      "loss": 0.5691,
+      "step": 5073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3747377025398934e-06,
+      "loss": 0.9904,
+      "step": 5074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3728755168465432e-06,
+      "loss": 1.0916,
+      "step": 5075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3710138343912587e-06,
+      "loss": 0.9239,
+      "step": 5076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.36915265553065e-06,
+      "loss": 0.8383,
+      "step": 5077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3672919806212367e-06,
+      "loss": 0.9447,
+      "step": 5078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3654318100194384e-06,
+      "loss": 0.8142,
+      "step": 5079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3635721440815823e-06,
+      "loss": 0.5982,
+      "step": 5080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3617129831638924e-06,
+      "loss": 0.7563,
+      "step": 5081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3598543276225032e-06,
+      "loss": 0.9937,
+      "step": 5082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3579961778134457e-06,
+      "loss": 0.5149,
+      "step": 5083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3561385340926607e-06,
+      "loss": 1.0725,
+      "step": 5084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.354281396815985e-06,
+      "loss": 0.7688,
+      "step": 5085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.352424766339165e-06,
+      "loss": 1.0931,
+      "step": 5086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3505686430178447e-06,
+      "loss": 0.7611,
+      "step": 5087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.348713027207576e-06,
+      "loss": 0.5273,
+      "step": 5088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3468579192638074e-06,
+      "loss": 1.0146,
+      "step": 5089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.345003319541897e-06,
+      "loss": 0.6819,
+      "step": 5090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3431492283970987e-06,
+      "loss": 0.809,
+      "step": 5091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3412956461845744e-06,
+      "loss": 0.9233,
+      "step": 5092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.339442573259388e-06,
+      "loss": 0.7212,
+      "step": 5093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3375900099765e-06,
+      "loss": 1.0073,
+      "step": 5094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.335737956690783e-06,
+      "loss": 0.8282,
+      "step": 5095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3338864137570018e-06,
+      "loss": 0.7253,
+      "step": 5096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.332035381529831e-06,
+      "loss": 0.8175,
+      "step": 5097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.330184860363843e-06,
+      "loss": 1.003,
+      "step": 5098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3283348506135155e-06,
+      "loss": 0.3594,
+      "step": 5099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.326485352633224e-06,
+      "loss": 1.0789,
+      "step": 5100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.324636366777252e-06,
+      "loss": 0.6253,
+      "step": 5101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3227878933997776e-06,
+      "loss": 0.66,
+      "step": 5102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.320939932854888e-06,
+      "loss": 0.681,
+      "step": 5103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3190924854965664e-06,
+      "loss": 0.8589,
+      "step": 5104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.317245551678702e-06,
+      "loss": 0.8442,
+      "step": 5105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.315399131755081e-06,
+      "loss": 0.9343,
+      "step": 5106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3135532260793968e-06,
+      "loss": 0.7993,
+      "step": 5107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3117078350052384e-06,
+      "loss": 0.6622,
+      "step": 5108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3098629588861025e-06,
+      "loss": 0.7254,
+      "step": 5109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3080185980753805e-06,
+      "loss": 0.9034,
+      "step": 5110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.306174752926372e-06,
+      "loss": 0.9955,
+      "step": 5111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.304331423792269e-06,
+      "loss": 0.5404,
+      "step": 5112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3024886110261757e-06,
+      "loss": 1.155,
+      "step": 5113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.300646314981087e-06,
+      "loss": 0.9396,
+      "step": 5114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2988045360099063e-06,
+      "loss": 0.6637,
+      "step": 5115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.296963274465433e-06,
+      "loss": 0.705,
+      "step": 5116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.295122530700372e-06,
+      "loss": 0.6433,
+      "step": 5117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2932823050673224e-06,
+      "loss": 0.5675,
+      "step": 5118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.291442597918791e-06,
+      "loss": 0.7379,
+      "step": 5119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.289603409607184e-06,
+      "loss": 1.0892,
+      "step": 5120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2877647404848018e-06,
+      "loss": 0.7709,
+      "step": 5121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2859265909038554e-06,
+      "loss": 0.4089,
+      "step": 5122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2840889612164457e-06,
+      "loss": 0.8859,
+      "step": 5123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2822518517745846e-06,
+      "loss": 0.9669,
+      "step": 5124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.280415262930174e-06,
+      "loss": 1.1625,
+      "step": 5125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2785791950350258e-06,
+      "loss": 0.7285,
+      "step": 5126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2767436484408426e-06,
+      "loss": 1.0318,
+      "step": 5127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2749086234992375e-06,
+      "loss": 0.8738,
+      "step": 5128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.273074120561713e-06,
+      "loss": 0.839,
+      "step": 5129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.271240139979681e-06,
+      "loss": 0.5963,
+      "step": 5130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2694066821044452e-06,
+      "loss": 0.4664,
+      "step": 5131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2675737472872173e-06,
+      "loss": 0.6139,
+      "step": 5132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2657413358791e-06,
+      "loss": 0.6643,
+      "step": 5133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.263909448231105e-06,
+      "loss": 0.9242,
+      "step": 5134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.262078084694134e-06,
+      "loss": 0.8587,
+      "step": 5135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2602472456189987e-06,
+      "loss": 0.3264,
+      "step": 5136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2584169313564e-06,
+      "loss": 1.0488,
+      "step": 5137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2565871422569473e-06,
+      "loss": 0.8824,
+      "step": 5138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.254757878671142e-06,
+      "loss": 1.052,
+      "step": 5139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2529291409493913e-06,
+      "loss": 1.044,
+      "step": 5140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.251100929441995e-06,
+      "loss": 0.768,
+      "step": 5141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.249273244499159e-06,
+      "loss": 0.4178,
+      "step": 5142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2474460864709825e-06,
+      "loss": 1.117,
+      "step": 5143
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2456194557074683e-06,
+      "loss": 1.0529,
+      "step": 5144
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2437933525585132e-06,
+      "loss": 0.6403,
+      "step": 5145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2419677773739195e-06,
+      "loss": 0.6464,
+      "step": 5146
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.240142730503381e-06,
+      "loss": 0.7197,
+      "step": 5147
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2383182122964956e-06,
+      "loss": 0.9829,
+      "step": 5148
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2364942231027608e-06,
+      "loss": 0.6095,
+      "step": 5149
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2346707632715663e-06,
+      "loss": 0.9501,
+      "step": 5150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.232847833152208e-06,
+      "loss": 0.6538,
+      "step": 5151
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.231025433093873e-06,
+      "loss": 0.5297,
+      "step": 5152
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.229203563445655e-06,
+      "loss": 0.896,
+      "step": 5153
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.227382224556537e-06,
+      "loss": 0.7486,
+      "step": 5154
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2255614167754093e-06,
+      "loss": 0.8525,
+      "step": 5155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2237411404510524e-06,
+      "loss": 1.3943,
+      "step": 5156
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.221921395932153e-06,
+      "loss": 0.9565,
+      "step": 5157
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.220102183567287e-06,
+      "loss": 0.7674,
+      "step": 5158
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2182835037049372e-06,
+      "loss": 0.8368,
+      "step": 5159
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.216465356693477e-06,
+      "loss": 0.7112,
+      "step": 5160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2146477428811846e-06,
+      "loss": 0.6875,
+      "step": 5161
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.212830662616228e-06,
+      "loss": 0.2814,
+      "step": 5162
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2110141162466823e-06,
+      "loss": 0.794,
+      "step": 5163
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2091981041205115e-06,
+      "loss": 0.7911,
+      "step": 5164
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2073826265855845e-06,
+      "loss": 0.8419,
+      "step": 5165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2055676839896613e-06,
+      "loss": 0.5404,
+      "step": 5166
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2037532766804058e-06,
+      "loss": 0.6408,
+      "step": 5167
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2019394050053737e-06,
+      "loss": 0.8373,
+      "step": 5168
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2001260693120236e-06,
+      "loss": 0.6944,
+      "step": 5169
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1983132699477044e-06,
+      "loss": 0.7768,
+      "step": 5170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.196501007259671e-06,
+      "loss": 0.9457,
+      "step": 5171
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1946892815950664e-06,
+      "loss": 0.7773,
+      "step": 5172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.192878093300939e-06,
+      "loss": 0.7151,
+      "step": 5173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1910674427242275e-06,
+      "loss": 0.724,
+      "step": 5174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.189257330211773e-06,
+      "loss": 1.1869,
+      "step": 5175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1874477561103087e-06,
+      "loss": 0.9264,
+      "step": 5176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1856387207664693e-06,
+      "loss": 0.5821,
+      "step": 5177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1838302245267818e-06,
+      "loss": 0.2794,
+      "step": 5178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1820222677376747e-06,
+      "loss": 0.6516,
+      "step": 5179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1802148507454675e-06,
+      "loss": 1.0478,
+      "step": 5180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1784079738963826e-06,
+      "loss": 0.8499,
+      "step": 5181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.176601637536533e-06,
+      "loss": 0.5469,
+      "step": 5182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.174795842011934e-06,
+      "loss": 0.6674,
+      "step": 5183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1729905876684907e-06,
+      "loss": 0.6895,
+      "step": 5184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.171185874852012e-06,
+      "loss": 0.5177,
+      "step": 5185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.169381703908195e-06,
+      "loss": 0.9872,
+      "step": 5186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.167578075182641e-06,
+      "loss": 0.7674,
+      "step": 5187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1657749890208394e-06,
+      "loss": 0.9859,
+      "step": 5188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.163972445768184e-06,
+      "loss": 0.6392,
+      "step": 5189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1621704457699577e-06,
+      "loss": 0.3289,
+      "step": 5190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.160368989371344e-06,
+      "loss": 0.7556,
+      "step": 5191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.158568076917418e-06,
+      "loss": 0.8402,
+      "step": 5192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1567677087531563e-06,
+      "loss": 0.9082,
+      "step": 5193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1549678852234236e-06,
+      "loss": 0.8297,
+      "step": 5194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.153168606672989e-06,
+      "loss": 0.9234,
+      "step": 5195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.151369873446509e-06,
+      "loss": 0.1979,
+      "step": 5196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1495716858885425e-06,
+      "loss": 0.8787,
+      "step": 5197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.147774044343538e-06,
+      "loss": 0.5247,
+      "step": 5198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1459769491558447e-06,
+      "loss": 0.9261,
+      "step": 5199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.144180400669702e-06,
+      "loss": 0.5833,
+      "step": 5200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.142384399229251e-06,
+      "loss": 0.5947,
+      "step": 5201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.14058894517852e-06,
+      "loss": 1.0706,
+      "step": 5202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1387940388614405e-06,
+      "loss": 0.6303,
+      "step": 5203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1369996806218308e-06,
+      "loss": 0.8154,
+      "step": 5204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.135205870803414e-06,
+      "loss": 0.2924,
+      "step": 5205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1334126097497975e-06,
+      "loss": 0.6742,
+      "step": 5206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.131619897804494e-06,
+      "loss": 1.029,
+      "step": 5207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1298277353109016e-06,
+      "loss": 0.7796,
+      "step": 5208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1280361226123207e-06,
+      "loss": 0.8627,
+      "step": 5209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1262450600519397e-06,
+      "loss": 1.0096,
+      "step": 5210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1244545479728497e-06,
+      "loss": 0.6283,
+      "step": 5211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.122664586718026e-06,
+      "loss": 0.8215,
+      "step": 5212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1208751766303505e-06,
+      "loss": 1.0772,
+      "step": 5213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.119086318052587e-06,
+      "loss": 0.8441,
+      "step": 5214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.117298011327405e-06,
+      "loss": 0.7955,
+      "step": 5215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1155102567973585e-06,
+      "loss": 0.7342,
+      "step": 5216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1137230548049042e-06,
+      "loss": 0.9092,
+      "step": 5217
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1119364056923858e-06,
+      "loss": 0.8418,
+      "step": 5218
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1101503098020477e-06,
+      "loss": 0.998,
+      "step": 5219
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.108364767476021e-06,
+      "loss": 0.6827,
+      "step": 5220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.106579779056339e-06,
+      "loss": 0.3962,
+      "step": 5221
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1047953448849205e-06,
+      "loss": 0.8757,
+      "step": 5222
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1030114653035868e-06,
+      "loss": 0.865,
+      "step": 5223
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1012281406540436e-06,
+      "loss": 0.6939,
+      "step": 5224
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0994453712778993e-06,
+      "loss": 0.9656,
+      "step": 5225
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0976631575166477e-06,
+      "loss": 0.6674,
+      "step": 5226
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.095881499711685e-06,
+      "loss": 0.7623,
+      "step": 5227
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0941003982042916e-06,
+      "loss": 0.8425,
+      "step": 5228
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0923198533356497e-06,
+      "loss": 0.6012,
+      "step": 5229
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0905398654468275e-06,
+      "loss": 0.7414,
+      "step": 5230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0887604348787917e-06,
+      "loss": 0.6765,
+      "step": 5231
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.086981561972402e-06,
+      "loss": 0.5817,
+      "step": 5232
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0852032470684076e-06,
+      "loss": 1.2972,
+      "step": 5233
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.083425490507455e-06,
+      "loss": 1.0471,
+      "step": 5234
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0816482926300802e-06,
+      "loss": 0.765,
+      "step": 5235
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0798716537767154e-06,
+      "loss": 0.6558,
+      "step": 5236
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.078095574287682e-06,
+      "loss": 0.7998,
+      "step": 5237
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.076320054503199e-06,
+      "loss": 0.9822,
+      "step": 5238
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0745450947633733e-06,
+      "loss": 0.8974,
+      "step": 5239
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.072770695408209e-06,
+      "loss": 0.8651,
+      "step": 5240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.070996856777597e-06,
+      "loss": 0.9428,
+      "step": 5241
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0692235792113298e-06,
+      "loss": 0.578,
+      "step": 5242
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.067450863049081e-06,
+      "loss": 0.9726,
+      "step": 5243
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0656787086304287e-06,
+      "loss": 0.7144,
+      "step": 5244
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.063907116294832e-06,
+      "loss": 0.5896,
+      "step": 5245
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0621360863816527e-06,
+      "loss": 0.588,
+      "step": 5246
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.060365619230135e-06,
+      "loss": 0.5851,
+      "step": 5247
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0585957151794256e-06,
+      "loss": 0.8953,
+      "step": 5248
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0568263745685523e-06,
+      "loss": 0.5815,
+      "step": 5249
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0550575977364454e-06,
+      "loss": 0.7448,
+      "step": 5250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053289385021919e-06,
+      "loss": 0.789,
+      "step": 5251
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.051521736763686e-06,
+      "loss": 0.9163,
+      "step": 5252
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.049754653300343e-06,
+      "loss": 0.664,
+      "step": 5253
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0479881349703885e-06,
+      "loss": 0.4606,
+      "step": 5254
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0462221821122024e-06,
+      "loss": 0.7906,
+      "step": 5255
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.044456795064066e-06,
+      "loss": 1.0158,
+      "step": 5256
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0426919741641434e-06,
+      "loss": 0.6458,
+      "step": 5257
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.040927719750496e-06,
+      "loss": 0.5459,
+      "step": 5258
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0391640321610767e-06,
+      "loss": 1.1036,
+      "step": 5259
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.037400911733725e-06,
+      "loss": 0.8258,
+      "step": 5260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0356383588061785e-06,
+      "loss": 0.5442,
+      "step": 5261
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0338763737160586e-06,
+      "loss": 0.6588,
+      "step": 5262
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0321149568008858e-06,
+      "loss": 0.8234,
+      "step": 5263
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0303541083980644e-06,
+      "loss": 0.5641,
+      "step": 5264
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.028593828844896e-06,
+      "loss": 0.6651,
+      "step": 5265
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0268341184785674e-06,
+      "loss": 0.699,
+      "step": 5266
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.025074977636162e-06,
+      "loss": 0.7718,
+      "step": 5267
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0233164066546495e-06,
+      "loss": 1.0673,
+      "step": 5268
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.021558405870895e-06,
+      "loss": 0.7286,
+      "step": 5269
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0198009756216497e-06,
+      "loss": 0.9446,
+      "step": 5270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0180441162435588e-06,
+      "loss": 0.8462,
+      "step": 5271
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0162878280731552e-06,
+      "loss": 0.7629,
+      "step": 5272
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0145321114468675e-06,
+      "loss": 0.8294,
+      "step": 5273
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.012776966701007e-06,
+      "loss": 0.9993,
+      "step": 5274
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0110223941717842e-06,
+      "loss": 0.5841,
+      "step": 5275
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0092683941952922e-06,
+      "loss": 1.0548,
+      "step": 5276
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0075149671075215e-06,
+      "loss": 0.6198,
+      "step": 5277
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.005762113244346e-06,
+      "loss": 0.6808,
+      "step": 5278
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0040098329415363e-06,
+      "loss": 0.8296,
+      "step": 5279
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.002258126534746e-06,
+      "loss": 1.0505,
+      "step": 5280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0005069943595273e-06,
+      "loss": 0.821,
+      "step": 5281
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.998756436751313e-06,
+      "loss": 1.1512,
+      "step": 5282
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9970064540454353e-06,
+      "loss": 0.9657,
+      "step": 5283
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9952570465771077e-06,
+      "loss": 0.5247,
+      "step": 5284
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9935082146814405e-06,
+      "loss": 0.7535,
+      "step": 5285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9917599586934277e-06,
+      "loss": 0.8173,
+      "step": 5286
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9900122789479576e-06,
+      "loss": 1.0053,
+      "step": 5287
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9882651757798083e-06,
+      "loss": 0.7246,
+      "step": 5288
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9865186495236423e-06,
+      "loss": 0.3737,
+      "step": 5289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.984772700514018e-06,
+      "loss": 0.7147,
+      "step": 5290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.983027329085377e-06,
+      "loss": 1.1212,
+      "step": 5291
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9812825355720562e-06,
+      "loss": 1.0312,
+      "step": 5292
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9795383203082763e-06,
+      "loss": 0.8398,
+      "step": 5293
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9777946836281535e-06,
+      "loss": 0.5875,
+      "step": 5294
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.976051625865686e-06,
+      "loss": 0.8375,
+      "step": 5295
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9743091473547676e-06,
+      "loss": 0.7228,
+      "step": 5296
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9725672484291752e-06,
+      "loss": 0.8253,
+      "step": 5297
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.970825929422582e-06,
+      "loss": 0.805,
+      "step": 5298
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9690851906685404e-06,
+      "loss": 0.7344,
+      "step": 5299
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.967345032500503e-06,
+      "loss": 0.9713,
+      "step": 5300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9656054552518e-06,
+      "loss": 0.988,
+      "step": 5301
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9638664592556595e-06,
+      "loss": 0.6987,
+      "step": 5302
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9621280448451914e-06,
+      "loss": 0.9382,
+      "step": 5303
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9603902123534003e-06,
+      "loss": 0.9428,
+      "step": 5304
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9586529621131727e-06,
+      "loss": 0.7107,
+      "step": 5305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9569162944572905e-06,
+      "loss": 0.8987,
+      "step": 5306
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9551802097184174e-06,
+      "loss": 0.6212,
+      "step": 5307
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.953444708229112e-06,
+      "loss": 0.8021,
+      "step": 5308
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.951709790321814e-06,
+      "loss": 0.4538,
+      "step": 5309
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9499754563288587e-06,
+      "loss": 1.0108,
+      "step": 5310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9482417065824628e-06,
+      "loss": 0.1904,
+      "step": 5311
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9465085414147373e-06,
+      "loss": 0.9443,
+      "step": 5312
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9447759611576743e-06,
+      "loss": 0.6944,
+      "step": 5313
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9430439661431603e-06,
+      "loss": 0.8644,
+      "step": 5314
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9413125567029685e-06,
+      "loss": 0.5523,
+      "step": 5315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.939581733168755e-06,
+      "loss": 0.4709,
+      "step": 5316
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.937851495872071e-06,
+      "loss": 0.9026,
+      "step": 5317
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.936121845144347e-06,
+      "loss": 0.5382,
+      "step": 5318
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9343927813169106e-06,
+      "loss": 0.9559,
+      "step": 5319
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9326643047209676e-06,
+      "loss": 0.8761,
+      "step": 5320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.93093641568762e-06,
+      "loss": 0.7835,
+      "step": 5321
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.92920911454785e-06,
+      "loss": 1.1985,
+      "step": 5322
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.927482401632532e-06,
+      "loss": 0.8745,
+      "step": 5323
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.925756277272425e-06,
+      "loss": 0.8712,
+      "step": 5324
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9240307417981775e-06,
+      "loss": 0.9196,
+      "step": 5325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.922305795540321e-06,
+      "loss": 0.6664,
+      "step": 5326
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9205814388292813e-06,
+      "loss": 0.5226,
+      "step": 5327
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9188576719953635e-06,
+      "loss": 0.9099,
+      "step": 5328
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.917134495368766e-06,
+      "loss": 0.9541,
+      "step": 5329
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.915411909279568e-06,
+      "loss": 1.0511,
+      "step": 5330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9136899140577435e-06,
+      "loss": 0.7037,
+      "step": 5331
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9119685100331444e-06,
+      "loss": 0.6922,
+      "step": 5332
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9102476975355167e-06,
+      "loss": 0.6952,
+      "step": 5333
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9085274768944877e-06,
+      "loss": 1.0128,
+      "step": 5334
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9068078484395768e-06,
+      "loss": 0.9706,
+      "step": 5335
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9050888125001838e-06,
+      "loss": 1.0875,
+      "step": 5336
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9033703694056016e-06,
+      "loss": 0.6123,
+      "step": 5337
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9016525194850017e-06,
+      "loss": 0.6232,
+      "step": 5338
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8999352630674506e-06,
+      "loss": 0.9547,
+      "step": 5339
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8982186004818932e-06,
+      "loss": 0.4453,
+      "step": 5340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8965025320571678e-06,
+      "loss": 0.776,
+      "step": 5341
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8947870581219923e-06,
+      "loss": 0.9526,
+      "step": 5342
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.893072179004975e-06,
+      "loss": 0.702,
+      "step": 5343
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8913578950346107e-06,
+      "loss": 0.879,
+      "step": 5344
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8896442065392755e-06,
+      "loss": 0.5834,
+      "step": 5345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8879311138472378e-06,
+      "loss": 0.5286,
+      "step": 5346
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8862186172866448e-06,
+      "loss": 0.8749,
+      "step": 5347
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8845067171855375e-06,
+      "loss": 0.6199,
+      "step": 5348
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.882795413871834e-06,
+      "loss": 0.8945,
+      "step": 5349
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8810847076733473e-06,
+      "loss": 0.954,
+      "step": 5350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8793745989177665e-06,
+      "loss": 0.8373,
+      "step": 5351
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8776650879326747e-06,
+      "loss": 0.7449,
+      "step": 5352
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.875956175045534e-06,
+      "loss": 0.6792,
+      "step": 5353
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8742478605836971e-06,
+      "loss": 0.5386,
+      "step": 5354
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8725401448743974e-06,
+      "loss": 0.8257,
+      "step": 5355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8708330282447583e-06,
+      "loss": 0.5797,
+      "step": 5356
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8691265110217833e-06,
+      "loss": 0.8681,
+      "step": 5357
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.867420593532367e-06,
+      "loss": 0.6326,
+      "step": 5358
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.865715276103282e-06,
+      "loss": 0.9316,
+      "step": 5359
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8640105590611945e-06,
+      "loss": 0.8811,
+      "step": 5360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8623064427326465e-06,
+      "loss": 0.8477,
+      "step": 5361
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8606029274440735e-06,
+      "loss": 0.9755,
+      "step": 5362
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8589000135217882e-06,
+      "loss": 0.7551,
+      "step": 5363
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8571977012919951e-06,
+      "loss": 1.0946,
+      "step": 5364
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8554959910807773e-06,
+      "loss": 0.7943,
+      "step": 5365
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8537948832141084e-06,
+      "loss": 0.93,
+      "step": 5366
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8520943780178402e-06,
+      "loss": 0.7036,
+      "step": 5367
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8503944758177155e-06,
+      "loss": 0.6396,
+      "step": 5368
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8486951769393562e-06,
+      "loss": 0.8679,
+      "step": 5369
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.846996481708272e-06,
+      "loss": 0.8647,
+      "step": 5370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8452983904498573e-06,
+      "loss": 0.6893,
+      "step": 5371
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.843600903489387e-06,
+      "loss": 0.6147,
+      "step": 5372
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.841904021152025e-06,
+      "loss": 0.8726,
+      "step": 5373
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8402077437628152e-06,
+      "loss": 0.9244,
+      "step": 5374
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8385120716466899e-06,
+      "loss": 0.7755,
+      "step": 5375
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8368170051284594e-06,
+      "loss": 0.6685,
+      "step": 5376
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8351225445328257e-06,
+      "loss": 0.5856,
+      "step": 5377
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8334286901843673e-06,
+      "loss": 1.0416,
+      "step": 5378
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8317354424075529e-06,
+      "loss": 1.0486,
+      "step": 5379
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8300428015267291e-06,
+      "loss": 0.582,
+      "step": 5380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8283507678661327e-06,
+      "loss": 0.8382,
+      "step": 5381
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8266593417498774e-06,
+      "loss": 0.9683,
+      "step": 5382
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8249685235019665e-06,
+      "loss": 0.6727,
+      "step": 5383
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8232783134462828e-06,
+      "loss": 0.7883,
+      "step": 5384
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8215887119065928e-06,
+      "loss": 0.9651,
+      "step": 5385
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8198997192065503e-06,
+      "loss": 0.7547,
+      "step": 5386
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8182113356696873e-06,
+      "loss": 0.792,
+      "step": 5387
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8165235616194243e-06,
+      "loss": 0.8577,
+      "step": 5388
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8148363973790595e-06,
+      "loss": 0.6118,
+      "step": 5389
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8131498432717798e-06,
+      "loss": 0.7708,
+      "step": 5390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8114638996206496e-06,
+      "loss": 0.8605,
+      "step": 5391
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.809778566748623e-06,
+      "loss": 0.8019,
+      "step": 5392
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8080938449785295e-06,
+      "loss": 0.9319,
+      "step": 5393
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8064097346330889e-06,
+      "loss": 1.0627,
+      "step": 5394
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8047262360348977e-06,
+      "loss": 0.6088,
+      "step": 5395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8030433495064414e-06,
+      "loss": 1.0919,
+      "step": 5396
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8013610753700805e-06,
+      "loss": 0.8178,
+      "step": 5397
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7996794139480667e-06,
+      "loss": 0.9034,
+      "step": 5398
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7979983655625265e-06,
+      "loss": 0.5231,
+      "step": 5399
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7963179305354767e-06,
+      "loss": 0.4308,
+      "step": 5400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7946381091888077e-06,
+      "loss": 0.8742,
+      "step": 5401
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7929589018443016e-06,
+      "loss": 0.5845,
+      "step": 5402
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7912803088236153e-06,
+      "loss": 0.3928,
+      "step": 5403
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.789602330448294e-06,
+      "loss": 0.7313,
+      "step": 5404
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7879249670397592e-06,
+      "loss": 0.4047,
+      "step": 5405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7862482189193208e-06,
+      "loss": 0.6524,
+      "step": 5406
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7845720864081644e-06,
+      "loss": 0.8384,
+      "step": 5407
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7828965698273653e-06,
+      "loss": 0.9393,
+      "step": 5408
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.781221669497873e-06,
+      "loss": 0.7556,
+      "step": 5409
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7795473857405248e-06,
+      "loss": 0.9361,
+      "step": 5410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.777873718876036e-06,
+      "loss": 0.798,
+      "step": 5411
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7762006692250067e-06,
+      "loss": 0.7984,
+      "step": 5412
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.774528237107916e-06,
+      "loss": 1.1272,
+      "step": 5413
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7728564228451283e-06,
+      "loss": 0.6084,
+      "step": 5414
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7711852267568846e-06,
+      "loss": 0.7872,
+      "step": 5415
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7695146491633135e-06,
+      "loss": 1.1922,
+      "step": 5416
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7678446903844186e-06,
+      "loss": 0.8205,
+      "step": 5417
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7661753507400926e-06,
+      "loss": 0.6108,
+      "step": 5418
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7645066305501008e-06,
+      "loss": 0.7959,
+      "step": 5419
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7628385301340983e-06,
+      "loss": 1.0408,
+      "step": 5420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7611710498116148e-06,
+      "loss": 0.9527,
+      "step": 5421
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.759504189902066e-06,
+      "loss": 0.5765,
+      "step": 5422
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7578379507247444e-06,
+      "loss": 1.1774,
+      "step": 5423
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7561723325988294e-06,
+      "loss": 0.6677,
+      "step": 5424
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7545073358433745e-06,
+      "loss": 0.8433,
+      "step": 5425
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7528429607773185e-06,
+      "loss": 0.4676,
+      "step": 5426
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7511792077194833e-06,
+      "loss": 0.3978,
+      "step": 5427
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7495160769885643e-06,
+      "loss": 0.8856,
+      "step": 5428
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7478535689031456e-06,
+      "loss": 0.7698,
+      "step": 5429
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7461916837816855e-06,
+      "loss": 0.9094,
+      "step": 5430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7445304219425286e-06,
+      "loss": 0.9863,
+      "step": 5431
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.742869783703895e-06,
+      "loss": 0.8193,
+      "step": 5432
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7412097693838896e-06,
+      "loss": 0.9042,
+      "step": 5433
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7395503793004948e-06,
+      "loss": 0.9444,
+      "step": 5434
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7378916137715757e-06,
+      "loss": 0.7602,
+      "step": 5435
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.736233473114875e-06,
+      "loss": 0.518,
+      "step": 5436
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7345759576480192e-06,
+      "loss": 0.6763,
+      "step": 5437
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7329190676885116e-06,
+      "loss": 0.9631,
+      "step": 5438
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7312628035537388e-06,
+      "loss": 0.8864,
+      "step": 5439
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7296071655609643e-06,
+      "loss": 1.016,
+      "step": 5440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7279521540273353e-06,
+      "loss": 0.5381,
+      "step": 5441
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7262977692698746e-06,
+      "loss": 0.856,
+      "step": 5442
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7246440116054901e-06,
+      "loss": 0.4073,
+      "step": 5443
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7229908813509644e-06,
+      "loss": 0.6111,
+      "step": 5444
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7213383788229647e-06,
+      "loss": 1.0611,
+      "step": 5445
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7196865043380335e-06,
+      "loss": 1.11,
+      "step": 5446
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.718035258212597e-06,
+      "loss": 0.9977,
+      "step": 5447
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716384640762957e-06,
+      "loss": 0.9739,
+      "step": 5448
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7147346523052999e-06,
+      "loss": 0.4601,
+      "step": 5449
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7130852931556858e-06,
+      "loss": 0.9208,
+      "step": 5450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7114365636300595e-06,
+      "loss": 1.1486,
+      "step": 5451
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7097884640442407e-06,
+      "loss": 0.7252,
+      "step": 5452
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7081409947139321e-06,
+      "loss": 0.7413,
+      "step": 5453
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7064941559547154e-06,
+      "loss": 0.9906,
+      "step": 5454
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7048479480820479e-06,
+      "loss": 0.5796,
+      "step": 5455
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7032023714112706e-06,
+      "loss": 0.9337,
+      "step": 5456
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7015574262575985e-06,
+      "loss": 0.8935,
+      "step": 5457
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.699913112936133e-06,
+      "loss": 0.7533,
+      "step": 5458
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6982694317618448e-06,
+      "loss": 1.0343,
+      "step": 5459
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6966263830495939e-06,
+      "loss": 0.8338,
+      "step": 5460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6949839671141094e-06,
+      "loss": 1.0267,
+      "step": 5461
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6933421842700081e-06,
+      "loss": 0.5429,
+      "step": 5462
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6917010348317774e-06,
+      "loss": 0.8494,
+      "step": 5463
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6900605191137904e-06,
+      "loss": 0.8045,
+      "step": 5464
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6884206374302925e-06,
+      "loss": 0.6404,
+      "step": 5465
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6867813900954144e-06,
+      "loss": 0.8,
+      "step": 5466
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6851427774231578e-06,
+      "loss": 0.3383,
+      "step": 5467
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.68350479972741e-06,
+      "loss": 0.8072,
+      "step": 5468
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6818674573219307e-06,
+      "loss": 1.0373,
+      "step": 5469
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.680230750520363e-06,
+      "loss": 1.0713,
+      "step": 5470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6785946796362223e-06,
+      "loss": 0.6608,
+      "step": 5471
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6769592449829103e-06,
+      "loss": 0.8062,
+      "step": 5472
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6753244468736978e-06,
+      "loss": 0.5202,
+      "step": 5473
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6736902856217419e-06,
+      "loss": 0.71,
+      "step": 5474
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6720567615400708e-06,
+      "loss": 0.6994,
+      "step": 5475
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6704238749415958e-06,
+      "loss": 0.4217,
+      "step": 5476
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6687916261391018e-06,
+      "loss": 0.9482,
+      "step": 5477
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6671600154452573e-06,
+      "loss": 0.5031,
+      "step": 5478
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6655290431726007e-06,
+      "loss": 1.016,
+      "step": 5479
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6638987096335562e-06,
+      "loss": 0.7269,
+      "step": 5480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6622690151404187e-06,
+      "loss": 0.9036,
+      "step": 5481
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6606399600053653e-06,
+      "loss": 1.0017,
+      "step": 5482
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.659011544540451e-06,
+      "loss": 0.6616,
+      "step": 5483
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6573837690576028e-06,
+      "loss": 0.6657,
+      "step": 5484
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6557566338686325e-06,
+      "loss": 0.9888,
+      "step": 5485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6541301392852222e-06,
+      "loss": 0.7251,
+      "step": 5486
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6525042856189378e-06,
+      "loss": 0.7597,
+      "step": 5487
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6508790731812152e-06,
+      "loss": 0.7509,
+      "step": 5488
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.649254502283376e-06,
+      "loss": 0.651,
+      "step": 5489
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6476305732366106e-06,
+      "loss": 0.7126,
+      "step": 5490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6460072863519928e-06,
+      "loss": 0.6734,
+      "step": 5491
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6443846419404685e-06,
+      "loss": 0.5976,
+      "step": 5492
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6427626403128666e-06,
+      "loss": 0.815,
+      "step": 5493
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6411412817798844e-06,
+      "loss": 0.9584,
+      "step": 5494
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6395205666521052e-06,
+      "loss": 1.1317,
+      "step": 5495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6379004952399808e-06,
+      "loss": 0.712,
+      "step": 5496
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.636281067853847e-06,
+      "loss": 0.6924,
+      "step": 5497
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.634662284803909e-06,
+      "loss": 1.1509,
+      "step": 5498
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6330441464002567e-06,
+      "loss": 0.9132,
+      "step": 5499
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6314266529528478e-06,
+      "loss": 0.6153,
+      "step": 5500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6298098047715244e-06,
+      "loss": 0.922,
+      "step": 5501
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6281936021659977e-06,
+      "loss": 0.7009,
+      "step": 5502
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.626578045445863e-06,
+      "loss": 0.9104,
+      "step": 5503
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6249631349205842e-06,
+      "loss": 1.092,
+      "step": 5504
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6233488708995081e-06,
+      "loss": 0.7519,
+      "step": 5505
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6217352536918512e-06,
+      "loss": 0.5223,
+      "step": 5506
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6201222836067132e-06,
+      "loss": 0.7878,
+      "step": 5507
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6185099609530624e-06,
+      "loss": 0.8154,
+      "step": 5508
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6168982860397492e-06,
+      "loss": 1.0524,
+      "step": 5509
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.615287259175498e-06,
+      "loss": 0.9342,
+      "step": 5510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6136768806689056e-06,
+      "loss": 0.9504,
+      "step": 5511
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6120671508284518e-06,
+      "loss": 0.7831,
+      "step": 5512
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6104580699624839e-06,
+      "loss": 0.8119,
+      "step": 5513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6088496383792317e-06,
+      "loss": 0.4246,
+      "step": 5514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.607241856386796e-06,
+      "loss": 1.107,
+      "step": 5515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.605634724293157e-06,
+      "loss": 0.6816,
+      "step": 5516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6040282424061654e-06,
+      "loss": 0.3004,
+      "step": 5517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6024224110335546e-06,
+      "loss": 0.5099,
+      "step": 5518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6008172304829244e-06,
+      "loss": 0.6249,
+      "step": 5519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5992127010617585e-06,
+      "loss": 0.8515,
+      "step": 5520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5976088230774096e-06,
+      "loss": 0.8895,
+      "step": 5521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5960055968371102e-06,
+      "loss": 0.7287,
+      "step": 5522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5944030226479634e-06,
+      "loss": 0.6455,
+      "step": 5523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.592801100816953e-06,
+      "loss": 1.0237,
+      "step": 5524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5911998316509309e-06,
+      "loss": 0.9162,
+      "step": 5525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5895992154566314e-06,
+      "loss": 0.8916,
+      "step": 5526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5879992525406562e-06,
+      "loss": 0.6524,
+      "step": 5527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5863999432094902e-06,
+      "loss": 0.8908,
+      "step": 5528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.584801287769484e-06,
+      "loss": 0.8736,
+      "step": 5529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.583203286526872e-06,
+      "loss": 0.5555,
+      "step": 5530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5816059397877553e-06,
+      "loss": 0.8671,
+      "step": 5531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5800092478581158e-06,
+      "loss": 0.7345,
+      "step": 5532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5784132110438045e-06,
+      "loss": 1.1256,
+      "step": 5533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5768178296505531e-06,
+      "loss": 0.6332,
+      "step": 5534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5752231039839616e-06,
+      "loss": 0.8953,
+      "step": 5535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5736290343495082e-06,
+      "loss": 0.6073,
+      "step": 5536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5720356210525462e-06,
+      "loss": 0.926,
+      "step": 5537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5704428643982976e-06,
+      "loss": 1.0521,
+      "step": 5538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5688507646918671e-06,
+      "loss": 0.7106,
+      "step": 5539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5672593222382243e-06,
+      "loss": 0.8247,
+      "step": 5540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5656685373422214e-06,
+      "loss": 0.7511,
+      "step": 5541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5640784103085777e-06,
+      "loss": 0.8272,
+      "step": 5542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5624889414418925e-06,
+      "loss": 0.8065,
+      "step": 5543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.560900131046632e-06,
+      "loss": 0.5192,
+      "step": 5544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.559311979427145e-06,
+      "loss": 0.8447,
+      "step": 5545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.557724486887645e-06,
+      "loss": 0.9412,
+      "step": 5546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5561376537322275e-06,
+      "loss": 0.8035,
+      "step": 5547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5545514802648548e-06,
+      "loss": 0.7641,
+      "step": 5548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5529659667893686e-06,
+      "loss": 0.8261,
+      "step": 5549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5513811136094786e-06,
+      "loss": 0.5588,
+      "step": 5550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5497969210287738e-06,
+      "loss": 0.9916,
+      "step": 5551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5482133893507113e-06,
+      "loss": 0.9027,
+      "step": 5552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5466305188786257e-06,
+      "loss": 0.7051,
+      "step": 5553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5450483099157214e-06,
+      "loss": 0.9976,
+      "step": 5554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5434667627650813e-06,
+      "loss": 0.55,
+      "step": 5555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5418858777296542e-06,
+      "loss": 1.2141,
+      "step": 5556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5403056551122697e-06,
+      "loss": 0.626,
+      "step": 5557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5387260952156236e-06,
+      "loss": 0.8915,
+      "step": 5558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5371471983422909e-06,
+      "loss": 0.8239,
+      "step": 5559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.535568964794714e-06,
+      "loss": 0.7728,
+      "step": 5560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5339913948752143e-06,
+      "loss": 0.5571,
+      "step": 5561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5324144888859793e-06,
+      "loss": 0.7575,
+      "step": 5562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5308382471290755e-06,
+      "loss": 1.1635,
+      "step": 5563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5292626699064372e-06,
+      "loss": 0.6763,
+      "step": 5564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.527687757519874e-06,
+      "loss": 0.6934,
+      "step": 5565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5261135102710706e-06,
+      "loss": 0.6228,
+      "step": 5566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5245399284615774e-06,
+      "loss": 0.7693,
+      "step": 5567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5229670123928253e-06,
+      "loss": 0.7618,
+      "step": 5568
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.52139476236611e-06,
+      "loss": 0.6867,
+      "step": 5569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5198231786826074e-06,
+      "loss": 0.4843,
+      "step": 5570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5182522616433576e-06,
+      "loss": 0.7813,
+      "step": 5571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5166820115492802e-06,
+      "loss": 0.8063,
+      "step": 5572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5151124287011621e-06,
+      "loss": 0.8197,
+      "step": 5573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5135435133996667e-06,
+      "loss": 0.5591,
+      "step": 5574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5119752659453235e-06,
+      "loss": 0.94,
+      "step": 5575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5104076866385414e-06,
+      "loss": 0.6937,
+      "step": 5576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5088407757795948e-06,
+      "loss": 0.6796,
+      "step": 5577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.507274533668635e-06,
+      "loss": 0.7255,
+      "step": 5578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5057089606056808e-06,
+      "loss": 0.6399,
+      "step": 5579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5041440568906275e-06,
+      "loss": 0.258,
+      "step": 5580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502579822823237e-06,
+      "loss": 1.3004,
+      "step": 5581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501016258703149e-06,
+      "loss": 0.6954,
+      "step": 5582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4994533648298687e-06,
+      "loss": 0.7975,
+      "step": 5583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497891141502778e-06,
+      "loss": 1.2094,
+      "step": 5584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4963295890211255e-06,
+      "loss": 0.9996,
+      "step": 5585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4947687076840373e-06,
+      "loss": 0.6448,
+      "step": 5586
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4932084977905043e-06,
+      "loss": 0.7483,
+      "step": 5587
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4916489596393945e-06,
+      "loss": 0.4671,
+      "step": 5588
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4900900935294427e-06,
+      "loss": 0.5818,
+      "step": 5589
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4885318997592596e-06,
+      "loss": 1.2542,
+      "step": 5590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4869743786273216e-06,
+      "loss": 0.6728,
+      "step": 5591
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4854175304319818e-06,
+      "loss": 0.9973,
+      "step": 5592
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4838613554714598e-06,
+      "loss": 0.9898,
+      "step": 5593
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.48230585404385e-06,
+      "loss": 0.7474,
+      "step": 5594
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4807510264471143e-06,
+      "loss": 0.6367,
+      "step": 5595
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.479196872979089e-06,
+      "loss": 0.7622,
+      "step": 5596
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4776433939374774e-06,
+      "loss": 0.8563,
+      "step": 5597
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4760905896198585e-06,
+      "loss": 0.5037,
+      "step": 5598
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4745384603236763e-06,
+      "loss": 0.6527,
+      "step": 5599
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.472987006346251e-06,
+      "loss": 0.4379,
+      "step": 5600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4714362279847688e-06,
+      "loss": 0.7147,
+      "step": 5601
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4698861255362907e-06,
+      "loss": 0.8933,
+      "step": 5602
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4683366992977433e-06,
+      "loss": 0.5508,
+      "step": 5603
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4667879495659303e-06,
+      "loss": 0.8363,
+      "step": 5604
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.465239876637518e-06,
+      "loss": 0.8333,
+      "step": 5605
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4636924808090508e-06,
+      "loss": 0.6857,
+      "step": 5606
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.462145762376936e-06,
+      "loss": 0.8876,
+      "step": 5607
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4605997216374583e-06,
+      "loss": 0.7664,
+      "step": 5608
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4590543588867662e-06,
+      "loss": 0.5861,
+      "step": 5609
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4575096744208832e-06,
+      "loss": 0.9197,
+      "step": 5610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4559656685356993e-06,
+      "loss": 0.7681,
+      "step": 5611
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4544223415269782e-06,
+      "loss": 0.4513,
+      "step": 5612
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4528796936903484e-06,
+      "loss": 0.5607,
+      "step": 5613
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4513377253213145e-06,
+      "loss": 0.7961,
+      "step": 5614
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.449796436715245e-06,
+      "loss": 0.872,
+      "step": 5615
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4482558281673841e-06,
+      "loss": 0.9648,
+      "step": 5616
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4467158999728398e-06,
+      "loss": 0.6626,
+      "step": 5617
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4451766524265948e-06,
+      "loss": 0.4439,
+      "step": 5618
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.443638085823496e-06,
+      "loss": 0.6379,
+      "step": 5619
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.442100200458268e-06,
+      "loss": 1.0825,
+      "step": 5620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4405629966254947e-06,
+      "loss": 0.6534,
+      "step": 5621
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4390264746196391e-06,
+      "loss": 0.4215,
+      "step": 5622
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4374906347350255e-06,
+      "loss": 0.7478,
+      "step": 5623
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4359554772658551e-06,
+      "loss": 0.5417,
+      "step": 5624
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4344210025061906e-06,
+      "loss": 0.7175,
+      "step": 5625
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4328872107499713e-06,
+      "loss": 1.0815,
+      "step": 5626
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4313541022909994e-06,
+      "loss": 1.0006,
+      "step": 5627
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4298216774229524e-06,
+      "loss": 0.4249,
+      "step": 5628
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4282899364393694e-06,
+      "loss": 0.5041,
+      "step": 5629
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4267588796336663e-06,
+      "loss": 0.9947,
+      "step": 5630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4252285072991207e-06,
+      "loss": 0.9365,
+      "step": 5631
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4236988197288865e-06,
+      "loss": 0.7395,
+      "step": 5632
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4221698172159786e-06,
+      "loss": 0.7226,
+      "step": 5633
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4206415000532885e-06,
+      "loss": 0.8513,
+      "step": 5634
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4191138685335694e-06,
+      "loss": 0.9919,
+      "step": 5635
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4175869229494487e-06,
+      "loss": 1.0219,
+      "step": 5636
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4160606635934177e-06,
+      "loss": 0.7096,
+      "step": 5637
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4145350907578414e-06,
+      "loss": 0.6958,
+      "step": 5638
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.413010204734947e-06,
+      "loss": 0.5426,
+      "step": 5639
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4114860058168368e-06,
+      "loss": 0.9687,
+      "step": 5640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.409962494295476e-06,
+      "loss": 0.965,
+      "step": 5641
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4084396704627024e-06,
+      "loss": 0.7311,
+      "step": 5642
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.406917534610217e-06,
+      "loss": 0.9572,
+      "step": 5643
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4053960870295962e-06,
+      "loss": 0.854,
+      "step": 5644
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4038753280122768e-06,
+      "loss": 0.7218,
+      "step": 5645
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4023552578495704e-06,
+      "loss": 0.7743,
+      "step": 5646
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4008358768326502e-06,
+      "loss": 0.8868,
+      "step": 5647
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3993171852525628e-06,
+      "loss": 0.8383,
+      "step": 5648
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3977991834002219e-06,
+      "loss": 0.7211,
+      "step": 5649
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.396281871566405e-06,
+      "loss": 0.8001,
+      "step": 5650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3947652500417641e-06,
+      "loss": 0.9978,
+      "step": 5651
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3932493191168106e-06,
+      "loss": 0.7894,
+      "step": 5652
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3917340790819329e-06,
+      "loss": 0.9343,
+      "step": 5653
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.390219530227378e-06,
+      "loss": 0.779,
+      "step": 5654
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3887056728432685e-06,
+      "loss": 0.7365,
+      "step": 5655
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.387192507219588e-06,
+      "loss": 0.7711,
+      "step": 5656
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3856800336461934e-06,
+      "loss": 1.0101,
+      "step": 5657
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3841682524128024e-06,
+      "loss": 0.8048,
+      "step": 5658
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3826571638090086e-06,
+      "loss": 0.6513,
+      "step": 5659
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.381146768124263e-06,
+      "loss": 0.726,
+      "step": 5660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3796370656478936e-06,
+      "loss": 0.9743,
+      "step": 5661
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3781280566690875e-06,
+      "loss": 0.8794,
+      "step": 5662
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.376619741476905e-06,
+      "loss": 0.8181,
+      "step": 5663
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3751121203602686e-06,
+      "loss": 0.6499,
+      "step": 5664
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.373605193607973e-06,
+      "loss": 0.7396,
+      "step": 5665
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3720989615086739e-06,
+      "loss": 0.8412,
+      "step": 5666
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3705934243509007e-06,
+      "loss": 0.5477,
+      "step": 5667
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3690885824230426e-06,
+      "loss": 0.9224,
+      "step": 5668
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3675844360133623e-06,
+      "loss": 0.7876,
+      "step": 5669
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3660809854099833e-06,
+      "loss": 0.4985,
+      "step": 5670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3645782309009015e-06,
+      "loss": 0.414,
+      "step": 5671
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3630761727739733e-06,
+      "loss": 1.0223,
+      "step": 5672
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.361574811316929e-06,
+      "loss": 0.8424,
+      "step": 5673
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3600741468173573e-06,
+      "loss": 0.6892,
+      "step": 5674
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.358574179562721e-06,
+      "loss": 0.7628,
+      "step": 5675
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3570749098403435e-06,
+      "loss": 0.8362,
+      "step": 5676
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.355576337937417e-06,
+      "loss": 0.6254,
+      "step": 5677
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3540784641410031e-06,
+      "loss": 0.9788,
+      "step": 5678
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.352581288738023e-06,
+      "loss": 0.7638,
+      "step": 5679
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3510848120152703e-06,
+      "loss": 0.9049,
+      "step": 5680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3495890342593992e-06,
+      "loss": 0.5652,
+      "step": 5681
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3480939557569362e-06,
+      "loss": 0.7223,
+      "step": 5682
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.346599576794268e-06,
+      "loss": 1.0416,
+      "step": 5683
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3451058976576526e-06,
+      "loss": 1.0574,
+      "step": 5684
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3436129186332082e-06,
+      "loss": 0.7552,
+      "step": 5685
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3421206400069249e-06,
+      "loss": 0.7773,
+      "step": 5686
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3406290620646522e-06,
+      "loss": 0.5949,
+      "step": 5687
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.339138185092113e-06,
+      "loss": 0.6947,
+      "step": 5688
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3376480093748884e-06,
+      "loss": 0.7079,
+      "step": 5689
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3361585351984313e-06,
+      "loss": 0.8434,
+      "step": 5690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3346697628480548e-06,
+      "loss": 0.8284,
+      "step": 5691
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.333181692608943e-06,
+      "loss": 0.8475,
+      "step": 5692
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3316943247661397e-06,
+      "loss": 0.6159,
+      "step": 5693
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3302076596045604e-06,
+      "loss": 0.7843,
+      "step": 5694
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.32872169740898e-06,
+      "loss": 0.7633,
+      "step": 5695
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.327236438464044e-06,
+      "loss": 0.9678,
+      "step": 5696
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.325751883054258e-06,
+      "loss": 0.5062,
+      "step": 5697
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3242680314639995e-06,
+      "loss": 0.8608,
+      "step": 5698
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3227848839775027e-06,
+      "loss": 0.8521,
+      "step": 5699
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.321302440878875e-06,
+      "loss": 0.7809,
+      "step": 5700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.319820702452083e-06,
+      "loss": 0.8065,
+      "step": 5701
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.318339668980963e-06,
+      "loss": 0.6723,
+      "step": 5702
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3168593407492114e-06,
+      "loss": 0.4445,
+      "step": 5703
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3153797180403927e-06,
+      "loss": 0.4163,
+      "step": 5704
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3139008011379383e-06,
+      "loss": 1.1485,
+      "step": 5705
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3124225903251375e-06,
+      "loss": 0.5005,
+      "step": 5706
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.310945085885153e-06,
+      "loss": 1.0882,
+      "step": 5707
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3094682881010028e-06,
+      "loss": 0.9952,
+      "step": 5708
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3079921972555793e-06,
+      "loss": 0.6426,
+      "step": 5709
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3065168136316308e-06,
+      "loss": 0.9327,
+      "step": 5710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3050421375117767e-06,
+      "loss": 0.8102,
+      "step": 5711
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3035681691784956e-06,
+      "loss": 0.5054,
+      "step": 5712
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3020949089141365e-06,
+      "loss": 1.0262,
+      "step": 5713
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3006223570009052e-06,
+      "loss": 0.6932,
+      "step": 5714
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2991505137208798e-06,
+      "loss": 0.6248,
+      "step": 5715
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2976793793559956e-06,
+      "loss": 0.8483,
+      "step": 5716
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2962089541880585e-06,
+      "loss": 0.9209,
+      "step": 5717
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2947392384987318e-06,
+      "loss": 0.6157,
+      "step": 5718
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.29327023256955e-06,
+      "loss": 0.2961,
+      "step": 5719
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.291801936681905e-06,
+      "loss": 0.7762,
+      "step": 5720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2903343511170584e-06,
+      "loss": 0.919,
+      "step": 5721
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2888674761561309e-06,
+      "loss": 1.0223,
+      "step": 5722
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2874013120801116e-06,
+      "loss": 0.9576,
+      "step": 5723
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2859358591698483e-06,
+      "loss": 0.5547,
+      "step": 5724
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2844711177060586e-06,
+      "loss": 0.497,
+      "step": 5725
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2830070879693174e-06,
+      "loss": 0.7475,
+      "step": 5726
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2815437702400696e-06,
+      "loss": 1.0495,
+      "step": 5727
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2800811647986177e-06,
+      "loss": 0.6044,
+      "step": 5728
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2786192719251332e-06,
+      "loss": 0.7703,
+      "step": 5729
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2771580918996463e-06,
+      "loss": 0.945,
+      "step": 5730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2756976250020537e-06,
+      "loss": 0.6961,
+      "step": 5731
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2742378715121168e-06,
+      "loss": 0.8619,
+      "step": 5732
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2727788317094552e-06,
+      "loss": 0.6276,
+      "step": 5733
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2713205058735573e-06,
+      "loss": 1.0091,
+      "step": 5734
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2698628942837698e-06,
+      "loss": 0.848,
+      "step": 5735
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2684059972193075e-06,
+      "loss": 0.7076,
+      "step": 5736
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2669498149592436e-06,
+      "loss": 1.1271,
+      "step": 5737
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2654943477825192e-06,
+      "loss": 0.82,
+      "step": 5738
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.264039595967933e-06,
+      "loss": 0.5208,
+      "step": 5739
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2625855597941527e-06,
+      "loss": 0.7489,
+      "step": 5740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2611322395397023e-06,
+      "loss": 0.8743,
+      "step": 5741
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259679635482976e-06,
+      "loss": 0.7797,
+      "step": 5742
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2582277479022231e-06,
+      "loss": 0.6778,
+      "step": 5743
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.256776577075563e-06,
+      "loss": 0.8381,
+      "step": 5744
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2553261232809715e-06,
+      "loss": 0.5123,
+      "step": 5745
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2538763867962921e-06,
+      "loss": 0.5619,
+      "step": 5746
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.252427367899226e-06,
+      "loss": 0.7081,
+      "step": 5747
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2509790668673427e-06,
+      "loss": 0.5823,
+      "step": 5748
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2495314839780681e-06,
+      "loss": 0.9329,
+      "step": 5749
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.248084619508696e-06,
+      "loss": 0.7184,
+      "step": 5750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.246638473736378e-06,
+      "loss": 0.6439,
+      "step": 5751
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.245193046938133e-06,
+      "loss": 0.6439,
+      "step": 5752
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2437483393908351e-06,
+      "loss": 0.635,
+      "step": 5753
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.242304351371229e-06,
+      "loss": 0.9841,
+      "step": 5754
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2408610831559149e-06,
+      "loss": 1.0113,
+      "step": 5755
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.239418535021359e-06,
+      "loss": 0.8489,
+      "step": 5756
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2379767072438863e-06,
+      "loss": 0.526,
+      "step": 5757
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.236535600099688e-06,
+      "loss": 0.7113,
+      "step": 5758
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2350952138648132e-06,
+      "loss": 0.8658,
+      "step": 5759
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2336555488151746e-06,
+      "loss": 0.7832,
+      "step": 5760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2322166052265494e-06,
+      "loss": 0.8439,
+      "step": 5761
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2307783833745706e-06,
+      "loss": 1.1504,
+      "step": 5762
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2293408835347397e-06,
+      "loss": 0.5377,
+      "step": 5763
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2279041059824133e-06,
+      "loss": 0.776,
+      "step": 5764
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2264680509928151e-06,
+      "loss": 1.02,
+      "step": 5765
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2250327188410267e-06,
+      "loss": 0.6901,
+      "step": 5766
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.223598109801995e-06,
+      "loss": 1.0954,
+      "step": 5767
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2221642241505232e-06,
+      "loss": 0.3232,
+      "step": 5768
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2207310621612817e-06,
+      "loss": 0.9958,
+      "step": 5769
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2192986241087967e-06,
+      "loss": 0.5054,
+      "step": 5770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.217866910267461e-06,
+      "loss": 0.3821,
+      "step": 5771
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2164359209115235e-06,
+      "loss": 0.692,
+      "step": 5772
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2150056563150996e-06,
+      "loss": 1.0141,
+      "step": 5773
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.213576116752161e-06,
+      "loss": 0.6763,
+      "step": 5774
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2121473024965452e-06,
+      "loss": 0.6391,
+      "step": 5775
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2107192138219447e-06,
+      "loss": 0.8266,
+      "step": 5776
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2092918510019208e-06,
+      "loss": 0.7409,
+      "step": 5777
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.207865214309888e-06,
+      "loss": 0.8508,
+      "step": 5778
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2064393040191276e-06,
+      "loss": 0.7667,
+      "step": 5779
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2050141204027776e-06,
+      "loss": 0.9375,
+      "step": 5780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2035896637338413e-06,
+      "loss": 0.61,
+      "step": 5781
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2021659342851765e-06,
+      "loss": 1.0166,
+      "step": 5782
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2007429323295088e-06,
+      "loss": 0.8594,
+      "step": 5783
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1993206581394179e-06,
+      "loss": 0.9336,
+      "step": 5784
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1978991119873495e-06,
+      "loss": 0.7952,
+      "step": 5785
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1964782941456055e-06,
+      "loss": 0.427,
+      "step": 5786
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1950582048863514e-06,
+      "loss": 0.7268,
+      "step": 5787
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1936388444816127e-06,
+      "loss": 0.7143,
+      "step": 5788
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1922202132032728e-06,
+      "loss": 0.775,
+      "step": 5789
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1908023113230792e-06,
+      "loss": 0.8361,
+      "step": 5790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1893851391126355e-06,
+      "loss": 0.8355,
+      "step": 5791
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1879686968434096e-06,
+      "loss": 1.1415,
+      "step": 5792
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1865529847867258e-06,
+      "loss": 0.7807,
+      "step": 5793
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1851380032137727e-06,
+      "loss": 0.4911,
+      "step": 5794
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1837237523955942e-06,
+      "loss": 0.6427,
+      "step": 5795
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1823102326030995e-06,
+      "loss": 0.8316,
+      "step": 5796
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1808974441070515e-06,
+      "loss": 0.8902,
+      "step": 5797
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1794853871780797e-06,
+      "loss": 0.7233,
+      "step": 5798
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1780740620866681e-06,
+      "loss": 0.6339,
+      "step": 5799
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1766634691031648e-06,
+      "loss": 0.5842,
+      "step": 5800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1752536084977727e-06,
+      "loss": 0.6782,
+      "step": 5801
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1738444805405608e-06,
+      "loss": 0.9415,
+      "step": 5802
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.17243608550145e-06,
+      "loss": 0.8079,
+      "step": 5803
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.171028423650229e-06,
+      "loss": 0.5801,
+      "step": 5804
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1696214952565387e-06,
+      "loss": 0.6212,
+      "step": 5805
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1682153005898855e-06,
+      "loss": 0.9907,
+      "step": 5806
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1668098399196304e-06,
+      "loss": 0.4268,
+      "step": 5807
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1654051135149985e-06,
+      "loss": 0.7622,
+      "step": 5808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.164001121645069e-06,
+      "loss": 0.9648,
+      "step": 5809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.162597864578786e-06,
+      "loss": 0.4302,
+      "step": 5810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.161195342584947e-06,
+      "loss": 0.827,
+      "step": 5811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.159793555932215e-06,
+      "loss": 0.9794,
+      "step": 5812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1583925048891048e-06,
+      "loss": 0.8687,
+      "step": 5813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.156992189723999e-06,
+      "loss": 0.4914,
+      "step": 5814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1555926107051296e-06,
+      "loss": 1.017,
+      "step": 5815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1541937681005972e-06,
+      "loss": 0.5538,
+      "step": 5816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.152795662178353e-06,
+      "loss": 0.6567,
+      "step": 5817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1513982932062135e-06,
+      "loss": 0.5989,
+      "step": 5818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1500016614518489e-06,
+      "loss": 0.7242,
+      "step": 5819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.148605767182794e-06,
+      "loss": 0.3953,
+      "step": 5820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1472106106664342e-06,
+      "loss": 1.0107,
+      "step": 5821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1458161921700228e-06,
+      "loss": 0.9907,
+      "step": 5822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.144422511960664e-06,
+      "loss": 0.4759,
+      "step": 5823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1430295703053257e-06,
+      "loss": 0.4765,
+      "step": 5824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1416373674708314e-06,
+      "loss": 1.1325,
+      "step": 5825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1402459037238651e-06,
+      "loss": 0.7379,
+      "step": 5826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1388551793309666e-06,
+      "loss": 0.6354,
+      "step": 5827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1374651945585385e-06,
+      "loss": 0.4209,
+      "step": 5828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.136075949672837e-06,
+      "loss": 0.9499,
+      "step": 5829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.134687444939977e-06,
+      "loss": 0.8136,
+      "step": 5830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.133299680625936e-06,
+      "loss": 0.2654,
+      "step": 5831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.131912656996545e-06,
+      "loss": 1.2203,
+      "step": 5832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.130526374317496e-06,
+      "loss": 0.4068,
+      "step": 5833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.129140832854337e-06,
+      "loss": 0.9102,
+      "step": 5834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1277560328724763e-06,
+      "loss": 0.5775,
+      "step": 5835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1263719746371764e-06,
+      "loss": 0.454,
+      "step": 5836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1249886584135634e-06,
+      "loss": 1.0729,
+      "step": 5837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1236060844666152e-06,
+      "loss": 0.8617,
+      "step": 5838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1222242530611727e-06,
+      "loss": 0.9689,
+      "step": 5839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.12084316446193e-06,
+      "loss": 0.1606,
+      "step": 5840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1194628189334433e-06,
+      "loss": 0.7752,
+      "step": 5841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1180832167401224e-06,
+      "loss": 1.1359,
+      "step": 5842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1167043581462366e-06,
+      "loss": 0.5575,
+      "step": 5843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1153262434159151e-06,
+      "loss": 0.9991,
+      "step": 5844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1139488728131399e-06,
+      "loss": 0.9673,
+      "step": 5845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1125722466017547e-06,
+      "loss": 0.6307,
+      "step": 5846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1111963650454561e-06,
+      "loss": 0.4105,
+      "step": 5847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1098212284078037e-06,
+      "loss": 1.0318,
+      "step": 5848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1084468369522083e-06,
+      "loss": 0.5166,
+      "step": 5849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1070731909419446e-06,
+      "loss": 0.6431,
+      "step": 5850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1057002906401372e-06,
+      "loss": 0.9314,
+      "step": 5851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1043281363097751e-06,
+      "loss": 0.598,
+      "step": 5852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1029567282136977e-06,
+      "loss": 0.7617,
+      "step": 5853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1015860666146084e-06,
+      "loss": 0.5641,
+      "step": 5854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1002161517750598e-06,
+      "loss": 0.4066,
+      "step": 5855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.098846983957469e-06,
+      "loss": 0.7013,
+      "step": 5856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0974785634241032e-06,
+      "loss": 0.8346,
+      "step": 5857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0961108904370937e-06,
+      "loss": 0.5304,
+      "step": 5858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0947439652584207e-06,
+      "loss": 0.7581,
+      "step": 5859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0933777881499285e-06,
+      "loss": 0.8824,
+      "step": 5860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0920123593733123e-06,
+      "loss": 0.5504,
+      "step": 5861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0906476791901289e-06,
+      "loss": 0.8464,
+      "step": 5862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0892837478617863e-06,
+      "loss": 0.9888,
+      "step": 5863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0879205656495546e-06,
+      "loss": 0.7377,
+      "step": 5864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0865581328145552e-06,
+      "loss": 0.5208,
+      "step": 5865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.085196449617772e-06,
+      "loss": 0.7344,
+      "step": 5866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0838355163200376e-06,
+      "loss": 0.4592,
+      "step": 5867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0824753331820488e-06,
+      "loss": 0.8227,
+      "step": 5868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0811159004643523e-06,
+      "loss": 0.6493,
+      "step": 5869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0797572184273553e-06,
+      "loss": 0.7712,
+      "step": 5870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.078399287331321e-06,
+      "loss": 0.7611,
+      "step": 5871
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0770421074363647e-06,
+      "loss": 1.0661,
+      "step": 5872
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0756856790024633e-06,
+      "loss": 0.6087,
+      "step": 5873
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.074330002289445e-06,
+      "loss": 0.8787,
+      "step": 5874
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0729750775569968e-06,
+      "loss": 0.6504,
+      "step": 5875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.071620905064661e-06,
+      "loss": 0.9086,
+      "step": 5876
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0702674850718359e-06,
+      "loss": 0.6769,
+      "step": 5877
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0689148178377746e-06,
+      "loss": 0.9168,
+      "step": 5878
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0675629036215885e-06,
+      "loss": 1.0944,
+      "step": 5879
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0662117426822404e-06,
+      "loss": 1.4913,
+      "step": 5880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0648613352785552e-06,
+      "loss": 0.6169,
+      "step": 5881
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0635116816692064e-06,
+      "loss": 0.8171,
+      "step": 5882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.062162782112729e-06,
+      "loss": 1.0199,
+      "step": 5883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0608146368675087e-06,
+      "loss": 0.6458,
+      "step": 5884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0594672461917921e-06,
+      "loss": 0.8979,
+      "step": 5885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0581206103436748e-06,
+      "loss": 0.5474,
+      "step": 5886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0567747295811143e-06,
+      "loss": 1.1613,
+      "step": 5887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0554296041619177e-06,
+      "loss": 0.7705,
+      "step": 5888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.054085234343753e-06,
+      "loss": 0.7042,
+      "step": 5889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0527416203841378e-06,
+      "loss": 0.8055,
+      "step": 5890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0513987625404498e-06,
+      "loss": 0.3999,
+      "step": 5891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.050056661069918e-06,
+      "loss": 0.8044,
+      "step": 5892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0487153162296299e-06,
+      "loss": 0.6256,
+      "step": 5893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0473747282765246e-06,
+      "loss": 0.8824,
+      "step": 5894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0460348974674e-06,
+      "loss": 0.4198,
+      "step": 5895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.044695824058905e-06,
+      "loss": 0.8044,
+      "step": 5896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0433575083075482e-06,
+      "loss": 0.6754,
+      "step": 5897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0420199504696864e-06,
+      "loss": 1.2506,
+      "step": 5898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.040683150801538e-06,
+      "loss": 0.7498,
+      "step": 5899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0393471095591733e-06,
+      "loss": 0.9829,
+      "step": 5900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0380118269985157e-06,
+      "loss": 1.0947,
+      "step": 5901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0366773033753462e-06,
+      "loss": 1.0824,
+      "step": 5902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353435389452981e-06,
+      "loss": 0.5434,
+      "step": 5903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0340105339638618e-06,
+      "loss": 0.682,
+      "step": 5904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0326782886863778e-06,
+      "loss": 1.1154,
+      "step": 5905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0313468033680473e-06,
+      "loss": 0.5357,
+      "step": 5906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0300160782639196e-06,
+      "loss": 1.0442,
+      "step": 5907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0286861136289045e-06,
+      "loss": 0.9095,
+      "step": 5908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.02735690971776e-06,
+      "loss": 0.8521,
+      "step": 5909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0260284667851033e-06,
+      "loss": 0.6818,
+      "step": 5910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0247007850854024e-06,
+      "loss": 0.8965,
+      "step": 5911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.023373864872983e-06,
+      "loss": 0.5313,
+      "step": 5912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.02204770640202e-06,
+      "loss": 0.6265,
+      "step": 5913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.020722309926549e-06,
+      "loss": 0.638,
+      "step": 5914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0193976757004525e-06,
+      "loss": 0.2029,
+      "step": 5915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0180738039774728e-06,
+      "loss": 0.5861,
+      "step": 5916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0167506950112021e-06,
+      "loss": 0.9143,
+      "step": 5917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0154283490550898e-06,
+      "loss": 0.9462,
+      "step": 5918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0141067663624354e-06,
+      "loss": 0.9666,
+      "step": 5919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.012785947186397e-06,
+      "loss": 0.5324,
+      "step": 5920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0114658917799808e-06,
+      "loss": 0.516,
+      "step": 5921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.010146600396052e-06,
+      "loss": 0.5738,
+      "step": 5922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.008828073287325e-06,
+      "loss": 0.9033,
+      "step": 5923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0075103107063721e-06,
+      "loss": 0.5887,
+      "step": 5924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.006193312905615e-06,
+      "loss": 1.2837,
+      "step": 5925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0048770801373313e-06,
+      "loss": 1.0023,
+      "step": 5926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0035616126536535e-06,
+      "loss": 0.3557,
+      "step": 5927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0022469107065625e-06,
+      "loss": 0.6518,
+      "step": 5928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0009329745478986e-06,
+      "loss": 0.8697,
+      "step": 5929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.996198044293493e-07,
+      "loss": 1.0336,
+      "step": 5930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.983074006024617e-07,
+      "loss": 0.9375,
+      "step": 5931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.969957633186306e-07,
+      "loss": 0.58,
+      "step": 5932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.95684892829108e-07,
+      "loss": 0.6919,
+      "step": 5933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.943747893849954e-07,
+      "loss": 1.1562,
+      "step": 5934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.930654532372518e-07,
+      "loss": 0.7262,
+      "step": 5935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.917568846366838e-07,
+      "loss": 0.9386,
+      "step": 5936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.904490838339564e-07,
+      "loss": 0.8854,
+      "step": 5937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.89142051079583e-07,
+      "loss": 0.8879,
+      "step": 5938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.87835786623934e-07,
+      "loss": 0.2968,
+      "step": 5939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.86530290717228e-07,
+      "loss": 0.9621,
+      "step": 5940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.85225563609542e-07,
+      "loss": 0.5102,
+      "step": 5941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.839216055507988e-07,
+      "loss": 0.9295,
+      "step": 5942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.826184167907815e-07,
+      "loss": 0.6383,
+      "step": 5943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.813159975791187e-07,
+      "loss": 0.8662,
+      "step": 5944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.80014348165298e-07,
+      "loss": 0.6072,
+      "step": 5945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.787134687986538e-07,
+      "loss": 0.8528,
+      "step": 5946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.77413359728378e-07,
+      "loss": 0.8318,
+      "step": 5947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.761140212035098e-07,
+      "loss": 0.9832,
+      "step": 5948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.74815453472947e-07,
+      "loss": 0.6876,
+      "step": 5949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.735176567854333e-07,
+      "loss": 0.8675,
+      "step": 5950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.722206313895705e-07,
+      "loss": 0.5253,
+      "step": 5951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.709243775338068e-07,
+      "loss": 0.846,
+      "step": 5952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.69628895466449e-07,
+      "loss": 0.5137,
+      "step": 5953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.683341854356504e-07,
+      "loss": 0.3328,
+      "step": 5954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.670402476894192e-07,
+      "loss": 1.1125,
+      "step": 5955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.65747082475617e-07,
+      "loss": 0.9756,
+      "step": 5956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.644546900419533e-07,
+      "loss": 0.7999,
+      "step": 5957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.63163070635994e-07,
+      "loss": 1.0526,
+      "step": 5958
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.618722245051532e-07,
+      "loss": 0.6444,
+      "step": 5959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.605821518967001e-07,
+      "loss": 0.6128,
+      "step": 5960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.592928530577523e-07,
+      "loss": 0.5489,
+      "step": 5961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.580043282352835e-07,
+      "loss": 0.9039,
+      "step": 5962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.56716577676114e-07,
+      "loss": 0.535,
+      "step": 5963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.55429601626921e-07,
+      "loss": 0.5887,
+      "step": 5964
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.541434003342282e-07,
+      "loss": 1.0062,
+      "step": 5965
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.52857974044416e-07,
+      "loss": 0.4595,
+      "step": 5966
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.515733230037121e-07,
+      "loss": 0.5347,
+      "step": 5967
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.502894474581992e-07,
+      "loss": 1.1081,
+      "step": 5968
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.490063476538075e-07,
+      "loss": 0.3394,
+      "step": 5969
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.477240238363233e-07,
+      "loss": 0.873,
+      "step": 5970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.464424762513791e-07,
+      "loss": 0.6255,
+      "step": 5971
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.45161705144464e-07,
+      "loss": 0.8546,
+      "step": 5972
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.438817107609127e-07,
+      "loss": 0.7078,
+      "step": 5973
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.426024933459177e-07,
+      "loss": 0.6809,
+      "step": 5974
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.413240531445161e-07,
+      "loss": 0.199,
+      "step": 5975
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.40046390401601e-07,
+      "loss": 0.5404,
+      "step": 5976
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.387695053619134e-07,
+      "loss": 0.6917,
+      "step": 5977
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.374933982700485e-07,
+      "loss": 0.5937,
+      "step": 5978
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.362180693704481e-07,
+      "loss": 0.7673,
+      "step": 5979
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.349435189074102e-07,
+      "loss": 0.6604,
+      "step": 5980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.336697471250778e-07,
+      "loss": 0.5349,
+      "step": 5981
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.323967542674501e-07,
+      "loss": 0.7099,
+      "step": 5982
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.311245405783753e-07,
+      "loss": 0.7733,
+      "step": 5983
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.298531063015498e-07,
+      "loss": 0.5465,
+      "step": 5984
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.285824516805253e-07,
+      "loss": 0.8012,
+      "step": 5985
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.273125769586994e-07,
+      "loss": 0.6236,
+      "step": 5986
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.260434823793246e-07,
+      "loss": 0.8032,
+      "step": 5987
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.247751681854999e-07,
+      "loss": 0.7373,
+      "step": 5988
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.235076346201788e-07,
+      "loss": 1.1953,
+      "step": 5989
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.222408819261614e-07,
+      "loss": 0.858,
+      "step": 5990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.209749103461024e-07,
+      "loss": 0.4414,
+      "step": 5991
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.197097201225019e-07,
+      "loss": 0.7251,
+      "step": 5992
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.18445311497716e-07,
+      "loss": 0.4209,
+      "step": 5993
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.171816847139447e-07,
+      "loss": 0.8015,
+      "step": 5994
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.159188400132457e-07,
+      "loss": 0.797,
+      "step": 5995
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.146567776375193e-07,
+      "loss": 0.7962,
+      "step": 5996
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.13395497828522e-07,
+      "loss": 0.7789,
+      "step": 5997
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.121350008278556e-07,
+      "loss": 1.0783,
+      "step": 5998
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.108752868769766e-07,
+      "loss": 1.0054,
+      "step": 5999
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.096163562171867e-07,
+      "loss": 0.5821,
+      "step": 6000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.083582090896425e-07,
+      "loss": 0.7456,
+      "step": 6001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.071008457353459e-07,
+      "loss": 0.6122,
+      "step": 6002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.058442663951528e-07,
+      "loss": 0.538,
+      "step": 6003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.045884713097642e-07,
+      "loss": 0.6012,
+      "step": 6004
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.03333460719737e-07,
+      "loss": 0.8362,
+      "step": 6005
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.020792348654706e-07,
+      "loss": 0.6785,
+      "step": 6006
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.008257939872211e-07,
+      "loss": 0.7889,
+      "step": 6007
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.995731383250883e-07,
+      "loss": 0.9698,
+      "step": 6008
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.983212681190273e-07,
+      "loss": 0.8145,
+      "step": 6009
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.970701836088369e-07,
+      "loss": 0.9978,
+      "step": 6010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.958198850341704e-07,
+      "loss": 0.7607,
+      "step": 6011
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.945703726345262e-07,
+      "loss": 0.8802,
+      "step": 6012
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.933216466492567e-07,
+      "loss": 0.9662,
+      "step": 6013
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.920737073175584e-07,
+      "loss": 0.7416,
+      "step": 6014
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.908265548784834e-07,
+      "loss": 1.3199,
+      "step": 6015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.895801895709261e-07,
+      "loss": 1.2877,
+      "step": 6016
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.883346116336367e-07,
+      "loss": 0.7759,
+      "step": 6017
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.870898213052092e-07,
+      "loss": 0.6295,
+      "step": 6018
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.858458188240909e-07,
+      "loss": 0.905,
+      "step": 6019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.846026044285744e-07,
+      "loss": 0.7461,
+      "step": 6020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.833601783568052e-07,
+      "loss": 0.7087,
+      "step": 6021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.821185408467742e-07,
+      "loss": 0.7507,
+      "step": 6022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.808776921363243e-07,
+      "loss": 0.5276,
+      "step": 6023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.796376324631445e-07,
+      "loss": 0.4673,
+      "step": 6024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.783983620647762e-07,
+      "loss": 0.8535,
+      "step": 6025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.771598811786047e-07,
+      "loss": 0.5289,
+      "step": 6026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.759221900418697e-07,
+      "loss": 0.7821,
+      "step": 6027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.746852888916541e-07,
+      "loss": 0.7275,
+      "step": 6028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.73449177964894e-07,
+      "loss": 0.9637,
+      "step": 6029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.722138574983713e-07,
+      "loss": 1.055,
+      "step": 6030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.709793277287182e-07,
+      "loss": 0.8758,
+      "step": 6031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.697455888924134e-07,
+      "loss": 0.5175,
+      "step": 6032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.685126412257877e-07,
+      "loss": 0.3451,
+      "step": 6033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.672804849650152e-07,
+      "loss": 0.8701,
+      "step": 6034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.660491203461241e-07,
+      "loss": 0.6457,
+      "step": 6035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.648185476049852e-07,
+      "loss": 1.0508,
+      "step": 6036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.63588766977323e-07,
+      "loss": 0.6064,
+      "step": 6037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.623597786987059e-07,
+      "loss": 1.0017,
+      "step": 6038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.611315830045541e-07,
+      "loss": 0.4812,
+      "step": 6039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.599041801301323e-07,
+      "loss": 0.9379,
+      "step": 6040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.586775703105571e-07,
+      "loss": 0.9364,
+      "step": 6041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.574517537807897e-07,
+      "loss": 0.6979,
+      "step": 6042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.562267307756434e-07,
+      "loss": 1.0949,
+      "step": 6043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.550025015297741e-07,
+      "loss": 0.8961,
+      "step": 6044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.537790662776918e-07,
+      "loss": 0.8112,
+      "step": 6045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.525564252537483e-07,
+      "loss": 0.7006,
+      "step": 6046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.513345786921489e-07,
+      "loss": 0.6247,
+      "step": 6047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.501135268269422e-07,
+      "loss": 0.6684,
+      "step": 6048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.488932698920277e-07,
+      "loss": 0.7436,
+      "step": 6049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.4767380812115e-07,
+      "loss": 1.0336,
+      "step": 6050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.464551417479044e-07,
+      "loss": 0.6912,
+      "step": 6051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.452372710057299e-07,
+      "loss": 0.7497,
+      "step": 6052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.440201961279182e-07,
+      "loss": 0.7058,
+      "step": 6053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.428039173476033e-07,
+      "loss": 0.6735,
+      "step": 6054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.41588434897771e-07,
+      "loss": 0.8256,
+      "step": 6055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.403737490112502e-07,
+      "loss": 0.9761,
+      "step": 6056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.391598599207229e-07,
+      "loss": 0.5796,
+      "step": 6057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.379467678587128e-07,
+      "loss": 0.6469,
+      "step": 6058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.367344730575949e-07,
+      "loss": 0.8396,
+      "step": 6059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.35522975749588e-07,
+      "loss": 0.5804,
+      "step": 6060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.343122761667633e-07,
+      "loss": 0.7846,
+      "step": 6061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.331023745410327e-07,
+      "loss": 0.6075,
+      "step": 6062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.318932711041611e-07,
+      "loss": 0.5771,
+      "step": 6063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.306849660877553e-07,
+      "loss": 1.0903,
+      "step": 6064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.294774597232741e-07,
+      "loss": 0.7861,
+      "step": 6065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.282707522420208e-07,
+      "loss": 0.9412,
+      "step": 6066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.27064843875145e-07,
+      "loss": 0.8616,
+      "step": 6067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.258597348536452e-07,
+      "loss": 0.6713,
+      "step": 6068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.246554254083639e-07,
+      "loss": 0.9278,
+      "step": 6069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.234519157699944e-07,
+      "loss": 0.2163,
+      "step": 6070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.222492061690724e-07,
+      "loss": 0.6592,
+      "step": 6071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.210472968359851e-07,
+      "loss": 0.8708,
+      "step": 6072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.198461880009617e-07,
+      "loss": 0.9523,
+      "step": 6073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.186458798940816e-07,
+      "loss": 0.8557,
+      "step": 6074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.174463727452686e-07,
+      "loss": 0.9738,
+      "step": 6075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.16247666784295e-07,
+      "loss": 0.6791,
+      "step": 6076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.150497622407772e-07,
+      "loss": 0.6218,
+      "step": 6077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.138526593441809e-07,
+      "loss": 0.531,
+      "step": 6078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.12656358323815e-07,
+      "loss": 0.8368,
+      "step": 6079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.114608594088391e-07,
+      "loss": 0.6198,
+      "step": 6080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.102661628282538e-07,
+      "loss": 0.4163,
+      "step": 6081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.090722688109115e-07,
+      "loss": 1.0907,
+      "step": 6082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.07879177585506e-07,
+      "loss": 0.5689,
+      "step": 6083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.066868893805812e-07,
+      "loss": 0.6112,
+      "step": 6084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.054954044245245e-07,
+      "loss": 1.0072,
+      "step": 6085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.043047229455714e-07,
+      "loss": 0.5498,
+      "step": 6086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.031148451718007e-07,
+      "loss": 0.7098,
+      "step": 6087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.019257713311418e-07,
+      "loss": 0.9388,
+      "step": 6088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.007375016513646e-07,
+      "loss": 0.4914,
+      "step": 6089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.9955003636009e-07,
+      "loss": 0.6321,
+      "step": 6090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.983633756847808e-07,
+      "loss": 0.8722,
+      "step": 6091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.971775198527493e-07,
+      "loss": 0.54,
+      "step": 6092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.959924690911497e-07,
+      "loss": 0.9631,
+      "step": 6093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.94808223626985e-07,
+      "loss": 0.7572,
+      "step": 6094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.936247836871041e-07,
+      "loss": 0.7385,
+      "step": 6095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.924421494981988e-07,
+      "loss": 0.4648,
+      "step": 6096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.912603212868097e-07,
+      "loss": 0.5513,
+      "step": 6097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.9007929927932e-07,
+      "loss": 0.4433,
+      "step": 6098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.888990837019617e-07,
+      "loss": 0.7311,
+      "step": 6099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.877196747808086e-07,
+      "loss": 0.7081,
+      "step": 6100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.865410727417844e-07,
+      "loss": 0.9345,
+      "step": 6101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.853632778106534e-07,
+      "loss": 1.0262,
+      "step": 6102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.841862902130298e-07,
+      "loss": 0.9848,
+      "step": 6103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.830101101743698e-07,
+      "loss": 0.5407,
+      "step": 6104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.818347379199781e-07,
+      "loss": 0.9268,
+      "step": 6105
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.806601736749997e-07,
+      "loss": 0.9825,
+      "step": 6106
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.794864176644313e-07,
+      "loss": 0.5941,
+      "step": 6107
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.783134701131084e-07,
+      "loss": 0.6244,
+      "step": 6108
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.771413312457171e-07,
+      "loss": 0.6472,
+      "step": 6109
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.759700012867837e-07,
+      "loss": 0.7404,
+      "step": 6110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.747994804606851e-07,
+      "loss": 0.7555,
+      "step": 6111
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.736297689916367e-07,
+      "loss": 0.7106,
+      "step": 6112
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.724608671037048e-07,
+      "loss": 0.5436,
+      "step": 6113
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.71292775020796e-07,
+      "loss": 0.856,
+      "step": 6114
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.701254929666657e-07,
+      "loss": 0.9005,
+      "step": 6115
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.689590211649106e-07,
+      "loss": 0.7418,
+      "step": 6116
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.677933598389759e-07,
+      "loss": 0.9047,
+      "step": 6117
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.666285092121473e-07,
+      "loss": 0.8567,
+      "step": 6118
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.654644695075597e-07,
+      "loss": 0.7524,
+      "step": 6119
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.643012409481876e-07,
+      "loss": 0.6289,
+      "step": 6120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.631388237568549e-07,
+      "loss": 0.7062,
+      "step": 6121
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.619772181562285e-07,
+      "loss": 0.3089,
+      "step": 6122
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.60816424368817e-07,
+      "loss": 0.6211,
+      "step": 6123
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.596564426169795e-07,
+      "loss": 0.7425,
+      "step": 6124
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.584972731229123e-07,
+      "loss": 0.6905,
+      "step": 6125
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.573389161086625e-07,
+      "loss": 0.8742,
+      "step": 6126
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.56181371796117e-07,
+      "loss": 0.8541,
+      "step": 6127
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.550246404070105e-07,
+      "loss": 1.1113,
+      "step": 6128
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.538687221629181e-07,
+      "loss": 0.9713,
+      "step": 6129
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.527136172852639e-07,
+      "loss": 0.7321,
+      "step": 6130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.515593259953119e-07,
+      "loss": 0.6543,
+      "step": 6131
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.504058485141735e-07,
+      "loss": 0.8523,
+      "step": 6132
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.492531850628004e-07,
+      "loss": 0.9118,
+      "step": 6133
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.481013358619937e-07,
+      "loss": 0.5859,
+      "step": 6134
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.469503011323931e-07,
+      "loss": 0.801,
+      "step": 6135
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.458000810944865e-07,
+      "loss": 0.9547,
+      "step": 6136
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.446506759686017e-07,
+      "loss": 0.678,
+      "step": 6137
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.435020859749159e-07,
+      "loss": 0.7374,
+      "step": 6138
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.423543113334436e-07,
+      "loss": 1.0654,
+      "step": 6139
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.41207352264049e-07,
+      "loss": 0.6124,
+      "step": 6140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.400612089864356e-07,
+      "loss": 0.7229,
+      "step": 6141
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.389158817201541e-07,
+      "loss": 0.6336,
+      "step": 6142
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.377713706845962e-07,
+      "loss": 0.9387,
+      "step": 6143
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.366276760989993e-07,
+      "loss": 0.9925,
+      "step": 6144
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.35484798182442e-07,
+      "loss": 0.8005,
+      "step": 6145
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.343427371538503e-07,
+      "loss": 1.0332,
+      "step": 6146
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.332014932319886e-07,
+      "loss": 1.0232,
+      "step": 6147
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.320610666354705e-07,
+      "loss": 0.6719,
+      "step": 6148
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.309214575827467e-07,
+      "loss": 1.0159,
+      "step": 6149
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.29782666292117e-07,
+      "loss": 0.7522,
+      "step": 6150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.286446929817225e-07,
+      "loss": 0.9478,
+      "step": 6151
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.275075378695462e-07,
+      "loss": 1.1955,
+      "step": 6152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.263712011734164e-07,
+      "loss": 0.5022,
+      "step": 6153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.252356831110024e-07,
+      "loss": 0.8989,
+      "step": 6154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.241009838998198e-07,
+      "loss": 1.0233,
+      "step": 6155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.229671037572233e-07,
+      "loss": 0.5954,
+      "step": 6156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.218340429004162e-07,
+      "loss": 0.3672,
+      "step": 6157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.207018015464379e-07,
+      "loss": 0.6989,
+      "step": 6158
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.195703799121778e-07,
+      "loss": 0.9813,
+      "step": 6159
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.184397782143626e-07,
+      "loss": 0.6401,
+      "step": 6160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.173099966695668e-07,
+      "loss": 0.6401,
+      "step": 6161
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.161810354942028e-07,
+      "loss": 1.063,
+      "step": 6162
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.150528949045304e-07,
+      "loss": 1.0486,
+      "step": 6163
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.139255751166485e-07,
+      "loss": 0.4366,
+      "step": 6164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.127990763465031e-07,
+      "loss": 0.8902,
+      "step": 6165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.116733988098773e-07,
+      "loss": 0.7429,
+      "step": 6166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.105485427224023e-07,
+      "loss": 0.5097,
+      "step": 6167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.094245082995482e-07,
+      "loss": 1.0298,
+      "step": 6168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.083012957566299e-07,
+      "loss": 0.8263,
+      "step": 6169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.071789053088029e-07,
+      "loss": 0.6537,
+      "step": 6170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.060573371710683e-07,
+      "loss": 0.5744,
+      "step": 6171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.049365915582656e-07,
+      "loss": 0.8921,
+      "step": 6172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.038166686850811e-07,
+      "loss": 0.6197,
+      "step": 6173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.026975687660386e-07,
+      "loss": 0.9299,
+      "step": 6174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.015792920155102e-07,
+      "loss": 0.6768,
+      "step": 6175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.004618386477041e-07,
+      "loss": 0.8065,
+      "step": 6176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.99345208876675e-07,
+      "loss": 0.6335,
+      "step": 6177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.982294029163195e-07,
+      "loss": 0.5345,
+      "step": 6178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.971144209803738e-07,
+      "loss": 0.9536,
+      "step": 6179
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.960002632824192e-07,
+      "loss": 0.6833,
+      "step": 6180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.948869300358769e-07,
+      "loss": 0.7416,
+      "step": 6181
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.937744214540126e-07,
+      "loss": 0.661,
+      "step": 6182
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.926627377499301e-07,
+      "loss": 0.7189,
+      "step": 6183
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.915518791365805e-07,
+      "loss": 0.7515,
+      "step": 6184
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.904418458267514e-07,
+      "loss": 0.7333,
+      "step": 6185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.893326380330778e-07,
+      "loss": 0.8851,
+      "step": 6186
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.882242559680303e-07,
+      "loss": 0.8875,
+      "step": 6187
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.871166998439277e-07,
+      "loss": 0.7073,
+      "step": 6188
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.86009969872925e-07,
+      "loss": 0.8291,
+      "step": 6189
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.849040662670237e-07,
+      "loss": 0.6351,
+      "step": 6190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.837989892380636e-07,
+      "loss": 1.0184,
+      "step": 6191
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.826947389977284e-07,
+      "loss": 0.9577,
+      "step": 6192
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.815913157575404e-07,
+      "loss": 0.9873,
+      "step": 6193
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.804887197288684e-07,
+      "loss": 0.9614,
+      "step": 6194
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.793869511229173e-07,
+      "loss": 0.8964,
+      "step": 6195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.782860101507383e-07,
+      "loss": 0.9659,
+      "step": 6196
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.771858970232198e-07,
+      "loss": 0.6258,
+      "step": 6197
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.760866119510951e-07,
+      "loss": 0.9535,
+      "step": 6198
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.749881551449361e-07,
+      "loss": 1.1807,
+      "step": 6199
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.738905268151591e-07,
+      "loss": 0.6384,
+      "step": 6200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.727937271720175e-07,
+      "loss": 0.4478,
+      "step": 6201
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.716977564256117e-07,
+      "loss": 0.8609,
+      "step": 6202
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.706026147858769e-07,
+      "loss": 0.556,
+      "step": 6203
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.695083024625943e-07,
+      "loss": 0.7931,
+      "step": 6204
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.68414819665385e-07,
+      "loss": 1.0231,
+      "step": 6205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.673221666037089e-07,
+      "loss": 0.7916,
+      "step": 6206
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.662303434868711e-07,
+      "loss": 0.9311,
+      "step": 6207
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.651393505240139e-07,
+      "loss": 0.7588,
+      "step": 6208
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.640491879241235e-07,
+      "loss": 0.6417,
+      "step": 6209
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.629598558960232e-07,
+      "loss": 0.8122,
+      "step": 6210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.618713546483829e-07,
+      "loss": 0.8748,
+      "step": 6211
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.607836843897069e-07,
+      "loss": 0.479,
+      "step": 6212
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.59696845328347e-07,
+      "loss": 0.5716,
+      "step": 6213
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.586108376724886e-07,
+      "loss": 0.9648,
+      "step": 6214
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.575256616301651e-07,
+      "loss": 0.8382,
+      "step": 6215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.564413174092443e-07,
+      "loss": 0.7683,
+      "step": 6216
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.553578052174398e-07,
+      "loss": 0.9664,
+      "step": 6217
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.542751252623009e-07,
+      "loss": 0.7482,
+      "step": 6218
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.531932777512228e-07,
+      "loss": 0.7521,
+      "step": 6219
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.521122628914356e-07,
+      "loss": 1.0465,
+      "step": 6220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.510320808900155e-07,
+      "loss": 0.5704,
+      "step": 6221
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.499527319538746e-07,
+      "loss": 0.9553,
+      "step": 6222
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.488742162897688e-07,
+      "loss": 0.7894,
+      "step": 6223
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.477965341042908e-07,
+      "loss": 1.262,
+      "step": 6224
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.467196856038777e-07,
+      "loss": 0.898,
+      "step": 6225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.456436709948033e-07,
+      "loss": 1.1664,
+      "step": 6226
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.445684904831851e-07,
+      "loss": 0.7313,
+      "step": 6227
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.434941442749764e-07,
+      "loss": 0.771,
+      "step": 6228
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.424206325759763e-07,
+      "loss": 0.6266,
+      "step": 6229
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.413479555918178e-07,
+      "loss": 0.9882,
+      "step": 6230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.402761135279806e-07,
+      "loss": 0.4481,
+      "step": 6231
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.392051065897781e-07,
+      "loss": 0.775,
+      "step": 6232
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.381349349823696e-07,
+      "loss": 0.6789,
+      "step": 6233
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.370655989107483e-07,
+      "loss": 0.9025,
+      "step": 6234
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.359970985797537e-07,
+      "loss": 1.0291,
+      "step": 6235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.349294341940593e-07,
+      "loss": 1.0386,
+      "step": 6236
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.338626059581831e-07,
+      "loss": 0.6403,
+      "step": 6237
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.327966140764802e-07,
+      "loss": 0.8604,
+      "step": 6238
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.317314587531476e-07,
+      "loss": 0.8427,
+      "step": 6239
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.306671401922182e-07,
+      "loss": 0.7102,
+      "step": 6240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.296036585975706e-07,
+      "loss": 0.6787,
+      "step": 6241
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.285410141729165e-07,
+      "loss": 0.9313,
+      "step": 6242
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.274792071218133e-07,
+      "loss": 0.7475,
+      "step": 6243
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.26418237647653e-07,
+      "loss": 1.0152,
+      "step": 6244
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.253581059536706e-07,
+      "loss": 1.0766,
+      "step": 6245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.242988122429378e-07,
+      "loss": 0.3478,
+      "step": 6246
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.232403567183698e-07,
+      "loss": 0.8082,
+      "step": 6247
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.221827395827157e-07,
+      "loss": 0.5583,
+      "step": 6248
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.211259610385701e-07,
+      "loss": 0.6783,
+      "step": 6249
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.200700212883609e-07,
+      "loss": 0.8189,
+      "step": 6250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.190149205343609e-07,
+      "loss": 0.7776,
+      "step": 6251
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.179606589786779e-07,
+      "loss": 0.7957,
+      "step": 6252
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.16907236823262e-07,
+      "loss": 0.9932,
+      "step": 6253
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.158546542698996e-07,
+      "loss": 0.8798,
+      "step": 6254
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.148029115202198e-07,
+      "loss": 0.5483,
+      "step": 6255
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.137520087756865e-07,
+      "loss": 0.5702,
+      "step": 6256
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.127019462376077e-07,
+      "loss": 0.9164,
+      "step": 6257
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.116527241071251e-07,
+      "loss": 0.7557,
+      "step": 6258
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.106043425852248e-07,
+      "loss": 0.8215,
+      "step": 6259
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.095568018727266e-07,
+      "loss": 0.9749,
+      "step": 6260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.085101021702944e-07,
+      "loss": 0.566,
+      "step": 6261
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.074642436784267e-07,
+      "loss": 1.3014,
+      "step": 6262
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.064192265974634e-07,
+      "loss": 0.9977,
+      "step": 6263
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.053750511275819e-07,
+      "loss": 0.5061,
+      "step": 6264
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.043317174687996e-07,
+      "loss": 0.6311,
+      "step": 6265
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.03289225820971e-07,
+      "loss": 0.6312,
+      "step": 6266
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.02247576383792e-07,
+      "loss": 1.0748,
+      "step": 6267
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.01206769356793e-07,
+      "loss": 0.5089,
+      "step": 6268
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.001668049393483e-07,
+      "loss": 0.809,
+      "step": 6269
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.991276833306648e-07,
+      "loss": 0.8998,
+      "step": 6270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.980894047297947e-07,
+      "loss": 0.4844,
+      "step": 6271
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.970519693356214e-07,
+      "loss": 0.7896,
+      "step": 6272
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.96015377346874e-07,
+      "loss": 0.4293,
+      "step": 6273
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.94979628962114e-07,
+      "loss": 1.0701,
+      "step": 6274
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.939447243797458e-07,
+      "loss": 0.8775,
+      "step": 6275
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.929106637980087e-07,
+      "loss": 0.6605,
+      "step": 6276
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.91877447414983e-07,
+      "loss": 0.6728,
+      "step": 6277
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.90845075428585e-07,
+      "loss": 0.6215,
+      "step": 6278
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.898135480365719e-07,
+      "loss": 0.7122,
+      "step": 6279
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.887828654365358e-07,
+      "loss": 0.6014,
+      "step": 6280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.877530278259108e-07,
+      "loss": 0.7415,
+      "step": 6281
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.867240354019649e-07,
+      "loss": 0.8901,
+      "step": 6282
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.856958883618091e-07,
+      "loss": 1.0331,
+      "step": 6283
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.846685869023877e-07,
+      "loss": 0.7874,
+      "step": 6284
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.836421312204865e-07,
+      "loss": 0.8265,
+      "step": 6285
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.82616521512726e-07,
+      "loss": 0.6406,
+      "step": 6286
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.815917579755692e-07,
+      "loss": 0.9744,
+      "step": 6287
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.805678408053117e-07,
+      "loss": 0.3801,
+      "step": 6288
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.795447701980917e-07,
+      "loss": 0.8774,
+      "step": 6289
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.785225463498828e-07,
+      "loss": 0.8761,
+      "step": 6290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.775011694564953e-07,
+      "loss": 0.4868,
+      "step": 6291
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.764806397135814e-07,
+      "loss": 0.8942,
+      "step": 6292
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.754609573166253e-07,
+      "loss": 0.8351,
+      "step": 6293
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.744421224609553e-07,
+      "loss": 0.8616,
+      "step": 6294
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.734241353417308e-07,
+      "loss": 0.6485,
+      "step": 6295
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.724069961539547e-07,
+      "loss": 0.9201,
+      "step": 6296
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.713907050924622e-07,
+      "loss": 1.01,
+      "step": 6297
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.70375262351931e-07,
+      "loss": 0.7448,
+      "step": 6298
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.693606681268727e-07,
+      "loss": 0.3107,
+      "step": 6299
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.683469226116378e-07,
+      "loss": 0.677,
+      "step": 6300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.67334026000414e-07,
+      "loss": 0.7046,
+      "step": 6301
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.66321978487227e-07,
+      "loss": 0.6241,
+      "step": 6302
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.65310780265938e-07,
+      "loss": 1.0033,
+      "step": 6303
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.643004315302486e-07,
+      "loss": 0.54,
+      "step": 6304
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.632909324736934e-07,
+      "loss": 1.0333,
+      "step": 6305
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.622822832896496e-07,
+      "loss": 0.9933,
+      "step": 6306
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.612744841713258e-07,
+      "loss": 0.8782,
+      "step": 6307
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.602675353117732e-07,
+      "loss": 0.5495,
+      "step": 6308
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.592614369038752e-07,
+      "loss": 0.5583,
+      "step": 6309
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.582561891403571e-07,
+      "loss": 0.8552,
+      "step": 6310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.572517922137771e-07,
+      "loss": 0.7962,
+      "step": 6311
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.562482463165331e-07,
+      "loss": 0.9673,
+      "step": 6312
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.552455516408584e-07,
+      "loss": 0.9477,
+      "step": 6313
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.542437083788249e-07,
+      "loss": 0.7908,
+      "step": 6314
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.53242716722338e-07,
+      "loss": 0.8785,
+      "step": 6315
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.52242576863145e-07,
+      "loss": 0.5454,
+      "step": 6316
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.512432889928271e-07,
+      "loss": 0.7629,
+      "step": 6317
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.502448533028004e-07,
+      "loss": 0.3068,
+      "step": 6318
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.492472699843232e-07,
+      "loss": 0.5028,
+      "step": 6319
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.482505392284842e-07,
+      "loss": 0.7061,
+      "step": 6320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.472546612262142e-07,
+      "loss": 0.9322,
+      "step": 6321
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.462596361682765e-07,
+      "loss": 1.1475,
+      "step": 6322
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.452654642452749e-07,
+      "loss": 0.7298,
+      "step": 6323
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.442721456476452e-07,
+      "loss": 0.7548,
+      "step": 6324
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.432796805656648e-07,
+      "loss": 0.8435,
+      "step": 6325
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.422880691894422e-07,
+      "loss": 0.924,
+      "step": 6326
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.412973117089288e-07,
+      "loss": 0.7295,
+      "step": 6327
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.403074083139048e-07,
+      "loss": 0.7478,
+      "step": 6328
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.393183591939949e-07,
+      "loss": 0.9037,
+      "step": 6329
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.38330164538653e-07,
+      "loss": 0.9911,
+      "step": 6330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.373428245371743e-07,
+      "loss": 0.5574,
+      "step": 6331
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.363563393786869e-07,
+      "loss": 1.012,
+      "step": 6332
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.353707092521581e-07,
+      "loss": 0.9761,
+      "step": 6333
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.343859343463886e-07,
+      "loss": 0.6509,
+      "step": 6334
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.334020148500191e-07,
+      "loss": 0.9003,
+      "step": 6335
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.324189509515209e-07,
+      "loss": 0.7071,
+      "step": 6336
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.31436742839207e-07,
+      "loss": 0.9554,
+      "step": 6337
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.304553907012216e-07,
+      "loss": 0.8848,
+      "step": 6338
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.294748947255502e-07,
+      "loss": 0.9811,
+      "step": 6339
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.284952551000089e-07,
+      "loss": 0.9048,
+      "step": 6340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.275164720122544e-07,
+      "loss": 1.0755,
+      "step": 6341
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.265385456497751e-07,
+      "loss": 0.7606,
+      "step": 6342
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.255614761998984e-07,
+      "loss": 0.87,
+      "step": 6343
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.245852638497872e-07,
+      "loss": 0.8299,
+      "step": 6344
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.236099087864382e-07,
+      "loss": 0.6795,
+      "step": 6345
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.226354111966875e-07,
+      "loss": 0.7218,
+      "step": 6346
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.216617712672018e-07,
+      "loss": 0.1884,
+      "step": 6347
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.206889891844886e-07,
+      "loss": 0.4385,
+      "step": 6348
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.197170651348871e-07,
+      "loss": 0.5197,
+      "step": 6349
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.187459993045762e-07,
+      "loss": 0.7638,
+      "step": 6350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.177757918795656e-07,
+      "loss": 0.8916,
+      "step": 6351
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.168064430457054e-07,
+      "loss": 0.8972,
+      "step": 6352
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.15837952988677e-07,
+      "loss": 0.8101,
+      "step": 6353
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.148703218940015e-07,
+      "loss": 0.7617,
+      "step": 6354
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.139035499470302e-07,
+      "loss": 0.7433,
+      "step": 6355
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.129376373329559e-07,
+      "loss": 1.0522,
+      "step": 6356
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.119725842368012e-07,
+      "loss": 0.8292,
+      "step": 6357
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.110083908434282e-07,
+      "loss": 0.6749,
+      "step": 6358
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.100450573375315e-07,
+      "loss": 1.0307,
+      "step": 6359
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.090825839036434e-07,
+      "loss": 0.7895,
+      "step": 6360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.081209707261292e-07,
+      "loss": 1.0523,
+      "step": 6361
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.07160217989191e-07,
+      "loss": 0.9063,
+      "step": 6362
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.06200325876865e-07,
+      "loss": 1.0377,
+      "step": 6363
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.05241294573024e-07,
+      "loss": 0.4104,
+      "step": 6364
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.042831242613733e-07,
+      "loss": 0.8413,
+      "step": 6365
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.033258151254572e-07,
+      "loss": 0.9045,
+      "step": 6366
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.023693673486502e-07,
+      "loss": 0.4651,
+      "step": 6367
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.014137811141673e-07,
+      "loss": 0.6544,
+      "step": 6368
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.004590566050527e-07,
+      "loss": 0.8784,
+      "step": 6369
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.995051940041901e-07,
+      "loss": 0.9555,
+      "step": 6370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.985521934942955e-07,
+      "loss": 0.7585,
+      "step": 6371
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.976000552579208e-07,
+      "loss": 0.679,
+      "step": 6372
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.966487794774538e-07,
+      "loss": 0.926,
+      "step": 6373
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.95698366335114e-07,
+      "loss": 0.9687,
+      "step": 6374
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.947488160129593e-07,
+      "loss": 0.8561,
+      "step": 6375
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.93800128692879e-07,
+      "loss": 0.509,
+      "step": 6376
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.928523045565997e-07,
+      "loss": 0.6912,
+      "step": 6377
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.919053437856802e-07,
+      "loss": 0.6755,
+      "step": 6378
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.909592465615171e-07,
+      "loss": 0.6927,
+      "step": 6379
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.90014013065338e-07,
+      "loss": 1.0534,
+      "step": 6380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.890696434782089e-07,
+      "loss": 0.814,
+      "step": 6381
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.881261379810254e-07,
+      "loss": 1.0256,
+      "step": 6382
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.871834967545236e-07,
+      "loss": 0.8146,
+      "step": 6383
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.862417199792685e-07,
+      "loss": 0.5709,
+      "step": 6384
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.85300807835663e-07,
+      "loss": 1.0378,
+      "step": 6385
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.843607605039424e-07,
+      "loss": 0.4193,
+      "step": 6386
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.834215781641787e-07,
+      "loss": 0.4238,
+      "step": 6387
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.824832609962748e-07,
+      "loss": 0.6783,
+      "step": 6388
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.815458091799719e-07,
+      "loss": 0.9513,
+      "step": 6389
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.806092228948405e-07,
+      "loss": 0.9354,
+      "step": 6390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.796735023202914e-07,
+      "loss": 0.5777,
+      "step": 6391
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.787386476355632e-07,
+      "loss": 0.7837,
+      "step": 6392
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.778046590197343e-07,
+      "loss": 0.9327,
+      "step": 6393
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.768715366517124e-07,
+      "loss": 0.712,
+      "step": 6394
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7593928071024363e-07,
+      "loss": 0.9209,
+      "step": 6395
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.750078913739037e-07,
+      "loss": 0.7074,
+      "step": 6396
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.74077368821107e-07,
+      "loss": 0.8674,
+      "step": 6397
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.731477132300971e-07,
+      "loss": 0.654,
+      "step": 6398
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7221892477895547e-07,
+      "loss": 0.5969,
+      "step": 6399
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.712910036455959e-07,
+      "loss": 0.4301,
+      "step": 6400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7036395000776556e-07,
+      "loss": 0.714,
+      "step": 6401
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.694377640430464e-07,
+      "loss": 0.8149,
+      "step": 6402
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6851244592885313e-07,
+      "loss": 0.6376,
+      "step": 6403
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.675879958424356e-07,
+      "loss": 0.8143,
+      "step": 6404
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.66664413960875e-07,
+      "loss": 0.6974,
+      "step": 6405
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6574170046109034e-07,
+      "loss": 0.8751,
+      "step": 6406
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6481985551982857e-07,
+      "loss": 0.7217,
+      "step": 6407
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.638988793136767e-07,
+      "loss": 0.8416,
+      "step": 6408
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6297877201904927e-07,
+      "loss": 1.0693,
+      "step": 6409
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6205953381219914e-07,
+      "loss": 0.6119,
+      "step": 6410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.611411648692088e-07,
+      "loss": 0.3997,
+      "step": 6411
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6022366536599863e-07,
+      "loss": 0.8651,
+      "step": 6412
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5930703547831743e-07,
+      "loss": 0.7145,
+      "step": 6413
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.583912753817521e-07,
+      "loss": 1.098,
+      "step": 6414
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.574763852517189e-07,
+      "loss": 0.7102,
+      "step": 6415
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.565623652634715e-07,
+      "loss": 0.5876,
+      "step": 6416
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5564921559209217e-07,
+      "loss": 1.0665,
+      "step": 6417
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.547369364125015e-07,
+      "loss": 0.4214,
+      "step": 6418
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.538255278994491e-07,
+      "loss": 0.6861,
+      "step": 6419
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.529149902275204e-07,
+      "loss": 0.8948,
+      "step": 6420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5200532357113304e-07,
+      "loss": 0.8355,
+      "step": 6421
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.510965281045382e-07,
+      "loss": 0.7785,
+      "step": 6422
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5018860400181885e-07,
+      "loss": 0.7467,
+      "step": 6423
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.492815514368937e-07,
+      "loss": 1.182,
+      "step": 6424
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4837537058351145e-07,
+      "loss": 0.6376,
+      "step": 6425
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4747006161525665e-07,
+      "loss": 0.6346,
+      "step": 6426
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.465656247055439e-07,
+      "loss": 0.6101,
+      "step": 6427
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.456620600276235e-07,
+      "loss": 0.5433,
+      "step": 6428
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.447593677545781e-07,
+      "loss": 0.4598,
+      "step": 6429
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.43857548059321e-07,
+      "loss": 0.8802,
+      "step": 6430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.429566011146019e-07,
+      "loss": 0.9224,
+      "step": 6431
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.420565270929994e-07,
+      "loss": 1.0675,
+      "step": 6432
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4115732616692886e-07,
+      "loss": 0.2931,
+      "step": 6433
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.402589985086353e-07,
+      "loss": 0.7596,
+      "step": 6434
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3936154429019717e-07,
+      "loss": 0.7051,
+      "step": 6435
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3846496368352756e-07,
+      "loss": 0.5518,
+      "step": 6436
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3756925686036897e-07,
+      "loss": 0.6734,
+      "step": 6437
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3667442399229985e-07,
+      "loss": 0.7652,
+      "step": 6438
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3578046525072805e-07,
+      "loss": 0.5477,
+      "step": 6439
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3488738080689764e-07,
+      "loss": 0.9525,
+      "step": 6440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.339951708318807e-07,
+      "loss": 0.8475,
+      "step": 6441
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3310383549658664e-07,
+      "loss": 0.9018,
+      "step": 6442
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3221337497175277e-07,
+      "loss": 0.8046,
+      "step": 6443
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.313237894279532e-07,
+      "loss": 0.818,
+      "step": 6444
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.304350790355899e-07,
+      "loss": 0.7242,
+      "step": 6445
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2954724396490175e-07,
+      "loss": 0.7424,
+      "step": 6446
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2866028438595607e-07,
+      "loss": 0.9022,
+      "step": 6447
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2777420046865524e-07,
+      "loss": 0.9214,
+      "step": 6448
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2688899238273187e-07,
+      "loss": 0.7877,
+      "step": 6449
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.26004660297753e-07,
+      "loss": 0.9213,
+      "step": 6450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2512120438311434e-07,
+      "loss": 0.916,
+      "step": 6451
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2423862480804876e-07,
+      "loss": 0.8916,
+      "step": 6452
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.233569217416167e-07,
+      "loss": 0.4481,
+      "step": 6453
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.224760953527135e-07,
+      "loss": 0.7443,
+      "step": 6454
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.215961458100648e-07,
+      "loss": 0.9929,
+      "step": 6455
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.207170732822302e-07,
+      "loss": 0.5789,
+      "step": 6456
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.198388779375984e-07,
+      "loss": 0.5733,
+      "step": 6457
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.189615599443936e-07,
+      "loss": 0.8204,
+      "step": 6458
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.180851194706692e-07,
+      "loss": 0.2504,
+      "step": 6459
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.172095566843121e-07,
+      "loss": 0.811,
+      "step": 6460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1633487175303967e-07,
+      "loss": 0.7932,
+      "step": 6461
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1546106484440353e-07,
+      "loss": 0.7037,
+      "step": 6462
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.145881361257831e-07,
+      "loss": 1.0906,
+      "step": 6463
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1371608576439404e-07,
+      "loss": 0.5097,
+      "step": 6464
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1284491392728055e-07,
+      "loss": 0.7529,
+      "step": 6465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.119746207813208e-07,
+      "loss": 0.8382,
+      "step": 6466
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1110520649322195e-07,
+      "loss": 0.6457,
+      "step": 6467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.102366712295264e-07,
+      "loss": 0.4278,
+      "step": 6468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.093690151566038e-07,
+      "loss": 0.751,
+      "step": 6469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0850223844066005e-07,
+      "loss": 0.8092,
+      "step": 6470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0763634124772853e-07,
+      "loss": 0.8193,
+      "step": 6471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.067713237436771e-07,
+      "loss": 0.8028,
+      "step": 6472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.059071860942032e-07,
+      "loss": 0.6617,
+      "step": 6473
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0504392846483775e-07,
+      "loss": 0.8978,
+      "step": 6474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.041815510209396e-07,
+      "loss": 0.881,
+      "step": 6475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.033200539277032e-07,
+      "loss": 1.106,
+      "step": 6476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.024594373501511e-07,
+      "loss": 0.8876,
+      "step": 6477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.015997014531398e-07,
+      "loss": 0.728,
+      "step": 6478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0074084640135346e-07,
+      "loss": 0.5147,
+      "step": 6479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.998828723593129e-07,
+      "loss": 0.7465,
+      "step": 6480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.990257794913643e-07,
+      "loss": 0.5816,
+      "step": 6481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.981695679616893e-07,
+      "loss": 0.7048,
+      "step": 6482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.973142379342992e-07,
+      "loss": 0.9237,
+      "step": 6483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9645978957303597e-07,
+      "loss": 0.8732,
+      "step": 6484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9560622304157417e-07,
+      "loss": 1.1546,
+      "step": 6485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9475353850341734e-07,
+      "loss": 0.1719,
+      "step": 6486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.939017361219022e-07,
+      "loss": 0.8523,
+      "step": 6487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.930508160601937e-07,
+      "loss": 0.6842,
+      "step": 6488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.922007784812926e-07,
+      "loss": 0.6136,
+      "step": 6489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.913516235480247e-07,
+      "loss": 0.943,
+      "step": 6490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9050335142305107e-07,
+      "loss": 0.9728,
+      "step": 6491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8965596226886173e-07,
+      "loss": 0.7102,
+      "step": 6492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.88809456247779e-07,
+      "loss": 0.7785,
+      "step": 6493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.879638335219538e-07,
+      "loss": 0.4688,
+      "step": 6494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8711909425337036e-07,
+      "loss": 0.8479,
+      "step": 6495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8627523860384097e-07,
+      "loss": 0.8075,
+      "step": 6496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.854322667350119e-07,
+      "loss": 0.7913,
+      "step": 6497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8459017880835667e-07,
+      "loss": 0.5989,
+      "step": 6498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8374897498518303e-07,
+      "loss": 0.8402,
+      "step": 6499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8290865542662537e-07,
+      "loss": 0.7744,
+      "step": 6500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8206922029365326e-07,
+      "loss": 0.5078,
+      "step": 6501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.812306697470619e-07,
+      "loss": 0.8461,
+      "step": 6502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8039300394748234e-07,
+      "loss": 0.7599,
+      "step": 6503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7955622305537065e-07,
+      "loss": 0.7973,
+      "step": 6504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7872032723101803e-07,
+      "loss": 0.9795,
+      "step": 6505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.778853166345436e-07,
+      "loss": 1.0481,
+      "step": 6506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7705119142589784e-07,
+      "loss": 0.8043,
+      "step": 6507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7621795176486064e-07,
+      "loss": 0.9757,
+      "step": 6508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.753855978110449e-07,
+      "loss": 0.7219,
+      "step": 6509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7455412972388915e-07,
+      "loss": 0.7957,
+      "step": 6510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7372354766266663e-07,
+      "loss": 0.5234,
+      "step": 6511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.728938517864794e-07,
+      "loss": 0.6907,
+      "step": 6512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.720650422542588e-07,
+      "loss": 0.9043,
+      "step": 6513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.712371192247688e-07,
+      "loss": 0.8439,
+      "step": 6514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.704100828565993e-07,
+      "loss": 0.4256,
+      "step": 6515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6958393330817576e-07,
+      "loss": 0.9238,
+      "step": 6516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6875867073774883e-07,
+      "loss": 0.558,
+      "step": 6517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6793429530340376e-07,
+      "loss": 0.8797,
+      "step": 6518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.671108071630508e-07,
+      "loss": 0.8354,
+      "step": 6519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.66288206474435e-07,
+      "loss": 0.8099,
+      "step": 6520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.654664933951285e-07,
+      "loss": 0.6438,
+      "step": 6521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.64645668082535e-07,
+      "loss": 0.7297,
+      "step": 6522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.638257306938864e-07,
+      "loss": 0.997,
+      "step": 6523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.630066813862465e-07,
+      "loss": 0.708,
+      "step": 6524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6218852031650765e-07,
+      "loss": 0.7103,
+      "step": 6525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6137124764139276e-07,
+      "loss": 0.7931,
+      "step": 6526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6055486351745327e-07,
+      "loss": 0.5934,
+      "step": 6527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.597393681010736e-07,
+      "loss": 0.94,
+      "step": 6528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.589247615484631e-07,
+      "loss": 0.707,
+      "step": 6529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5811104401566545e-07,
+      "loss": 0.6346,
+      "step": 6530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5729821565855026e-07,
+      "loss": 0.7724,
+      "step": 6531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.564862766328209e-07,
+      "loss": 0.2953,
+      "step": 6532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.556752270940056e-07,
+      "loss": 0.8733,
+      "step": 6533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5486506719746736e-07,
+      "loss": 0.3987,
+      "step": 6534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5405579709839376e-07,
+      "loss": 0.6165,
+      "step": 6535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5324741695180633e-07,
+      "loss": 0.84,
+      "step": 6536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.524399269125517e-07,
+      "loss": 0.835,
+      "step": 6537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5163332713531063e-07,
+      "loss": 0.7544,
+      "step": 6538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.508276177745901e-07,
+      "loss": 0.4048,
+      "step": 6539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.500227989847277e-07,
+      "loss": 0.682,
+      "step": 6540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4921887091989073e-07,
+      "loss": 0.8952,
+      "step": 6541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.484158337340737e-07,
+      "loss": 1.1873,
+      "step": 6542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.476136875811048e-07,
+      "loss": 0.7343,
+      "step": 6543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4681243261463607e-07,
+      "loss": 0.8029,
+      "step": 6544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.460120689881541e-07,
+      "loss": 0.8105,
+      "step": 6545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4521259685497076e-07,
+      "loss": 0.5024,
+      "step": 6546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.444140163682297e-07,
+      "loss": 0.813,
+      "step": 6547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.436163276809018e-07,
+      "loss": 0.7446,
+      "step": 6548
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4281953094578877e-07,
+      "loss": 1.1263,
+      "step": 6549
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.420236263155202e-07,
+      "loss": 0.7701,
+      "step": 6550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.412286139425569e-07,
+      "loss": 0.9369,
+      "step": 6551
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4043449397918484e-07,
+      "loss": 0.9521,
+      "step": 6552
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.396412665775234e-07,
+      "loss": 0.6901,
+      "step": 6553
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.388489318895177e-07,
+      "loss": 0.5735,
+      "step": 6554
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3805749006694424e-07,
+      "loss": 0.6446,
+      "step": 6555
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3726694126140657e-07,
+      "loss": 0.5292,
+      "step": 6556
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3647728562433925e-07,
+      "loss": 0.7814,
+      "step": 6557
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3568852330700286e-07,
+      "loss": 0.5184,
+      "step": 6558
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.349006544604899e-07,
+      "loss": 0.7824,
+      "step": 6559
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3411367923571913e-07,
+      "loss": 0.6172,
+      "step": 6560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3332759778344105e-07,
+      "loss": 0.5611,
+      "step": 6561
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3254241025423127e-07,
+      "loss": 0.9454,
+      "step": 6562
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3175811679849843e-07,
+      "loss": 0.6125,
+      "step": 6563
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3097471756647516e-07,
+      "loss": 0.7015,
+      "step": 6564
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3019221270822743e-07,
+      "loss": 0.8502,
+      "step": 6565
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.29410602373646e-07,
+      "loss": 0.871,
+      "step": 6566
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2862988671245324e-07,
+      "loss": 0.8589,
+      "step": 6567
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2785006587419844e-07,
+      "loss": 0.7284,
+      "step": 6568
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.270711400082599e-07,
+      "loss": 0.6343,
+      "step": 6569
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2629310926384493e-07,
+      "loss": 1.0658,
+      "step": 6570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.255159737899882e-07,
+      "loss": 0.6909,
+      "step": 6571
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.247397337355546e-07,
+      "loss": 0.8583,
+      "step": 6572
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2396438924923556e-07,
+      "loss": 0.9282,
+      "step": 6573
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2318994047955354e-07,
+      "loss": 1.0937,
+      "step": 6574
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.224163875748559e-07,
+      "loss": 0.4297,
+      "step": 6575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.21643730683322e-07,
+      "loss": 0.6025,
+      "step": 6576
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.208719699529561e-07,
+      "loss": 0.9186,
+      "step": 6577
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.20101105531595e-07,
+      "loss": 0.8437,
+      "step": 6578
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.193311375668989e-07,
+      "loss": 0.8825,
+      "step": 6579
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.18562066206361e-07,
+      "loss": 0.6213,
+      "step": 6580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1779389159729833e-07,
+      "loss": 1.0685,
+      "step": 6581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.170266138868605e-07,
+      "loss": 0.8496,
+      "step": 6582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.162602332220216e-07,
+      "loss": 0.941,
+      "step": 6583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1549474974958694e-07,
+      "loss": 0.823,
+      "step": 6584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.147301636161865e-07,
+      "loss": 0.4293,
+      "step": 6585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1396647496828245e-07,
+      "loss": 0.4961,
+      "step": 6586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1320368395216074e-07,
+      "loss": 0.6404,
+      "step": 6587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1244179071393987e-07,
+      "loss": 0.4416,
+      "step": 6588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1168079539956267e-07,
+      "loss": 0.7806,
+      "step": 6589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1092069815480186e-07,
+      "loss": 0.5411,
+      "step": 6590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.101614991252566e-07,
+      "loss": 0.7882,
+      "step": 6591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.094031984563572e-07,
+      "loss": 0.5081,
+      "step": 6592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.086457962933576e-07,
+      "loss": 1.085,
+      "step": 6593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.078892927813426e-07,
+      "loss": 0.8491,
+      "step": 6594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0713368806522483e-07,
+      "loss": 0.8146,
+      "step": 6595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0637898228974174e-07,
+      "loss": 1.1493,
+      "step": 6596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.056251755994638e-07,
+      "loss": 1.0414,
+      "step": 6597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0487226813878336e-07,
+      "loss": 1.0385,
+      "step": 6598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0412026005192496e-07,
+      "loss": 0.6289,
+      "step": 6599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.03369151482939e-07,
+      "loss": 0.7071,
+      "step": 6600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.026189425757042e-07,
+      "loss": 0.615,
+      "step": 6601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.018696334739252e-07,
+      "loss": 0.6482,
+      "step": 6602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0112122432113754e-07,
+      "loss": 0.623,
+      "step": 6603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0037371526070114e-07,
+      "loss": 0.4109,
+      "step": 6604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9962710643580584e-07,
+      "loss": 0.8969,
+      "step": 6605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9888139798946723e-07,
+      "loss": 0.5353,
+      "step": 6606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9813659006453053e-07,
+      "loss": 0.8528,
+      "step": 6607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9739268280366495e-07,
+      "loss": 0.8459,
+      "step": 6608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.96649676349372e-07,
+      "loss": 0.8841,
+      "step": 6609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9590757084397625e-07,
+      "loss": 0.9332,
+      "step": 6610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.951663664296328e-07,
+      "loss": 0.5312,
+      "step": 6611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9442606324832146e-07,
+      "loss": 0.8896,
+      "step": 6612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9368666144185266e-07,
+      "loss": 0.7273,
+      "step": 6613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9294816115185976e-07,
+      "loss": 0.7973,
+      "step": 6614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.922105625198085e-07,
+      "loss": 0.9009,
+      "step": 6615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9147386568698755e-07,
+      "loss": 0.6611,
+      "step": 6616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.907380707945162e-07,
+      "loss": 0.5012,
+      "step": 6617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9000317798333734e-07,
+      "loss": 0.2788,
+      "step": 6618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.892691873942255e-07,
+      "loss": 0.4611,
+      "step": 6619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8853609916777827e-07,
+      "loss": 0.5595,
+      "step": 6620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.878039134444222e-07,
+      "loss": 0.5247,
+      "step": 6621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.870726303644128e-07,
+      "loss": 0.8231,
+      "step": 6622
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8634225006782867e-07,
+      "loss": 0.9422,
+      "step": 6623
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.85612772694579e-07,
+      "loss": 1.0014,
+      "step": 6624
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.848841983843975e-07,
+      "loss": 0.7373,
+      "step": 6625
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.841565272768471e-07,
+      "loss": 0.582,
+      "step": 6626
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.834297595113156e-07,
+      "loss": 0.6333,
+      "step": 6627
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8270389522702003e-07,
+      "loss": 0.7125,
+      "step": 6628
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.819789345630014e-07,
+      "loss": 0.7648,
+      "step": 6629
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8125487765813085e-07,
+      "loss": 0.9016,
+      "step": 6630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.805317246511041e-07,
+      "loss": 0.7653,
+      "step": 6631
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.798094756804448e-07,
+      "loss": 0.5903,
+      "step": 6632
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.790881308845028e-07,
+      "loss": 1.041,
+      "step": 6633
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7836769040145593e-07,
+      "loss": 1.1934,
+      "step": 6634
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7764815436930713e-07,
+      "loss": 0.5367,
+      "step": 6635
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7692952292588726e-07,
+      "loss": 0.8324,
+      "step": 6636
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.762117962088529e-07,
+      "loss": 0.7565,
+      "step": 6637
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.754949743556895e-07,
+      "loss": 1.2179,
+      "step": 6638
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.747790575037057e-07,
+      "loss": 0.5978,
+      "step": 6639
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7406404579004067e-07,
+      "loss": 0.8928,
+      "step": 6640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73349939351657e-07,
+      "loss": 0.7294,
+      "step": 6641
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.726367383253459e-07,
+      "loss": 0.2944,
+      "step": 6642
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.71924442847723e-07,
+      "loss": 0.3919,
+      "step": 6643
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.712130530552337e-07,
+      "loss": 0.5349,
+      "step": 6644
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7050256908414674e-07,
+      "loss": 0.9084,
+      "step": 6645
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6979299107055936e-07,
+      "loss": 0.7951,
+      "step": 6646
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.690843191503939e-07,
+      "loss": 0.7004,
+      "step": 6647
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.683765534594013e-07,
+      "loss": 0.7506,
+      "step": 6648
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6766969413315524e-07,
+      "loss": 0.7962,
+      "step": 6649
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6696374130705973e-07,
+      "loss": 0.7759,
+      "step": 6650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.662586951163421e-07,
+      "loss": 0.9768,
+      "step": 6651
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.655545556960587e-07,
+      "loss": 0.3946,
+      "step": 6652
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6485132318108895e-07,
+      "loss": 0.8468,
+      "step": 6653
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.641489977061423e-07,
+      "loss": 0.8788,
+      "step": 6654
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.634475794057506e-07,
+      "loss": 0.9973,
+      "step": 6655
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6274706841427524e-07,
+      "loss": 0.3784,
+      "step": 6656
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.620474648659016e-07,
+      "loss": 0.7136,
+      "step": 6657
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.613487688946431e-07,
+      "loss": 0.9571,
+      "step": 6658
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6065098063433657e-07,
+      "loss": 0.7896,
+      "step": 6659
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.599541002186479e-07,
+      "loss": 0.6105,
+      "step": 6660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.592581277810674e-07,
+      "loss": 0.7425,
+      "step": 6661
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.585630634549119e-07,
+      "loss": 0.6855,
+      "step": 6662
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5786890737332424e-07,
+      "loss": 0.5331,
+      "step": 6663
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.571756596692737e-07,
+      "loss": 1.02,
+      "step": 6664
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.56483320475554e-07,
+      "loss": 1.0684,
+      "step": 6665
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.557918899247874e-07,
+      "loss": 0.7753,
+      "step": 6666
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5510136814941967e-07,
+      "loss": 0.6415,
+      "step": 6667
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5441175528172435e-07,
+      "loss": 0.9347,
+      "step": 6668
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.537230514537986e-07,
+      "loss": 0.5884,
+      "step": 6669
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.530352567975686e-07,
+      "loss": 0.5836,
+      "step": 6670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5234837144478333e-07,
+      "loss": 0.5716,
+      "step": 6671
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5166239552701974e-07,
+      "loss": 0.8082,
+      "step": 6672
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5097732917567896e-07,
+      "loss": 0.8978,
+      "step": 6673
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5029317252199037e-07,
+      "loss": 0.452,
+      "step": 6674
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.496099256970047e-07,
+      "loss": 0.5003,
+      "step": 6675
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4892758883160396e-07,
+      "loss": 0.6863,
+      "step": 6676
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4824616205649023e-07,
+      "loss": 1.0762,
+      "step": 6677
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4756564550219685e-07,
+      "loss": 0.9248,
+      "step": 6678
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.468860392990774e-07,
+      "loss": 0.8464,
+      "step": 6679
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4620734357731545e-07,
+      "loss": 0.8891,
+      "step": 6680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4552955846691705e-07,
+      "loss": 0.5719,
+      "step": 6681
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.448526840977161e-07,
+      "loss": 0.7548,
+      "step": 6682
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.441767205993706e-07,
+      "loss": 0.8437,
+      "step": 6683
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.435016681013652e-07,
+      "loss": 0.885,
+      "step": 6684
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.428275267330077e-07,
+      "loss": 0.6455,
+      "step": 6685
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4215429662343526e-07,
+      "loss": 0.6022,
+      "step": 6686
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.41481977901607e-07,
+      "loss": 1.0082,
+      "step": 6687
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4081057069630873e-07,
+      "loss": 0.4163,
+      "step": 6688
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.401400751361521e-07,
+      "loss": 0.744,
+      "step": 6689
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.394704913495738e-07,
+      "loss": 0.7177,
+      "step": 6690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.388018194648345e-07,
+      "loss": 0.9348,
+      "step": 6691
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3813405961002345e-07,
+      "loss": 0.6349,
+      "step": 6692
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.374672119130511e-07,
+      "loss": 0.481,
+      "step": 6693
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3680127650165696e-07,
+      "loss": 0.9734,
+      "step": 6694
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3613625350340286e-07,
+      "loss": 0.8926,
+      "step": 6695
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3547214304567855e-07,
+      "loss": 0.9202,
+      "step": 6696
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3480894525569564e-07,
+      "loss": 1.056,
+      "step": 6697
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.341466602604947e-07,
+      "loss": 0.8866,
+      "step": 6698
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3348528818693817e-07,
+      "loss": 0.6189,
+      "step": 6699
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3282482916171578e-07,
+      "loss": 0.7008,
+      "step": 6700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3216528331134026e-07,
+      "loss": 0.5646,
+      "step": 6701
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3150665076215274e-07,
+      "loss": 0.8824,
+      "step": 6702
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3084893164031563e-07,
+      "loss": 1.1295,
+      "step": 6703
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.301921260718193e-07,
+      "loss": 0.6537,
+      "step": 6704
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2953623418247694e-07,
+      "loss": 0.494,
+      "step": 6705
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.288812560979281e-07,
+      "loss": 0.6688,
+      "step": 6706
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2822719194363796e-07,
+      "loss": 0.718,
+      "step": 6707
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2757404184489351e-07,
+      "loss": 0.8394,
+      "step": 6708
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2692180592681133e-07,
+      "loss": 0.3997,
+      "step": 6709
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2627048431432764e-07,
+      "loss": 0.7571,
+      "step": 6710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2562007713220812e-07,
+      "loss": 0.8373,
+      "step": 6711
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.249705845050404e-07,
+      "loss": 0.7291,
+      "step": 6712
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2432200655723824e-07,
+      "loss": 0.5845,
+      "step": 6713
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2367434341303896e-07,
+      "loss": 0.9599,
+      "step": 6714
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2302759519650663e-07,
+      "loss": 0.7284,
+      "step": 6715
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2238176203152828e-07,
+      "loss": 0.8551,
+      "step": 6716
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.217368440418166e-07,
+      "loss": 1.0637,
+      "step": 6717
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2109284135090837e-07,
+      "loss": 0.3457,
+      "step": 6718
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.204497540821654e-07,
+      "loss": 0.7561,
+      "step": 6719
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.198075823587742e-07,
+      "loss": 0.5824,
+      "step": 6720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1916632630374579e-07,
+      "loss": 0.994,
+      "step": 6721
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1852598603991527e-07,
+      "loss": 0.7734,
+      "step": 6722
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1788656168994393e-07,
+      "loss": 0.5962,
+      "step": 6723
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.172480533763155e-07,
+      "loss": 1.0427,
+      "step": 6724
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1661046122133988e-07,
+      "loss": 0.9389,
+      "step": 6725
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.159737853471505e-07,
+      "loss": 0.5268,
+      "step": 6726
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1533802587570597e-07,
+      "loss": 0.549,
+      "step": 6727
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1470318292878768e-07,
+      "loss": 0.7793,
+      "step": 6728
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1406925662800504e-07,
+      "loss": 0.6269,
+      "step": 6729
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1343624709478815e-07,
+      "loss": 0.5446,
+      "step": 6730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1280415445039337e-07,
+      "loss": 0.8932,
+      "step": 6731
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.121729788159005e-07,
+      "loss": 1.208,
+      "step": 6732
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1154272031221446e-07,
+      "loss": 0.6681,
+      "step": 6733
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.109133790600648e-07,
+      "loss": 0.261,
+      "step": 6734
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1028495518000403e-07,
+      "loss": 1.0618,
+      "step": 6735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0965744879241078e-07,
+      "loss": 0.7483,
+      "step": 6736
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0903086001748507e-07,
+      "loss": 0.5789,
+      "step": 6737
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.084051889752553e-07,
+      "loss": 0.4915,
+      "step": 6738
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0778043578556894e-07,
+      "loss": 0.7341,
+      "step": 6739
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0715660056810305e-07,
+      "loss": 0.8971,
+      "step": 6740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0653368344235425e-07,
+      "loss": 0.6614,
+      "step": 6741
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0591168452764654e-07,
+      "loss": 0.8318,
+      "step": 6742
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.052906039431257e-07,
+      "loss": 0.4981,
+      "step": 6743
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0467044180776384e-07,
+      "loss": 0.8388,
+      "step": 6744
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.040511982403548e-07,
+      "loss": 0.657,
+      "step": 6745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0343287335951867e-07,
+      "loss": 0.7311,
+      "step": 6746
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0281546728369683e-07,
+      "loss": 0.9581,
+      "step": 6747
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0219898013115803e-07,
+      "loss": 0.6526,
+      "step": 6748
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.015834120199922e-07,
+      "loss": 0.5083,
+      "step": 6749
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0096876306811507e-07,
+      "loss": 0.9835,
+      "step": 6750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0035503339326523e-07,
+      "loss": 0.9297,
+      "step": 6751
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.997422231130053e-07,
+      "loss": 0.9147,
+      "step": 6752
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.991303323447219e-07,
+      "loss": 0.673,
+      "step": 6753
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9851936120562632e-07,
+      "loss": 0.55,
+      "step": 6754
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9790930981275213e-07,
+      "loss": 0.7865,
+      "step": 6755
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9730017828295812e-07,
+      "loss": 0.4343,
+      "step": 6756
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.966919667329259e-07,
+      "loss": 0.8463,
+      "step": 6757
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9608467527916174e-07,
+      "loss": 0.7205,
+      "step": 6758
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9547830403799373e-07,
+      "loss": 0.7396,
+      "step": 6759
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.948728531255778e-07,
+      "loss": 0.6632,
+      "step": 6760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9426832265788788e-07,
+      "loss": 0.8074,
+      "step": 6761
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9366471275072686e-07,
+      "loss": 0.7925,
+      "step": 6762
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9306202351971836e-07,
+      "loss": 1.1132,
+      "step": 6763
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9246025508031007e-07,
+      "loss": 0.699,
+      "step": 6764
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9185940754777421e-07,
+      "loss": 0.9428,
+      "step": 6765
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9125948103720537e-07,
+      "loss": 0.718,
+      "step": 6766
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9066047566352275e-07,
+      "loss": 0.8943,
+      "step": 6767
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.900623915414679e-07,
+      "loss": 0.7109,
+      "step": 6768
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8946522878560746e-07,
+      "loss": 0.8721,
+      "step": 6769
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8886898751033e-07,
+      "loss": 1.0946,
+      "step": 6770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8827366782984913e-07,
+      "loss": 0.6504,
+      "step": 6771
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8767926985820027e-07,
+      "loss": 0.9664,
+      "step": 6772
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8708579370924462e-07,
+      "loss": 0.5,
+      "step": 6773
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8649323949666343e-07,
+      "loss": 0.5521,
+      "step": 6774
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8590160733396479e-07,
+      "loss": 0.8942,
+      "step": 6775
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.853108973344775e-07,
+      "loss": 1.0127,
+      "step": 6776
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8472110961135604e-07,
+      "loss": 0.7402,
+      "step": 6777
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.841322442775756e-07,
+      "loss": 0.3994,
+      "step": 6778
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.835443014459376e-07,
+      "loss": 0.7179,
+      "step": 6779
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8295728122906365e-07,
+      "loss": 0.6895,
+      "step": 6780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.823711837394021e-07,
+      "loss": 1.0458,
+      "step": 6781
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8178600908922096e-07,
+      "loss": 0.6891,
+      "step": 6782
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.812017573906144e-07,
+      "loss": 0.8634,
+      "step": 6783
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.806184287554974e-07,
+      "loss": 0.5094,
+      "step": 6784
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.800360232956111e-07,
+      "loss": 1.1952,
+      "step": 6785
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.794545411225157e-07,
+      "loss": 0.1905,
+      "step": 6786
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.788739823475988e-07,
+      "loss": 0.8355,
+      "step": 6787
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7829434708206805e-07,
+      "loss": 0.8845,
+      "step": 6788
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.777156354369558e-07,
+      "loss": 0.5421,
+      "step": 6789
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7713784752311723e-07,
+      "loss": 0.8334,
+      "step": 6790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.765609834512294e-07,
+      "loss": 0.6053,
+      "step": 6791
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7598504333179445e-07,
+      "loss": 0.9765,
+      "step": 6792
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7541002727513578e-07,
+      "loss": 0.7278,
+      "step": 6793
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7483593539140031e-07,
+      "loss": 0.7245,
+      "step": 6794
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.742627677905584e-07,
+      "loss": 0.7483,
+      "step": 6795
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7369052458240332e-07,
+      "loss": 1.0072,
+      "step": 6796
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7311920587654962e-07,
+      "loss": 0.4456,
+      "step": 6797
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7254881178243755e-07,
+      "loss": 0.9124,
+      "step": 6798
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7197934240932801e-07,
+      "loss": 0.798,
+      "step": 6799
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.71410797866306e-07,
+      "loss": 0.6216,
+      "step": 6800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7084317826227825e-07,
+      "loss": 0.6389,
+      "step": 6801
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7027648370597617e-07,
+      "loss": 0.6703,
+      "step": 6802
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6971071430595065e-07,
+      "loss": 0.7631,
+      "step": 6803
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6914587017058005e-07,
+      "loss": 0.6741,
+      "step": 6804
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6858195140806111e-07,
+      "loss": 0.6382,
+      "step": 6805
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6801895812641577e-07,
+      "loss": 0.4314,
+      "step": 6806
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6745689043348779e-07,
+      "loss": 0.7732,
+      "step": 6807
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6689574843694433e-07,
+      "loss": 0.8566,
+      "step": 6808
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.663355322442739e-07,
+      "loss": 0.7946,
+      "step": 6809
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6577624196279007e-07,
+      "loss": 1.0852,
+      "step": 6810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.652178776996255e-07,
+      "loss": 0.7313,
+      "step": 6811
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.646604395617396e-07,
+      "loss": 0.4078,
+      "step": 6812
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.641039276559103e-07,
+      "loss": 0.2761,
+      "step": 6813
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6354834208874182e-07,
+      "loss": 0.7416,
+      "step": 6814
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6299368296665784e-07,
+      "loss": 0.8628,
+      "step": 6815
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6243995039590677e-07,
+      "loss": 0.515,
+      "step": 6816
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.618871444825587e-07,
+      "loss": 0.5356,
+      "step": 6817
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6133526533250566e-07,
+      "loss": 1.2219,
+      "step": 6818
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6078431305146304e-07,
+      "loss": 0.603,
+      "step": 6819
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6023428774496808e-07,
+      "loss": 0.8789,
+      "step": 6820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5968518951838153e-07,
+      "loss": 0.5945,
+      "step": 6821
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5913701847688422e-07,
+      "loss": 0.8723,
+      "step": 6822
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5858977472548276e-07,
+      "loss": 0.9109,
+      "step": 6823
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5804345836900215e-07,
+      "loss": 0.4111,
+      "step": 6824
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5749806951209368e-07,
+      "loss": 0.7117,
+      "step": 6825
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5695360825922824e-07,
+      "loss": 1.029,
+      "step": 6826
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.564100747147007e-07,
+      "loss": 0.6769,
+      "step": 6827
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5586746898262617e-07,
+      "loss": 1.2466,
+      "step": 6828
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5532579116694536e-07,
+      "loss": 0.7564,
+      "step": 6829
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5478504137141636e-07,
+      "loss": 0.5468,
+      "step": 6830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5424521969962525e-07,
+      "loss": 0.6896,
+      "step": 6831
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.537063262549754e-07,
+      "loss": 0.7457,
+      "step": 6832
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5316836114069532e-07,
+      "loss": 0.762,
+      "step": 6833
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.526313244598343e-07,
+      "loss": 0.5847,
+      "step": 6834
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5209521631526503e-07,
+      "loss": 0.6736,
+      "step": 6835
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5156003680968034e-07,
+      "loss": 0.6137,
+      "step": 6836
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.510257860455977e-07,
+      "loss": 0.7046,
+      "step": 6837
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.504924641253541e-07,
+      "loss": 0.4679,
+      "step": 6838
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4996007115111067e-07,
+      "loss": 0.5069,
+      "step": 6839
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4942860722484965e-07,
+      "loss": 0.521,
+      "step": 6840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4889807244837574e-07,
+      "loss": 1.1907,
+      "step": 6841
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.483684669233143e-07,
+      "loss": 0.9724,
+      "step": 6842
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4783979075111476e-07,
+      "loss": 0.7151,
+      "step": 6843
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4731204403304722e-07,
+      "loss": 0.5626,
+      "step": 6844
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4678522687020414e-07,
+      "loss": 0.497,
+      "step": 6845
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.462593393634998e-07,
+      "loss": 0.8478,
+      "step": 6846
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4573438161367027e-07,
+      "loss": 0.7987,
+      "step": 6847
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4521035372127402e-07,
+      "loss": 0.8802,
+      "step": 6848
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4468725578669075e-07,
+      "loss": 0.7664,
+      "step": 6849
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4416508791012252e-07,
+      "loss": 0.5141,
+      "step": 6850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4364385019159266e-07,
+      "loss": 0.8568,
+      "step": 6851
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.431235427309474e-07,
+      "loss": 0.7475,
+      "step": 6852
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4260416562785374e-07,
+      "loss": 0.9685,
+      "step": 6853
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4208571898180034e-07,
+      "loss": 0.6165,
+      "step": 6854
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4156820289209894e-07,
+      "loss": 0.6483,
+      "step": 6855
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4105161745788187e-07,
+      "loss": 0.8271,
+      "step": 6856
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4053596277810279e-07,
+      "loss": 0.7139,
+      "step": 6857
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4002123895153873e-07,
+      "loss": 0.524,
+      "step": 6858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3950744607678646e-07,
+      "loss": 0.8176,
+      "step": 6859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.389945842522672e-07,
+      "loss": 1.1602,
+      "step": 6860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3848265357622015e-07,
+      "loss": 1.1145,
+      "step": 6861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379716541467091e-07,
+      "loss": 1.072,
+      "step": 6862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3746158606161797e-07,
+      "loss": 0.8114,
+      "step": 6863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3695244941865306e-07,
+      "loss": 0.5431,
+      "step": 6864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.364442443153413e-07,
+      "loss": 0.7273,
+      "step": 6865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.359369708490327e-07,
+      "loss": 0.8646,
+      "step": 6866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3543062911689663e-07,
+      "loss": 0.3237,
+      "step": 6867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3492521921592672e-07,
+      "loss": 1.0029,
+      "step": 6868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3442074124293548e-07,
+      "loss": 0.8497,
+      "step": 6869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3391719529455837e-07,
+      "loss": 0.7215,
+      "step": 6870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.334145814672516e-07,
+      "loss": 0.8962,
+      "step": 6871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3291289985729427e-07,
+      "loss": 0.6975,
+      "step": 6872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3241215056078504e-07,
+      "loss": 1.0048,
+      "step": 6873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3191233367364497e-07,
+      "loss": 0.8793,
+      "step": 6874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3141344929161582e-07,
+      "loss": 0.7683,
+      "step": 6875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3091549751026224e-07,
+      "loss": 0.6221,
+      "step": 6876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.30418478424969e-07,
+      "loss": 0.6967,
+      "step": 6877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2992239213094172e-07,
+      "loss": 0.575,
+      "step": 6878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2942723872320872e-07,
+      "loss": 0.788,
+      "step": 6879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2893301829661865e-07,
+      "loss": 0.5438,
+      "step": 6880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.284397309458424e-07,
+      "loss": 0.8837,
+      "step": 6881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2794737676536993e-07,
+      "loss": 0.6635,
+      "step": 6882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2745595584951586e-07,
+      "loss": 1.0071,
+      "step": 6883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.269654682924132e-07,
+      "loss": 0.8437,
+      "step": 6884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.264759141880173e-07,
+      "loss": 0.6636,
+      "step": 6885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2598729363010432e-07,
+      "loss": 0.7866,
+      "step": 6886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2549960671227268e-07,
+      "loss": 0.6855,
+      "step": 6887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.250128535279399e-07,
+      "loss": 1.1946,
+      "step": 6888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2452703417034696e-07,
+      "loss": 0.9051,
+      "step": 6889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2404214873255382e-07,
+      "loss": 0.4268,
+      "step": 6890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2355819730744345e-07,
+      "loss": 0.4698,
+      "step": 6891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2307517998771834e-07,
+      "loss": 0.7996,
+      "step": 6892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2259309686590338e-07,
+      "loss": 0.8724,
+      "step": 6893
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.22111948034343e-07,
+      "loss": 0.6907,
+      "step": 6894
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.216317335852052e-07,
+      "loss": 0.7464,
+      "step": 6895
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2115245361047467e-07,
+      "loss": 0.4773,
+      "step": 6896
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2067410820196247e-07,
+      "loss": 0.692,
+      "step": 6897
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.201966974512958e-07,
+      "loss": 0.7271,
+      "step": 6898
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1972022144992656e-07,
+      "loss": 0.8893,
+      "step": 6899
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1924468028912505e-07,
+      "loss": 0.404,
+      "step": 6900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1877007405998286e-07,
+      "loss": 0.5675,
+      "step": 6901
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1829640285341504e-07,
+      "loss": 0.4503,
+      "step": 6902
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1782366676015344e-07,
+      "loss": 0.9832,
+      "step": 6903
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1735186587075398e-07,
+      "loss": 1.1502,
+      "step": 6904
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1688100027559212e-07,
+      "loss": 0.4192,
+      "step": 6905
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1641107006486463e-07,
+      "loss": 0.8596,
+      "step": 6906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1594207532858781e-07,
+      "loss": 1.0372,
+      "step": 6907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1547401615660147e-07,
+      "loss": 0.7839,
+      "step": 6908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1500689263856335e-07,
+      "loss": 0.7216,
+      "step": 6909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1454070486395352e-07,
+      "loss": 0.3304,
+      "step": 6910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1407545292207222e-07,
+      "loss": 1.1957,
+      "step": 6911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1361113690204095e-07,
+      "loss": 0.9484,
+      "step": 6912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1314775689280133e-07,
+      "loss": 1.0611,
+      "step": 6913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.126853129831168e-07,
+      "loss": 0.9219,
+      "step": 6914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1222380526156929e-07,
+      "loss": 0.9625,
+      "step": 6915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1176323381656417e-07,
+      "loss": 0.8138,
+      "step": 6916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1130359873632479e-07,
+      "loss": 0.9788,
+      "step": 6917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1084490010889737e-07,
+      "loss": 0.8025,
+      "step": 6918
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1038713802214718e-07,
+      "loss": 0.5258,
+      "step": 6919
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0993031256376129e-07,
+      "loss": 1.0332,
+      "step": 6920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0947442382124584e-07,
+      "loss": 0.9292,
+      "step": 6921
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0901947188192985e-07,
+      "loss": 0.5143,
+      "step": 6922
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0856545683296027e-07,
+      "loss": 0.7596,
+      "step": 6923
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0811237876130698e-07,
+      "loss": 0.6536,
+      "step": 6924
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0766023775375778e-07,
+      "loss": 0.784,
+      "step": 6925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0720903389692395e-07,
+      "loss": 0.8518,
+      "step": 6926
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0675876727723411e-07,
+      "loss": 0.7748,
+      "step": 6927
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0630943798093984e-07,
+      "loss": 0.8771,
+      "step": 6928
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.058610460941123e-07,
+      "loss": 1.0598,
+      "step": 6929
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0541359170264331e-07,
+      "loss": 0.5617,
+      "step": 6930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0496707489224433e-07,
+      "loss": 0.6208,
+      "step": 6931
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0452149574844749e-07,
+      "loss": 0.808,
+      "step": 6932
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0407685435660675e-07,
+      "loss": 0.5959,
+      "step": 6933
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0363315080189395e-07,
+      "loss": 0.8482,
+      "step": 6934
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.031903851693039e-07,
+      "loss": 0.6903,
+      "step": 6935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0274855754364931e-07,
+      "loss": 0.9131,
+      "step": 6936
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0230766800956471e-07,
+      "loss": 1.0793,
+      "step": 6937
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0186771665150475e-07,
+      "loss": 0.857,
+      "step": 6938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0142870355374424e-07,
+      "loss": 0.7578,
+      "step": 6939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0099062880037757e-07,
+      "loss": 0.6899,
+      "step": 6940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0055349247532098e-07,
+      "loss": 0.6991,
+      "step": 6941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0011729466230913e-07,
+      "loss": 0.6376,
+      "step": 6942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.968203544489852e-08,
+      "loss": 0.5777,
+      "step": 6943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.924771490646413e-08,
+      "loss": 1.0107,
+      "step": 6944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.881433313020328e-08,
+      "loss": 0.7286,
+      "step": 6945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.83818901991307e-08,
+      "loss": 0.6504,
+      "step": 6946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.795038619608511e-08,
+      "loss": 0.8605,
+      "step": 6947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.751982120372094e-08,
+      "loss": 0.8712,
+      "step": 6948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.709019530451614e-08,
+      "loss": 0.6088,
+      "step": 6949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.666150858076762e-08,
+      "loss": 1.1543,
+      "step": 6950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.623376111459193e-08,
+      "loss": 0.8498,
+      "step": 6951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.580695298792575e-08,
+      "loss": 0.6762,
+      "step": 6952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.538108428252702e-08,
+      "loss": 1.2539,
+      "step": 6953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.49561550799727e-08,
+      "loss": 0.8216,
+      "step": 6954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.453216546165933e-08,
+      "loss": 0.6816,
+      "step": 6955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.410911550880474e-08,
+      "loss": 0.7712,
+      "step": 6956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.368700530244578e-08,
+      "loss": 0.797,
+      "step": 6957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.326583492344055e-08,
+      "loss": 0.8785,
+      "step": 6958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.284560445246504e-08,
+      "loss": 0.7556,
+      "step": 6959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.242631397001711e-08,
+      "loss": 0.933,
+      "step": 6960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.200796355641361e-08,
+      "loss": 1.0119,
+      "step": 6961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.159055329179156e-08,
+      "loss": 0.949,
+      "step": 6962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.117408325610699e-08,
+      "loss": 0.7256,
+      "step": 6963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.075855352913831e-08,
+      "loss": 0.497,
+      "step": 6964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.03439641904813e-08,
+      "loss": 0.8083,
+      "step": 6965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.993031531955243e-08,
+      "loss": 0.7761,
+      "step": 6966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.951760699558831e-08,
+      "loss": 0.4594,
+      "step": 6967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.910583929764516e-08,
+      "loss": 0.7849,
+      "step": 6968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.869501230459876e-08,
+      "loss": 0.7732,
+      "step": 6969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.828512609514562e-08,
+      "loss": 0.4324,
+      "step": 6970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.787618074780068e-08,
+      "loss": 0.4419,
+      "step": 6971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.746817634089966e-08,
+      "loss": 0.9562,
+      "step": 6972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.706111295259723e-08,
+      "loss": 0.4837,
+      "step": 6973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.665499066086935e-08,
+      "loss": 0.9195,
+      "step": 6974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.624980954350936e-08,
+      "loss": 0.8178,
+      "step": 6975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.584556967813296e-08,
+      "loss": 0.4115,
+      "step": 6976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.544227114217262e-08,
+      "loss": 0.778,
+      "step": 6977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.503991401288381e-08,
+      "loss": 0.882,
+      "step": 6978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.463849836733873e-08,
+      "loss": 0.5747,
+      "step": 6979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.423802428243089e-08,
+      "loss": 0.9147,
+      "step": 6980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.38384918348728e-08,
+      "loss": 0.804,
+      "step": 6981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.34399011011966e-08,
+      "loss": 0.4297,
+      "step": 6982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.304225215775508e-08,
+      "loss": 0.8599,
+      "step": 6983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.264554508071843e-08,
+      "loss": 0.7689,
+      "step": 6984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.224977994607919e-08,
+      "loss": 0.8554,
+      "step": 6985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.185495682964672e-08,
+      "loss": 0.4195,
+      "step": 6986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.146107580705275e-08,
+      "loss": 1.0328,
+      "step": 6987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.106813695374527e-08,
+      "loss": 0.6672,
+      "step": 6988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.067614034499516e-08,
+      "loss": 0.885,
+      "step": 6989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.028508605589013e-08,
+      "loss": 0.526,
+      "step": 6990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.989497416133918e-08,
+      "loss": 0.6171,
+      "step": 6991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.950580473606973e-08,
+      "loss": 0.5892,
+      "step": 6992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.911757785462882e-08,
+      "loss": 0.6351,
+      "step": 6993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.873029359138362e-08,
+      "loss": 0.7473,
+      "step": 6994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.834395202051981e-08,
+      "loss": 0.6546,
+      "step": 6995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.795855321604317e-08,
+      "loss": 0.7461,
+      "step": 6996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.757409725177856e-08,
+      "loss": 0.6931,
+      "step": 6997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.719058420137038e-08,
+      "loss": 0.7775,
+      "step": 6998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.680801413828265e-08,
+      "loss": 0.6611,
+      "step": 6999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.642638713579786e-08,
+      "loss": 0.6433,
+      "step": 7000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.604570326701921e-08,
+      "loss": 0.7052,
+      "step": 7001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.566596260486725e-08,
+      "loss": 0.6987,
+      "step": 7002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.52871652220849e-08,
+      "loss": 0.7927,
+      "step": 7003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.490931119123079e-08,
+      "loss": 0.8525,
+      "step": 7004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.453240058468591e-08,
+      "loss": 0.5743,
+      "step": 7005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.415643347464862e-08,
+      "loss": 0.7577,
+      "step": 7006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.378140993313743e-08,
+      "loss": 0.5609,
+      "step": 7007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.34073300319893e-08,
+      "loss": 0.4781,
+      "step": 7008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.303419384286248e-08,
+      "loss": 0.7712,
+      "step": 7009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.26620014372309e-08,
+      "loss": 1.0876,
+      "step": 7010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.229075288639143e-08,
+      "loss": 1.0114,
+      "step": 7011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.192044826145772e-08,
+      "loss": 0.5615,
+      "step": 7012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.155108763336361e-08,
+      "loss": 0.6777,
+      "step": 7013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.118267107286191e-08,
+      "loss": 0.926,
+      "step": 7014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.0815198650524e-08,
+      "loss": 0.7423,
+      "step": 7015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.044867043674187e-08,
+      "loss": 0.6194,
+      "step": 7016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.00830865017249e-08,
+      "loss": 0.5196,
+      "step": 7017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.971844691550266e-08,
+      "loss": 0.9814,
+      "step": 7018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.935475174792372e-08,
+      "loss": 0.8131,
+      "step": 7019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.899200106865622e-08,
+      "loss": 0.9656,
+      "step": 7020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.863019494718515e-08,
+      "loss": 0.824,
+      "step": 7021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.826933345281784e-08,
+      "loss": 0.5922,
+      "step": 7022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.790941665467731e-08,
+      "loss": 0.8085,
+      "step": 7023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.755044462170956e-08,
+      "loss": 0.7289,
+      "step": 7024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.719241742267513e-08,
+      "loss": 0.8186,
+      "step": 7025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.683533512615747e-08,
+      "loss": 0.7799,
+      "step": 7026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.647919780055578e-08,
+      "loss": 0.6669,
+      "step": 7027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.61240055140916e-08,
+      "loss": 0.7467,
+      "step": 7028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.57697583348027e-08,
+      "loss": 0.6349,
+      "step": 7029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.54164563305465e-08,
+      "loss": 0.7079,
+      "step": 7030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.50640995690005e-08,
+      "loss": 0.8735,
+      "step": 7031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.471268811765962e-08,
+      "loss": 0.6302,
+      "step": 7032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.436222204383891e-08,
+      "loss": 0.6752,
+      "step": 7033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.401270141467131e-08,
+      "loss": 0.5507,
+      "step": 7034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.366412629710938e-08,
+      "loss": 0.8841,
+      "step": 7035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.331649675792473e-08,
+      "loss": 0.9774,
+      "step": 7036
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.29698128637063e-08,
+      "loss": 0.6676,
+      "step": 7037
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.262407468086428e-08,
+      "loss": 0.7983,
+      "step": 7038
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.227928227562574e-08,
+      "loss": 0.8697,
+      "step": 7039
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.193543571403727e-08,
+      "loss": 0.7337,
+      "step": 7040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.159253506196505e-08,
+      "loss": 0.4738,
+      "step": 7041
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.125058038509213e-08,
+      "loss": 0.774,
+      "step": 7042
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.09095717489222e-08,
+      "loss": 0.2697,
+      "step": 7043
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.056950921877746e-08,
+      "loss": 0.6944,
+      "step": 7044
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.023039285979803e-08,
+      "loss": 0.9184,
+      "step": 7045
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.98922227369425e-08,
+      "loss": 0.7848,
+      "step": 7046
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9554998914990724e-08,
+      "loss": 0.7669,
+      "step": 7047
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.921872145853769e-08,
+      "loss": 0.9468,
+      "step": 7048
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.88833904320002e-08,
+      "loss": 0.7651,
+      "step": 7049
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.854900589961188e-08,
+      "loss": 0.9506,
+      "step": 7050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8215567925425933e-08,
+      "loss": 0.4056,
+      "step": 7051
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.788307657331349e-08,
+      "loss": 0.9374,
+      "step": 7052
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7551531906965826e-08,
+      "loss": 0.473,
+      "step": 7053
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.722093398989104e-08,
+      "loss": 0.6713,
+      "step": 7054
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.689128288541679e-08,
+      "loss": 0.4874,
+      "step": 7055
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.656257865668924e-08,
+      "loss": 1.2131,
+      "step": 7056
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.623482136667413e-08,
+      "loss": 0.6642,
+      "step": 7057
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.590801107815347e-08,
+      "loss": 0.584,
+      "step": 7058
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.558214785373051e-08,
+      "loss": 0.7427,
+      "step": 7059
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5257231755825314e-08,
+      "loss": 0.5715,
+      "step": 7060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.493326284667699e-08,
+      "loss": 0.8237,
+      "step": 7061
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.461024118834368e-08,
+      "loss": 0.9155,
+      "step": 7062
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.428816684270144e-08,
+      "loss": 0.466,
+      "step": 7063
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.396703987144536e-08,
+      "loss": 0.4632,
+      "step": 7064
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3646860336089014e-08,
+      "loss": 0.5148,
+      "step": 7065
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.332762829796334e-08,
+      "loss": 0.716,
+      "step": 7066
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3009343818219985e-08,
+      "loss": 0.8457,
+      "step": 7067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.269200695782684e-08,
+      "loss": 0.7235,
+      "step": 7068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.237561777757194e-08,
+      "loss": 0.6153,
+      "step": 7069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.206017633806071e-08,
+      "loss": 0.6224,
+      "step": 7070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.174568269971758e-08,
+      "loss": 0.4807,
+      "step": 7071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.143213692278548e-08,
+      "loss": 0.6488,
+      "step": 7072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1119539067325255e-08,
+      "loss": 0.9685,
+      "step": 7073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.080788919321622e-08,
+      "loss": 0.9891,
+      "step": 7074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.049718736015729e-08,
+      "loss": 1.0397,
+      "step": 7075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0187433627664206e-08,
+      "loss": 0.7425,
+      "step": 7076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.987862805507226e-08,
+      "loss": 1.0686,
+      "step": 7077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.95707707015336e-08,
+      "loss": 0.9317,
+      "step": 7078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9263861626021037e-08,
+      "loss": 0.8765,
+      "step": 7079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.895790088732366e-08,
+      "loss": 0.5407,
+      "step": 7080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.865288854404959e-08,
+      "loss": 1.0433,
+      "step": 7081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.834882465462598e-08,
+      "loss": 0.3262,
+      "step": 7082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.804570927729735e-08,
+      "loss": 0.677,
+      "step": 7083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.774354247012669e-08,
+      "loss": 0.81,
+      "step": 7084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7442324290996044e-08,
+      "loss": 0.3302,
+      "step": 7085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.714205479760425e-08,
+      "loss": 0.8585,
+      "step": 7086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6842734047470306e-08,
+      "loss": 1.0173,
+      "step": 7087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6544362097930005e-08,
+      "loss": 0.7719,
+      "step": 7088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6246939006137635e-08,
+      "loss": 1.0838,
+      "step": 7089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.595046482906651e-08,
+      "loss": 0.4586,
+      "step": 7090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5654939623507866e-08,
+      "loss": 0.8465,
+      "step": 7091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.536036344607031e-08,
+      "loss": 0.8787,
+      "step": 7092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.506673635318204e-08,
+      "loss": 0.8094,
+      "step": 7093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4774058401087504e-08,
+      "loss": 1.177,
+      "step": 7094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4482329645851865e-08,
+      "loss": 0.8535,
+      "step": 7095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4191550143355964e-08,
+      "loss": 0.7704,
+      "step": 7096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.390171994930081e-08,
+      "loss": 0.5061,
+      "step": 7097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.361283911920477e-08,
+      "loss": 0.7803,
+      "step": 7098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3324907708403584e-08,
+      "loss": 0.803,
+      "step": 7099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.303792577205257e-08,
+      "loss": 0.5789,
+      "step": 7100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.275189336512442e-08,
+      "loss": 0.741,
+      "step": 7101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2466810542409755e-08,
+      "loss": 0.6702,
+      "step": 7102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.218267735851767e-08,
+      "loss": 0.5752,
+      "step": 7103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.189949386787462e-08,
+      "loss": 0.6608,
+      "step": 7104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1617260124726665e-08,
+      "loss": 0.8034,
+      "step": 7105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.133597618313612e-08,
+      "loss": 0.7522,
+      "step": 7106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1055642096985424e-08,
+      "loss": 0.5193,
+      "step": 7107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.077625791997275e-08,
+      "loss": 0.8067,
+      "step": 7108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0497823705615836e-08,
+      "loss": 0.6966,
+      "step": 7109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0220339507249794e-08,
+      "loss": 0.559,
+      "step": 7110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.994380537802933e-08,
+      "loss": 0.3944,
+      "step": 7111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9668221370923744e-08,
+      "loss": 0.896,
+      "step": 7112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9393587538724686e-08,
+      "loss": 0.7993,
+      "step": 7113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.911990393403786e-08,
+      "loss": 0.5472,
+      "step": 7114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.884717060928911e-08,
+      "loss": 0.8076,
+      "step": 7115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8575387616722195e-08,
+      "loss": 0.9194,
+      "step": 7116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.830455500839825e-08,
+      "loss": 0.7559,
+      "step": 7117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8034672836195774e-08,
+      "loss": 0.859,
+      "step": 7118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.776574115181342e-08,
+      "loss": 0.8901,
+      "step": 7119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.749776000676497e-08,
+      "loss": 1.0816,
+      "step": 7120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7230729452384376e-08,
+      "loss": 0.9822,
+      "step": 7121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6964649539822373e-08,
+      "loss": 0.7929,
+      "step": 7122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.669952032004709e-08,
+      "loss": 0.6892,
+      "step": 7123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.643534184384678e-08,
+      "loss": 0.5485,
+      "step": 7124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.617211416182431e-08,
+      "loss": 0.9053,
+      "step": 7125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5909837324403785e-08,
+      "loss": 0.3374,
+      "step": 7126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5648511381824457e-08,
+      "loss": 0.785,
+      "step": 7127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5388136384144625e-08,
+      "loss": 0.5894,
+      "step": 7128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.512871238124105e-08,
+      "loss": 0.5293,
+      "step": 7129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4870239422807316e-08,
+      "loss": 0.6881,
+      "step": 7130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.461271755835493e-08,
+      "loss": 0.9917,
+      "step": 7131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4356146837213334e-08,
+      "loss": 0.6598,
+      "step": 7132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.410052730853042e-08,
+      "loss": 0.7303,
+      "step": 7133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.384585902127091e-08,
+      "loss": 0.6783,
+      "step": 7134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.35921420242169e-08,
+      "loss": 0.6896,
+      "step": 7135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.333937636597118e-08,
+      "loss": 1.074,
+      "step": 7136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.308756209495001e-08,
+      "loss": 0.801,
+      "step": 7137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.283669925939093e-08,
+      "loss": 0.8185,
+      "step": 7138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2586787907347153e-08,
+      "loss": 0.5053,
+      "step": 7139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2337828086691506e-08,
+      "loss": 0.8061,
+      "step": 7140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2089819845111946e-08,
+      "loss": 0.7226,
+      "step": 7141
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1842763230116594e-08,
+      "loss": 0.6352,
+      "step": 7142
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1596658289029246e-08,
+      "loss": 0.8118,
+      "step": 7143
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1351505068994426e-08,
+      "loss": 0.6929,
+      "step": 7144
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.110730361697012e-08,
+      "loss": 0.9511,
+      "step": 7145
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.086405397973613e-08,
+      "loss": 0.4258,
+      "step": 7146
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0621756203886855e-08,
+      "loss": 0.6919,
+      "step": 7147
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.038041033583627e-08,
+      "loss": 0.9274,
+      "step": 7148
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0140016421814635e-08,
+      "loss": 0.7303,
+      "step": 7149
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.99005745078712e-08,
+      "loss": 0.5454,
+      "step": 7150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.966208463987208e-08,
+      "loss": 0.6609,
+      "step": 7151
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9424546863500715e-08,
+      "loss": 0.4991,
+      "step": 7152
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9187961224259044e-08,
+      "loss": 0.9171,
+      "step": 7153
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8952327767465814e-08,
+      "loss": 1.1197,
+      "step": 7154
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.871764653825826e-08,
+      "loss": 0.8289,
+      "step": 7155
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.848391758158986e-08,
+      "loss": 0.9258,
+      "step": 7156
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8251140942233135e-08,
+      "loss": 0.8539,
+      "step": 7157
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8019316664776864e-08,
+      "loss": 0.8224,
+      "step": 7158
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.778844479362941e-08,
+      "loss": 0.7797,
+      "step": 7159
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7558525373013735e-08,
+      "loss": 1.1048,
+      "step": 7160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7329558446972935e-08,
+      "loss": 0.9277,
+      "step": 7161
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7101544059366935e-08,
+      "loss": 0.4268,
+      "step": 7162
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6874482253872458e-08,
+      "loss": 0.7085,
+      "step": 7163
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6648373073984157e-08,
+      "loss": 0.6516,
+      "step": 7164
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6423216563014608e-08,
+      "loss": 1.181,
+      "step": 7165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.61990127640932e-08,
+      "loss": 0.733,
+      "step": 7166
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5975761720167804e-08,
+      "loss": 0.425,
+      "step": 7167
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5753463474003093e-08,
+      "loss": 0.7828,
+      "step": 7168
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.553211806818112e-08,
+      "loss": 0.6806,
+      "step": 7169
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5311725545101307e-08,
+      "loss": 0.9417,
+      "step": 7170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.50922859469821e-08,
+      "loss": 0.645,
+      "step": 7171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4873799315856538e-08,
+      "loss": 0.4725,
+      "step": 7172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4656265693578375e-08,
+      "loss": 0.8315,
+      "step": 7173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4439685121815383e-08,
+      "loss": 0.847,
+      "step": 7174
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4224057642056596e-08,
+      "loss": 0.9274,
+      "step": 7175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4009383295605092e-08,
+      "loss": 0.9266,
+      "step": 7176
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3795662123582975e-08,
+      "loss": 0.8323,
+      "step": 7177
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.358289416693027e-08,
+      "loss": 0.7499,
+      "step": 7178
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3371079466402715e-08,
+      "loss": 0.5564,
+      "step": 7179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.316021806257507e-08,
+      "loss": 0.5034,
+      "step": 7180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.295030999583836e-08,
+      "loss": 0.8532,
+      "step": 7181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.274135530640209e-08,
+      "loss": 0.7239,
+      "step": 7182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2533354034291465e-08,
+      "loss": 0.6738,
+      "step": 7183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2326306219351278e-08,
+      "loss": 0.5049,
+      "step": 7184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2120211901242027e-08,
+      "loss": 1.1231,
+      "step": 7185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1915071119442132e-08,
+      "loss": 0.7685,
+      "step": 7186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1710883913246827e-08,
+      "loss": 1.0771,
+      "step": 7187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1507650321769825e-08,
+      "loss": 0.6386,
+      "step": 7188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1305370383940537e-08,
+      "loss": 0.4749,
+      "step": 7189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1104044138507975e-08,
+      "loss": 0.5356,
+      "step": 7190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0903671624035726e-08,
+      "loss": 0.8107,
+      "step": 7191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0704252878906985e-08,
+      "loss": 0.5977,
+      "step": 7192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0505787941321187e-08,
+      "loss": 0.5109,
+      "step": 7193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.030827684929515e-08,
+      "loss": 0.9178,
+      "step": 7194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0111719640663052e-08,
+      "loss": 0.6282,
+      "step": 7195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.991611635307644e-08,
+      "loss": 0.604,
+      "step": 7196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.972146702400368e-08,
+      "loss": 0.7486,
+      "step": 7197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9527771690731613e-08,
+      "loss": 0.5297,
+      "step": 7198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9335030390362218e-08,
+      "loss": 0.9035,
+      "step": 7199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.914324315981708e-08,
+      "loss": 0.8727,
+      "step": 7200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.895241003583348e-08,
+      "loss": 0.7949,
+      "step": 7201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8762531054966615e-08,
+      "loss": 0.7431,
+      "step": 7202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8573606253588504e-08,
+      "loss": 0.5465,
+      "step": 7203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8385635667888536e-08,
+      "loss": 0.4333,
+      "step": 7204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.819861933387346e-08,
+      "loss": 0.7149,
+      "step": 7205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8012557287367394e-08,
+      "loss": 0.894,
+      "step": 7206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.782744956401128e-08,
+      "loss": 0.354,
+      "step": 7207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.764329619926286e-08,
+      "loss": 0.6914,
+      "step": 7208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7460097228398364e-08,
+      "loss": 1.2401,
+      "step": 7209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7277852686509722e-08,
+      "loss": 0.6783,
+      "step": 7210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7096562608507338e-08,
+      "loss": 0.5681,
+      "step": 7211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6916227029117327e-08,
+      "loss": 0.9995,
+      "step": 7212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.673684598288483e-08,
+      "loss": 0.7908,
+      "step": 7213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.65584195041707e-08,
+      "loss": 0.708,
+      "step": 7214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6380947627153143e-08,
+      "loss": 0.4843,
+      "step": 7215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.620443038582775e-08,
+      "loss": 1.1258,
+      "step": 7216
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.602886781400803e-08,
+      "loss": 0.7356,
+      "step": 7217
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5854259945322638e-08,
+      "loss": 1.3168,
+      "step": 7218
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5680606813219256e-08,
+      "loss": 0.8534,
+      "step": 7219
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.550790845096184e-08,
+      "loss": 0.5691,
+      "step": 7220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.53361648916317e-08,
+      "loss": 0.4292,
+      "step": 7221
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.516537616812641e-08,
+      "loss": 0.8501,
+      "step": 7222
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.499554231316258e-08,
+      "loss": 0.8043,
+      "step": 7223
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.482666335927141e-08,
+      "loss": 1.096,
+      "step": 7224
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4658739338803129e-08,
+      "loss": 0.9369,
+      "step": 7225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4491770283924788e-08,
+      "loss": 0.7227,
+      "step": 7226
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4325756226619136e-08,
+      "loss": 0.8498,
+      "step": 7227
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4160697198687401e-08,
+      "loss": 0.7237,
+      "step": 7228
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3996593231748178e-08,
+      "loss": 0.575,
+      "step": 7229
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.383344435723466e-08,
+      "loss": 0.6824,
+      "step": 7230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3671250606400177e-08,
+      "loss": 0.5849,
+      "step": 7231
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3510012010313212e-08,
+      "loss": 0.3784,
+      "step": 7232
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3349728599860168e-08,
+      "loss": 0.5752,
+      "step": 7233
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3190400405743153e-08,
+      "loss": 0.5377,
+      "step": 7234
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3032027458482754e-08,
+      "loss": 0.8723,
+      "step": 7235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2874609788416925e-08,
+      "loss": 0.8564,
+      "step": 7236
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2718147425698213e-08,
+      "loss": 0.3925,
+      "step": 7237
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2562640400298754e-08,
+      "loss": 0.75,
+      "step": 7238
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2408088742006385e-08,
+      "loss": 0.5063,
+      "step": 7239
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2254492480426316e-08,
+      "loss": 0.5248,
+      "step": 7240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2101851644980011e-08,
+      "loss": 0.86,
+      "step": 7241
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1950166264907415e-08,
+      "loss": 0.8988,
+      "step": 7242
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1799436369264172e-08,
+      "loss": 0.7769,
+      "step": 7243
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.16496619869233e-08,
+      "loss": 1.0368,
+      "step": 7244
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1500843146574626e-08,
+      "loss": 0.8196,
+      "step": 7245
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.13529798767259e-08,
+      "loss": 0.8921,
+      "step": 7246
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1206072205700025e-08,
+      "loss": 0.7821,
+      "step": 7247
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1060120161637821e-08,
+      "loss": 1.032,
+      "step": 7248
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0915123772498038e-08,
+      "loss": 0.9441,
+      "step": 7249
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0771083066054567e-08,
+      "loss": 0.7138,
+      "step": 7250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0627998069899781e-08,
+      "loss": 0.7789,
+      "step": 7251
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0485868811441757e-08,
+      "loss": 0.4882,
+      "step": 7252
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0344695317905939e-08,
+      "loss": 0.7533,
+      "step": 7253
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.020447761633514e-08,
+      "loss": 1.016,
+      "step": 7254
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0065215733588984e-08,
+      "loss": 0.4413,
+      "step": 7255
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.9269096963428e-09,
+      "loss": 0.7351,
+      "step": 7256
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.789559531090398e-09,
+      "loss": 0.8876,
+      "step": 7257
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.653165264142394e-09,
+      "loss": 0.6467,
+      "step": 7258
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.51772692162456e-09,
+      "loss": 0.7014,
+      "step": 7259
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.38324452948114e-09,
+      "loss": 0.5281,
+      "step": 7260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.249718113473194e-09,
+      "loss": 0.7129,
+      "step": 7261
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.117147699178597e-09,
+      "loss": 0.8627,
+      "step": 7262
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.985533311990925e-09,
+      "loss": 0.5338,
+      "step": 7263
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.85487497712223e-09,
+      "loss": 0.6097,
+      "step": 7264
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.725172719600272e-09,
+      "loss": 0.3814,
+      "step": 7265
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.596426564270733e-09,
+      "loss": 0.6477,
+      "step": 7266
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.468636535794995e-09,
+      "loss": 0.7846,
+      "step": 7267
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.341802658652364e-09,
+      "loss": 1.0612,
+      "step": 7268
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.215924957137855e-09,
+      "loss": 1.0897,
+      "step": 7269
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.091003455364398e-09,
+      "loss": 1.008,
+      "step": 7270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.967038177260633e-09,
+      "loss": 1.1053,
+      "step": 7271
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.844029146573673e-09,
+      "loss": 0.6127,
+      "step": 7272
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.721976386865782e-09,
+      "loss": 0.9792,
+      "step": 7273
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.600879921517146e-09,
+      "loss": 0.8702,
+      "step": 7274
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.48073977372421e-09,
+      "loss": 0.7235,
+      "step": 7275
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.361555966499678e-09,
+      "loss": 0.5002,
+      "step": 7276
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.243328522675286e-09,
+      "loss": 0.8724,
+      "step": 7277
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.126057464896807e-09,
+      "loss": 0.6996,
+      "step": 7278
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.009742815628495e-09,
+      "loss": 0.4206,
+      "step": 7279
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.8943845971508605e-09,
+      "loss": 0.6331,
+      "step": 7280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.779982831561227e-09,
+      "loss": 0.8326,
+      "step": 7281
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.666537540774287e-09,
+      "loss": 1.0202,
+      "step": 7282
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.554048746520436e-09,
+      "loss": 1.0738,
+      "step": 7283
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.442516470347993e-09,
+      "loss": 0.8415,
+      "step": 7284
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.331940733620978e-09,
+      "loss": 0.847,
+      "step": 7285
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.2223215575207825e-09,
+      "loss": 0.5946,
+      "step": 7286
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.11365896304561e-09,
+      "loss": 0.9055,
+      "step": 7287
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.005952971010476e-09,
+      "loss": 0.4166,
+      "step": 7288
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.899203602046655e-09,
+      "loss": 0.7151,
+      "step": 7289
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.793410876602235e-09,
+      "loss": 0.908,
+      "step": 7290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.688574814943226e-09,
+      "loss": 0.9352,
+      "step": 7291
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.584695437150234e-09,
+      "loss": 0.836,
+      "step": 7292
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.481772763122895e-09,
+      "loss": 1.1098,
+      "step": 7293
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.379806812576549e-09,
+      "loss": 0.7793,
+      "step": 7294
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.27879760504224e-09,
+      "loss": 0.6304,
+      "step": 7295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.178745159870047e-09,
+      "loss": 0.8296,
+      "step": 7296
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0796494962240816e-09,
+      "loss": 0.6428,
+      "step": 7297
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.981510633088049e-09,
+      "loss": 0.8406,
+      "step": 7298
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.884328589259691e-09,
+      "loss": 0.787,
+      "step": 7299
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.788103383355225e-09,
+      "loss": 0.486,
+      "step": 7300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.69283503380713e-09,
+      "loss": 0.8928,
+      "step": 7301
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.598523558864143e-09,
+      "loss": 1.0662,
+      "step": 7302
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.505168976592922e-09,
+      "loss": 0.6256,
+      "step": 7303
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.412771304874719e-09,
+      "loss": 0.6674,
+      "step": 7304
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.32133056140982e-09,
+      "loss": 0.7621,
+      "step": 7305
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2308467637131035e-09,
+      "loss": 0.8368,
+      "step": 7306
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.141319929118481e-09,
+      "loss": 0.7096,
+      "step": 7307
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0527500747739035e-09,
+      "loss": 0.9385,
+      "step": 7308
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.965137217646353e-09,
+      "loss": 0.846,
+      "step": 7309
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.878481374517406e-09,
+      "loss": 0.7765,
+      "step": 7310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.792782561987673e-09,
+      "loss": 0.7717,
+      "step": 7311
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7080407964729116e-09,
+      "loss": 1.2313,
+      "step": 7312
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.624256094204581e-09,
+      "loss": 0.5018,
+      "step": 7313
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5414284712337323e-09,
+      "loss": 0.7522,
+      "step": 7314
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.459557943424896e-09,
+      "loss": 0.8999,
+      "step": 7315
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3786445264621935e-09,
+      "loss": 0.7554,
+      "step": 7316
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2986882358437834e-09,
+      "loss": 1.0918,
+      "step": 7317
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.219689086885747e-09,
+      "loss": 1.0036,
+      "step": 7318
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.141647094721534e-09,
+      "loss": 0.5073,
+      "step": 7319
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0645622742997427e-09,
+      "loss": 0.6442,
+      "step": 7320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9884346403868945e-09,
+      "loss": 0.5241,
+      "step": 7321
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.913264207565214e-09,
+      "loss": 0.419,
+      "step": 7322
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.839050990233738e-09,
+      "loss": 0.9688,
+      "step": 7323
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7657950026094283e-09,
+      "loss": 0.6103,
+      "step": 7324
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.693496258723283e-09,
+      "loss": 0.7052,
+      "step": 7325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6221547724253337e-09,
+      "loss": 1.1142,
+      "step": 7326
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.551770557381872e-09,
+      "loss": 0.7734,
+      "step": 7327
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.48234362707378e-09,
+      "loss": 0.6552,
+      "step": 7328
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4138739948020852e-09,
+      "loss": 0.8315,
+      "step": 7329
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3463616736807415e-09,
+      "loss": 0.65,
+      "step": 7330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2798066766427373e-09,
+      "loss": 0.7934,
+      "step": 7331
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2142090164373188e-09,
+      "loss": 0.7168,
+      "step": 7332
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.14956870562999e-09,
+      "loss": 0.5465,
+      "step": 7333
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0858857566019576e-09,
+      "loss": 0.7189,
+      "step": 7334
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0231601815529078e-09,
+      "loss": 0.8716,
+      "step": 7335
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9613919924982293e-09,
+      "loss": 0.7662,
+      "step": 7336
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.900581201269569e-09,
+      "loss": 0.3888,
+      "step": 7337
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8407278195153866e-09,
+      "loss": 0.7739,
+      "step": 7338
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7818318587015103e-09,
+      "loss": 0.6248,
+      "step": 7339
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7238933301089166e-09,
+      "loss": 0.3768,
+      "step": 7340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.66691224483706e-09,
+      "loss": 1.0806,
+      "step": 7341
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6108886137994329e-09,
+      "loss": 0.7519,
+      "step": 7342
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5558224477291161e-09,
+      "loss": 0.9877,
+      "step": 7343
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5017137571732287e-09,
+      "loss": 0.5376,
+      "step": 7344
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4485625524973679e-09,
+      "loss": 0.7498,
+      "step": 7345
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3963688438822787e-09,
+      "loss": 1.0236,
+      "step": 7346
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3451326413266296e-09,
+      "loss": 0.9446,
+      "step": 7347
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2948539546436823e-09,
+      "loss": 0.7225,
+      "step": 7348
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2455327934657312e-09,
+      "loss": 0.5264,
+      "step": 7349
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1971691672402198e-09,
+      "loss": 0.8263,
+      "step": 7350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1497630852314035e-09,
+      "loss": 0.6212,
+      "step": 7351
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1033145565197966e-09,
+      "loss": 0.7741,
+      "step": 7352
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0578235900032819e-09,
+      "loss": 0.6597,
+      "step": 7353
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0132901943954444e-09,
+      "loss": 0.8048,
+      "step": 7354
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.697143782272378e-10,
+      "loss": 0.6575,
+      "step": 7355
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.27096149845319e-10,
+      "loss": 0.8007,
+      "step": 7356
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.85435517413713e-10,
+      "loss": 0.9165,
+      "step": 7357
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.44732488913258e-10,
+      "loss": 0.7416,
+      "step": 7358
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.0498707213994e-10,
+      "loss": 0.4734,
+      "step": 7359
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.661992747076686e-10,
+      "loss": 0.7791,
+      "step": 7360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.283691040466113e-10,
+      "loss": 0.9915,
+      "step": 7361
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.914965674026386e-10,
+      "loss": 0.8615,
+      "step": 7362
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.555816718389896e-10,
+      "loss": 0.81,
+      "step": 7363
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.206244242362713e-10,
+      "loss": 0.5944,
+      "step": 7364
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.866248312896839e-10,
+      "loss": 0.9539,
+      "step": 7365
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.535828995117954e-10,
+      "loss": 1.1966,
+      "step": 7366
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.214986352330976e-10,
+      "loss": 0.6337,
+      "step": 7367
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.903720445986749e-10,
+      "loss": 0.6817,
+      "step": 7368
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.602031335715351e-10,
+      "loss": 0.6787,
+      "step": 7369
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3099190792983416e-10,
+      "loss": 0.4803,
+      "step": 7370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0273837326965106e-10,
+      "loss": 0.5648,
+      "step": 7371
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7544253500332306e-10,
+      "loss": 0.7686,
+      "step": 7372
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4910439835889044e-10,
+      "loss": 1.1369,
+      "step": 7373
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.237239683817617e-10,
+      "loss": 0.9339,
+      "step": 7374
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9930124993415857e-10,
+      "loss": 1.1612,
+      "step": 7375
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7583624769345065e-10,
+      "loss": 0.6153,
+      "step": 7376
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.533289661554861e-10,
+      "loss": 0.5098,
+      "step": 7377
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3177940963070577e-10,
+      "loss": 0.537,
+      "step": 7378
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1118758224802916e-10,
+      "loss": 0.5287,
+      "step": 7379
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.915534879509684e-10,
+      "loss": 1.0527,
+      "step": 7380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.728771305009591e-10,
+      "loss": 0.805,
+      "step": 7381
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.55158513475695e-10,
+      "loss": 0.5384,
+      "step": 7382
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3839764026857271e-10,
+      "loss": 0.9675,
+      "step": 7383
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.225945140909124e-10,
+      "loss": 1.0498,
+      "step": 7384
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0774913797029218e-10,
+      "loss": 0.4712,
+      "step": 7385
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.386151474943817e-11,
+      "loss": 0.7943,
+      "step": 7386
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.09316470884447e-11,
+      "loss": 0.9356,
+      "step": 7387
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.895953746532957e-11,
+      "loss": 0.5814,
+      "step": 7388
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.7945188172148227e-11,
+      "loss": 0.9499,
+      "step": 7389
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.788860131943463e-11,
+      "loss": 0.9065,
+      "step": 7390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.878977883342572e-11,
+      "loss": 0.8127,
+      "step": 7391
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.064872245717165e-11,
+      "loss": 0.8655,
+      "step": 7392
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.346543374998067e-11,
+      "loss": 0.7647,
+      "step": 7393
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7239914087974208e-11,
+      "loss": 0.6337,
+      "step": 7394
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1972164662976683e-11,
+      "loss": 0.9778,
+      "step": 7395
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.662186485291045e-12,
+      "loss": 0.7889,
+      "step": 7396
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3099803792578944e-12,
+      "loss": 0.4769,
+      "step": 7397
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.915546988251471e-12,
+      "loss": 0.9913,
+      "step": 7398
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.788867702387734e-13,
+      "loss": 0.6042,
+      "step": 7399
+    },
+    {
+      "epoch": 1.0,
+      "step": 7399,
+      "total_flos": 1.3968111656857764e+18,
+      "train_loss": 0.8074634926559023,
+      "train_runtime": 48533.0079,
+      "train_samples_per_second": 19.514,
+      "train_steps_per_second": 0.152
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 7400,
+  "num_train_epochs": 1,
+  "save_steps": 2500,
+  "total_flos": 1.3968111656857764e+18,
+  "trial_name": null,
+  "trial_params": null
+}