diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,65635 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 100.0,
+  "global_step": 10900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0,
+      "loss": 16.1388,
+      "step": 1
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0,
+      "loss": 14.7638,
+      "step": 2
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 6e-07,
+      "loss": 14.203,
+      "step": 3
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2e-06,
+      "loss": 14.6627,
+      "step": 4
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8e-06,
+      "loss": 14.5492,
+      "step": 5
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.4e-06,
+      "loss": 17.2444,
+      "step": 6
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 15.1152,
+      "step": 7
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 3.6e-06,
+      "loss": 15.5723,
+      "step": 8
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.2e-06,
+      "loss": 15.5678,
+      "step": 9
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8e-06,
+      "loss": 14.5661,
+      "step": 10
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 5.399999999999999e-06,
+      "loss": 15.1542,
+      "step": 11
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 14.8203,
+      "step": 12
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.599999999999999e-06,
+      "loss": 14.9697,
+      "step": 13
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.2e-06,
+      "loss": 14.5007,
+      "step": 14
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.799999999999998e-06,
+      "loss": 14.5995,
+      "step": 15
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 7.799999999999998e-06,
+      "loss": 13.5446,
+      "step": 16
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.4e-06,
+      "loss": 13.8205,
+      "step": 17
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 13.3675,
+      "step": 18
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.6e-06,
+      "loss": 13.451,
+      "step": 19
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.02e-05,
+      "loss": 16.3684,
+      "step": 20
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.0799999999999998e-05,
+      "loss": 14.7638,
+      "step": 21
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.14e-05,
+      "loss": 14.5562,
+      "step": 22
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 13.251,
+      "step": 23
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.26e-05,
+      "loss": 13.2937,
+      "step": 24
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3199999999999997e-05,
+      "loss": 13.1883,
+      "step": 25
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.3799999999999998e-05,
+      "loss": 13.7261,
+      "step": 26
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.44e-05,
+      "loss": 13.5788,
+      "step": 27
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 13.5407,
+      "step": 28
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.5599999999999996e-05,
+      "loss": 13.3514,
+      "step": 29
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6199999999999997e-05,
+      "loss": 11.3071,
+      "step": 30
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68e-05,
+      "loss": 11.6792,
+      "step": 31
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.74e-05,
+      "loss": 11.8642,
+      "step": 32
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 13.2409,
+      "step": 33
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8599999999999998e-05,
+      "loss": 11.3985,
+      "step": 34
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.92e-05,
+      "loss": 11.1092,
+      "step": 35
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.98e-05,
+      "loss": 10.148,
+      "step": 36
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.04e-05,
+      "loss": 9.8871,
+      "step": 37
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.1e-05,
+      "loss": 10.0808,
+      "step": 38
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.1599999999999996e-05,
+      "loss": 10.485,
+      "step": 39
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2199999999999998e-05,
+      "loss": 9.1753,
+      "step": 40
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.28e-05,
+      "loss": 8.3659,
+      "step": 41
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.34e-05,
+      "loss": 8.2671,
+      "step": 42
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 7.5912,
+      "step": 43
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.4599999999999998e-05,
+      "loss": 8.2449,
+      "step": 44
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.52e-05,
+      "loss": 7.8486,
+      "step": 45
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.5799999999999997e-05,
+      "loss": 7.6262,
+      "step": 46
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.6399999999999995e-05,
+      "loss": 7.0738,
+      "step": 47
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 6.9934,
+      "step": 48
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7599999999999997e-05,
+      "loss": 6.2465,
+      "step": 49
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 7.2318,
+      "step": 50
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.88e-05,
+      "loss": 6.6769,
+      "step": 51
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.94e-05,
+      "loss": 6.5081,
+      "step": 52
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 5.7597,
+      "step": 53
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.06e-05,
+      "loss": 5.9698,
+      "step": 54
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.119999999999999e-05,
+      "loss": 5.7114,
+      "step": 55
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.1799999999999994e-05,
+      "loss": 5.5164,
+      "step": 56
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2399999999999995e-05,
+      "loss": 6.1398,
+      "step": 57
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 5.5499,
+      "step": 58
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.36e-05,
+      "loss": 5.4479,
+      "step": 59
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.42e-05,
+      "loss": 5.4245,
+      "step": 60
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.48e-05,
+      "loss": 5.0095,
+      "step": 61
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.539999999999999e-05,
+      "loss": 5.0945,
+      "step": 62
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 5.4826,
+      "step": 63
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6599999999999995e-05,
+      "loss": 5.2119,
+      "step": 64
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7199999999999996e-05,
+      "loss": 4.959,
+      "step": 65
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.78e-05,
+      "loss": 4.6605,
+      "step": 66
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.84e-05,
+      "loss": 4.7602,
+      "step": 67
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9e-05,
+      "loss": 4.5658,
+      "step": 68
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.96e-05,
+      "loss": 5.0442,
+      "step": 69
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.02e-05,
+      "loss": 4.7574,
+      "step": 70
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.08e-05,
+      "loss": 4.6685,
+      "step": 71
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.14e-05,
+      "loss": 4.6069,
+      "step": 72
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.2e-05,
+      "loss": 4.5866,
+      "step": 73
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.259999999999999e-05,
+      "loss": 4.3165,
+      "step": 74
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.319999999999999e-05,
+      "loss": 4.2273,
+      "step": 75
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3799999999999994e-05,
+      "loss": 4.6323,
+      "step": 76
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4399999999999995e-05,
+      "loss": 4.4517,
+      "step": 77
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 4.6156,
+      "step": 78
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.56e-05,
+      "loss": 4.2353,
+      "step": 79
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.62e-05,
+      "loss": 4.292,
+      "step": 80
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.68e-05,
+      "loss": 4.2268,
+      "step": 81
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.7399999999999993e-05,
+      "loss": 4.3657,
+      "step": 82
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 4.2215,
+      "step": 83
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.8599999999999995e-05,
+      "loss": 4.1501,
+      "step": 84
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9199999999999997e-05,
+      "loss": 4.2086,
+      "step": 85
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.98e-05,
+      "loss": 4.0437,
+      "step": 86
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.04e-05,
+      "loss": 4.1459,
+      "step": 87
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.1e-05,
+      "loss": 4.7229,
+      "step": 88
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.1599999999999994e-05,
+      "loss": 4.1088,
+      "step": 89
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.2199999999999995e-05,
+      "loss": 4.0703,
+      "step": 90
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.279999999999999e-05,
+      "loss": 3.9748,
+      "step": 91
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.339999999999999e-05,
+      "loss": 3.9837,
+      "step": 92
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 3.9323,
+      "step": 93
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.459999999999999e-05,
+      "loss": 4.1823,
+      "step": 94
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.519999999999999e-05,
+      "loss": 3.9706,
+      "step": 95
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.5799999999999994e-05,
+      "loss": 3.8621,
+      "step": 96
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 3.8612,
+      "step": 97
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 3.7782,
+      "step": 98
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.76e-05,
+      "loss": 3.7542,
+      "step": 99
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.82e-05,
+      "loss": 3.899,
+      "step": 100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.88e-05,
+      "loss": 3.8986,
+      "step": 101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.94e-05,
+      "loss": 3.8067,
+      "step": 102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 3.8077,
+      "step": 103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.0599999999999996e-05,
+      "loss": 3.7054,
+      "step": 104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.12e-05,
+      "loss": 3.6935,
+      "step": 105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.18e-05,
+      "loss": 3.7779,
+      "step": 106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.239999999999999e-05,
+      "loss": 3.8345,
+      "step": 107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 3.6368,
+      "step": 108
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.359999999999999e-05,
+      "loss": 3.7975,
+      "step": 109
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 6.419999999999999e-05,
+      "loss": 3.6648,
+      "step": 110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.479999999999999e-05,
+      "loss": 3.6839,
+      "step": 111
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 6.539999999999999e-05,
+      "loss": 3.583,
+      "step": 112
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 3.5326,
+      "step": 113
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 6.659999999999999e-05,
+      "loss": 3.508,
+      "step": 114
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 6.72e-05,
+      "loss": 3.971,
+      "step": 115
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 6.78e-05,
+      "loss": 3.5742,
+      "step": 116
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 6.84e-05,
+      "loss": 3.534,
+      "step": 117
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 6.9e-05,
+      "loss": 3.4967,
+      "step": 118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 6.96e-05,
+      "loss": 3.5018,
+      "step": 119
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.02e-05,
+      "loss": 3.4305,
+      "step": 120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.079999999999999e-05,
+      "loss": 3.5876,
+      "step": 121
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.139999999999999e-05,
+      "loss": 3.4753,
+      "step": 122
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 3.479,
+      "step": 123
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.259999999999999e-05,
+      "loss": 3.4577,
+      "step": 124
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.319999999999999e-05,
+      "loss": 3.4247,
+      "step": 125
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.379999999999999e-05,
+      "loss": 3.388,
+      "step": 126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.439999999999999e-05,
+      "loss": 3.3731,
+      "step": 127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.5e-05,
+      "loss": 3.43,
+      "step": 128
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.56e-05,
+      "loss": 3.3663,
+      "step": 129
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.62e-05,
+      "loss": 3.3643,
+      "step": 130
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.68e-05,
+      "loss": 3.3239,
+      "step": 131
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.74e-05,
+      "loss": 3.3568,
+      "step": 132
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.8e-05,
+      "loss": 3.3012,
+      "step": 133
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.86e-05,
+      "loss": 3.3442,
+      "step": 134
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.92e-05,
+      "loss": 3.3846,
+      "step": 135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.98e-05,
+      "loss": 3.2915,
+      "step": 136
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.04e-05,
+      "loss": 3.3022,
+      "step": 137
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.1e-05,
+      "loss": 3.2539,
+      "step": 138
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.16e-05,
+      "loss": 3.2323,
+      "step": 139
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.22e-05,
+      "loss": 3.3579,
+      "step": 140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.28e-05,
+      "loss": 3.337,
+      "step": 141
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.34e-05,
+      "loss": 3.274,
+      "step": 142
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.4e-05,
+      "loss": 3.2516,
+      "step": 143
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.459999999999998e-05,
+      "loss": 3.2522,
+      "step": 144
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.519999999999998e-05,
+      "loss": 3.209,
+      "step": 145
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.579999999999998e-05,
+      "loss": 3.199,
+      "step": 146
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.639999999999999e-05,
+      "loss": 3.3219,
+      "step": 147
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 3.2243,
+      "step": 148
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.759999999999999e-05,
+      "loss": 3.2115,
+      "step": 149
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.819999999999999e-05,
+      "loss": 3.1722,
+      "step": 150
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.879999999999999e-05,
+      "loss": 3.2155,
+      "step": 151
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.939999999999999e-05,
+      "loss": 3.1864,
+      "step": 152
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 3.1972,
+      "step": 153
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.059999999999999e-05,
+      "loss": 3.1867,
+      "step": 154
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 9.12e-05,
+      "loss": 3.1745,
+      "step": 155
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 9.18e-05,
+      "loss": 3.2028,
+      "step": 156
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.24e-05,
+      "loss": 3.1843,
+      "step": 157
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 9.3e-05,
+      "loss": 3.2242,
+      "step": 158
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 9.36e-05,
+      "loss": 3.3006,
+      "step": 159
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.419999999999999e-05,
+      "loss": 3.1746,
+      "step": 160
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.479999999999999e-05,
+      "loss": 3.1482,
+      "step": 161
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 9.539999999999999e-05,
+      "loss": 3.1883,
+      "step": 162
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 3.1903,
+      "step": 163
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.659999999999999e-05,
+      "loss": 3.1663,
+      "step": 164
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.719999999999999e-05,
+      "loss": 3.1931,
+      "step": 165
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.779999999999999e-05,
+      "loss": 3.1813,
+      "step": 166
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.839999999999999e-05,
+      "loss": 3.1649,
+      "step": 167
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.9e-05,
+      "loss": 3.13,
+      "step": 168
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.96e-05,
+      "loss": 3.1572,
+      "step": 169
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001002,
+      "loss": 3.1504,
+      "step": 170
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001008,
+      "loss": 3.1867,
+      "step": 171
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001014,
+      "loss": 3.2155,
+      "step": 172
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.000102,
+      "loss": 3.1583,
+      "step": 173
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001026,
+      "loss": 3.1541,
+      "step": 174
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00010319999999999999,
+      "loss": 3.1036,
+      "step": 175
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00010379999999999999,
+      "loss": 3.1464,
+      "step": 176
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00010439999999999999,
+      "loss": 3.1783,
+      "step": 177
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 3.2176,
+      "step": 178
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00010559999999999998,
+      "loss": 3.1353,
+      "step": 179
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00010619999999999998,
+      "loss": 3.1107,
+      "step": 180
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00010679999999999998,
+      "loss": 3.1562,
+      "step": 181
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00010739999999999998,
+      "loss": 3.1309,
+      "step": 182
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 3.1707,
+      "step": 183
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00010859999999999998,
+      "loss": 3.3286,
+      "step": 184
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00010919999999999998,
+      "loss": 3.1253,
+      "step": 185
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00010979999999999999,
+      "loss": 3.1367,
+      "step": 186
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00011039999999999999,
+      "loss": 3.1526,
+      "step": 187
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 3.1757,
+      "step": 188
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00011159999999999999,
+      "loss": 3.1801,
+      "step": 189
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00011219999999999999,
+      "loss": 3.1572,
+      "step": 190
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011279999999999999,
+      "loss": 3.2281,
+      "step": 191
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011339999999999999,
+      "loss": 3.1563,
+      "step": 192
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 3.104,
+      "step": 193
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001146,
+      "loss": 3.1815,
+      "step": 194
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001152,
+      "loss": 3.1655,
+      "step": 195
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001158,
+      "loss": 3.148,
+      "step": 196
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001164,
+      "loss": 3.222,
+      "step": 197
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.000117,
+      "loss": 3.138,
+      "step": 198
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001176,
+      "loss": 3.1277,
+      "step": 199
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001182,
+      "loss": 3.1316,
+      "step": 200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001188,
+      "loss": 3.1028,
+      "step": 201
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001194,
+      "loss": 3.1135,
+      "step": 202
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 3.2245,
+      "step": 203
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00012059999999999999,
+      "loss": 3.1546,
+      "step": 204
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00012119999999999999,
+      "loss": 3.1286,
+      "step": 205
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00012179999999999999,
+      "loss": 3.0974,
+      "step": 206
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001224,
+      "loss": 3.136,
+      "step": 207
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 3.168,
+      "step": 208
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001236,
+      "loss": 3.1192,
+      "step": 209
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00012419999999999998,
+      "loss": 3.0744,
+      "step": 210
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00012479999999999997,
+      "loss": 3.1006,
+      "step": 211
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00012539999999999999,
+      "loss": 3.0912,
+      "step": 212
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 3.1342,
+      "step": 213
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001266,
+      "loss": 3.1681,
+      "step": 214
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00012719999999999997,
+      "loss": 3.1582,
+      "step": 215
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0001278,
+      "loss": 3.1323,
+      "step": 216
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00012839999999999998,
+      "loss": 3.0959,
+      "step": 217
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.000129,
+      "loss": 3.1162,
+      "step": 218
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00012959999999999998,
+      "loss": 3.084,
+      "step": 219
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0001302,
+      "loss": 3.0845,
+      "step": 220
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00013079999999999998,
+      "loss": 3.0634,
+      "step": 221
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0001314,
+      "loss": 3.082,
+      "step": 222
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 3.1108,
+      "step": 223
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0001326,
+      "loss": 3.1663,
+      "step": 224
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00013319999999999999,
+      "loss": 3.1371,
+      "step": 225
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0001338,
+      "loss": 3.1042,
+      "step": 226
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0001344,
+      "loss": 3.0674,
+      "step": 227
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.000135,
+      "loss": 3.1057,
+      "step": 228
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0001356,
+      "loss": 3.0597,
+      "step": 229
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0001362,
+      "loss": 3.0892,
+      "step": 230
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0001368,
+      "loss": 3.1795,
+      "step": 231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0001374,
+      "loss": 3.073,
+      "step": 232
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.000138,
+      "loss": 3.073,
+      "step": 233
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0001386,
+      "loss": 3.0837,
+      "step": 234
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0001392,
+      "loss": 3.0763,
+      "step": 235
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00013979999999999998,
+      "loss": 3.0813,
+      "step": 236
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0001404,
+      "loss": 3.281,
+      "step": 237
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 3.0822,
+      "step": 238
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00014159999999999997,
+      "loss": 3.0708,
+      "step": 239
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0001422,
+      "loss": 3.0866,
+      "step": 240
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00014279999999999997,
+      "loss": 3.0903,
+      "step": 241
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0001434,
+      "loss": 3.0935,
+      "step": 242
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 3.1114,
+      "step": 243
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0001446,
+      "loss": 3.0885,
+      "step": 244
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00014519999999999998,
+      "loss": 3.1056,
+      "step": 245
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0001458,
+      "loss": 3.0828,
+      "step": 246
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00014639999999999998,
+      "loss": 3.0616,
+      "step": 247
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000147,
+      "loss": 3.0725,
+      "step": 248
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00014759999999999998,
+      "loss": 3.0719,
+      "step": 249
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0001482,
+      "loss": 3.1351,
+      "step": 250
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00014879999999999998,
+      "loss": 3.0717,
+      "step": 251
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0001494,
+      "loss": 3.0708,
+      "step": 252
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00015,
+      "loss": 3.0825,
+      "step": 253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00015059999999999997,
+      "loss": 3.0532,
+      "step": 254
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0001512,
+      "loss": 3.1436,
+      "step": 255
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00015179999999999998,
+      "loss": 3.0363,
+      "step": 256
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0001524,
+      "loss": 3.0774,
+      "step": 257
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 3.0521,
+      "step": 258
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0001536,
+      "loss": 3.0394,
+      "step": 259
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00015419999999999998,
+      "loss": 3.0705,
+      "step": 260
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0001548,
+      "loss": 3.0519,
+      "step": 261
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00015539999999999998,
+      "loss": 3.0563,
+      "step": 262
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.000156,
+      "loss": 3.0426,
+      "step": 263
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00015659999999999998,
+      "loss": 3.0367,
+      "step": 264
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0001572,
+      "loss": 3.0503,
+      "step": 265
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0001578,
+      "loss": 3.0466,
+      "step": 266
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0001584,
+      "loss": 3.0004,
+      "step": 267
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.000159,
+      "loss": 3.043,
+      "step": 268
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0001596,
+      "loss": 3.0422,
+      "step": 269
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0001602,
+      "loss": 3.0171,
+      "step": 270
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0001608,
+      "loss": 3.0206,
+      "step": 271
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0001614,
+      "loss": 3.0498,
+      "step": 272
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000162,
+      "loss": 3.021,
+      "step": 273
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0001626,
+      "loss": 3.0138,
+      "step": 274
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0001632,
+      "loss": 3.0486,
+      "step": 275
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0001638,
+      "loss": 3.0137,
+      "step": 276
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0001644,
+      "loss": 3.0379,
+      "step": 277
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.000165,
+      "loss": 3.0422,
+      "step": 278
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0001656,
+      "loss": 2.9999,
+      "step": 279
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0001662,
+      "loss": 3.0735,
+      "step": 280
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0001668,
+      "loss": 3.0187,
+      "step": 281
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0001674,
+      "loss": 3.044,
+      "step": 282
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.000168,
+      "loss": 3.0199,
+      "step": 283
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0001686,
+      "loss": 3.0556,
+      "step": 284
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00016919999999999997,
+      "loss": 3.0548,
+      "step": 285
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00016979999999999998,
+      "loss": 3.0227,
+      "step": 286
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00017039999999999997,
+      "loss": 3.0085,
+      "step": 287
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 3.0361,
+      "step": 288
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00017159999999999997,
+      "loss": 3.0181,
+      "step": 289
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00017219999999999998,
+      "loss": 3.0192,
+      "step": 290
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00017279999999999997,
+      "loss": 3.015,
+      "step": 291
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00017339999999999996,
+      "loss": 3.0491,
+      "step": 292
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 3.1035,
+      "step": 293
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00017459999999999996,
+      "loss": 3.0,
+      "step": 294
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00017519999999999998,
+      "loss": 3.0302,
+      "step": 295
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00017579999999999996,
+      "loss": 3.0742,
+      "step": 296
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00017639999999999998,
+      "loss": 3.0235,
+      "step": 297
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 3.0185,
+      "step": 298
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00017759999999999998,
+      "loss": 3.0257,
+      "step": 299
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00017819999999999997,
+      "loss": 3.0845,
+      "step": 300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00017879999999999998,
+      "loss": 3.0345,
+      "step": 301
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00017939999999999997,
+      "loss": 3.0276,
+      "step": 302
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 3.0464,
+      "step": 303
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00018059999999999997,
+      "loss": 3.0411,
+      "step": 304
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00018119999999999999,
+      "loss": 3.0183,
+      "step": 305
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00018179999999999997,
+      "loss": 3.0108,
+      "step": 306
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0001824,
+      "loss": 3.0168,
+      "step": 307
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 3.0104,
+      "step": 308
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0001836,
+      "loss": 2.9945,
+      "step": 309
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00018419999999999998,
+      "loss": 3.0081,
+      "step": 310
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0001848,
+      "loss": 3.0171,
+      "step": 311
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00018539999999999998,
+      "loss": 3.064,
+      "step": 312
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.000186,
+      "loss": 3.0435,
+      "step": 313
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00018659999999999998,
+      "loss": 2.9983,
+      "step": 314
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0001872,
+      "loss": 2.986,
+      "step": 315
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00018779999999999998,
+      "loss": 3.0006,
+      "step": 316
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00018839999999999997,
+      "loss": 2.9705,
+      "step": 317
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 3.0438,
+      "step": 318
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.00018959999999999997,
+      "loss": 3.0192,
+      "step": 319
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0001902,
+      "loss": 3.0054,
+      "step": 320
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00019079999999999998,
+      "loss": 2.9989,
+      "step": 321
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0001914,
+      "loss": 2.9742,
+      "step": 322
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 2.948,
+      "step": 323
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0001926,
+      "loss": 2.9967,
+      "step": 324
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00019319999999999998,
+      "loss": 2.9783,
+      "step": 325
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0001938,
+      "loss": 2.9504,
+      "step": 326
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00019439999999999998,
+      "loss": 2.9449,
+      "step": 327
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.000195,
+      "loss": 2.9565,
+      "step": 328
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00019559999999999998,
+      "loss": 2.9389,
+      "step": 329
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0001962,
+      "loss": 2.9903,
+      "step": 330
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00019679999999999999,
+      "loss": 2.9536,
+      "step": 331
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0001974,
+      "loss": 2.9785,
+      "step": 332
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.000198,
+      "loss": 2.9922,
+      "step": 333
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0001986,
+      "loss": 3.0289,
+      "step": 334
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0001992,
+      "loss": 2.9932,
+      "step": 335
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0001998,
+      "loss": 2.9612,
+      "step": 336
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002004,
+      "loss": 2.928,
+      "step": 337
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.000201,
+      "loss": 2.9267,
+      "step": 338
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002016,
+      "loss": 2.964,
+      "step": 339
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002022,
+      "loss": 2.9404,
+      "step": 340
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002028,
+      "loss": 2.9203,
+      "step": 341
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00020339999999999998,
+      "loss": 2.9491,
+      "step": 342
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.000204,
+      "loss": 2.9421,
+      "step": 343
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00020459999999999999,
+      "loss": 2.9345,
+      "step": 344
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002052,
+      "loss": 2.9269,
+      "step": 345
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002058,
+      "loss": 2.9366,
+      "step": 346
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00020639999999999998,
+      "loss": 2.9274,
+      "step": 347
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 2.9307,
+      "step": 348
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00020759999999999998,
+      "loss": 2.951,
+      "step": 349
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00020819999999999996,
+      "loss": 2.91,
+      "step": 350
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00020879999999999998,
+      "loss": 2.8697,
+      "step": 351
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00020939999999999997,
+      "loss": 2.8476,
+      "step": 352
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 2.9795,
+      "step": 353
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00021059999999999997,
+      "loss": 2.8879,
+      "step": 354
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00021119999999999996,
+      "loss": 2.8539,
+      "step": 355
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00021179999999999997,
+      "loss": 2.9072,
+      "step": 356
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00021239999999999996,
+      "loss": 2.9411,
+      "step": 357
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 2.8558,
+      "step": 358
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00021359999999999996,
+      "loss": 2.9355,
+      "step": 359
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00021419999999999998,
+      "loss": 2.8892,
+      "step": 360
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00021479999999999996,
+      "loss": 2.892,
+      "step": 361
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00021539999999999998,
+      "loss": 2.895,
+      "step": 362
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 2.8323,
+      "step": 363
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00021659999999999998,
+      "loss": 2.8535,
+      "step": 364
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00021719999999999997,
+      "loss": 2.8355,
+      "step": 365
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00021779999999999998,
+      "loss": 2.8535,
+      "step": 366
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00021839999999999997,
+      "loss": 2.8405,
+      "step": 367
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 2.8272,
+      "step": 368
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00021959999999999997,
+      "loss": 2.8116,
+      "step": 369
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00022019999999999999,
+      "loss": 2.7921,
+      "step": 370
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00022079999999999997,
+      "loss": 2.7819,
+      "step": 371
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002214,
+      "loss": 2.7962,
+      "step": 372
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 2.8001,
+      "step": 373
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002226,
+      "loss": 2.7157,
+      "step": 374
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00022319999999999998,
+      "loss": 2.7487,
+      "step": 375
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002238,
+      "loss": 2.721,
+      "step": 376
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00022439999999999998,
+      "loss": 2.7406,
+      "step": 377
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.000225,
+      "loss": 2.7688,
+      "step": 378
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00022559999999999998,
+      "loss": 2.7693,
+      "step": 379
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00022619999999999997,
+      "loss": 2.7143,
+      "step": 380
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00022679999999999998,
+      "loss": 2.7133,
+      "step": 381
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00022739999999999997,
+      "loss": 2.6917,
+      "step": 382
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 2.7178,
+      "step": 383
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00022859999999999997,
+      "loss": 2.6668,
+      "step": 384
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002292,
+      "loss": 2.6418,
+      "step": 385
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00022979999999999997,
+      "loss": 2.5992,
+      "step": 386
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002304,
+      "loss": 2.6048,
+      "step": 387
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 2.5943,
+      "step": 388
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002316,
+      "loss": 2.629,
+      "step": 389
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00023219999999999998,
+      "loss": 2.5857,
+      "step": 390
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002328,
+      "loss": 2.5712,
+      "step": 391
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00023339999999999998,
+      "loss": 2.5597,
+      "step": 392
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.000234,
+      "loss": 2.5636,
+      "step": 393
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00023459999999999998,
+      "loss": 2.5792,
+      "step": 394
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002352,
+      "loss": 2.4681,
+      "step": 395
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00023579999999999999,
+      "loss": 2.5453,
+      "step": 396
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002364,
+      "loss": 2.4647,
+      "step": 397
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.000237,
+      "loss": 2.4946,
+      "step": 398
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002376,
+      "loss": 2.4757,
+      "step": 399
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002382,
+      "loss": 2.4642,
+      "step": 400
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002388,
+      "loss": 2.3904,
+      "step": 401
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002394,
+      "loss": 2.4381,
+      "step": 402
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 2.4957,
+      "step": 403
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002406,
+      "loss": 2.3476,
+      "step": 404
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00024119999999999998,
+      "loss": 2.4591,
+      "step": 405
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002418,
+      "loss": 2.424,
+      "step": 406
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00024239999999999998,
+      "loss": 2.4619,
+      "step": 407
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.000243,
+      "loss": 2.4581,
+      "step": 408
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00024359999999999999,
+      "loss": 2.3042,
+      "step": 409
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00024419999999999997,
+      "loss": 2.4113,
+      "step": 410
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002448,
+      "loss": 2.3109,
+      "step": 411
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00024539999999999995,
+      "loss": 2.323,
+      "step": 412
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 2.3387,
+      "step": 413
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002466,
+      "loss": 2.2672,
+      "step": 414
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0002472,
+      "loss": 2.3961,
+      "step": 415
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00024779999999999995,
+      "loss": 2.2778,
+      "step": 416
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00024839999999999997,
+      "loss": 2.2023,
+      "step": 417
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000249,
+      "loss": 2.2,
+      "step": 418
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00024959999999999994,
+      "loss": 2.108,
+      "step": 419
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00025019999999999996,
+      "loss": 2.2234,
+      "step": 420
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00025079999999999997,
+      "loss": 2.195,
+      "step": 421
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002514,
+      "loss": 2.0259,
+      "step": 422
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 2.1704,
+      "step": 423
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025259999999999996,
+      "loss": 2.0296,
+      "step": 424
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002532,
+      "loss": 2.109,
+      "step": 425
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002538,
+      "loss": 2.1596,
+      "step": 426
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00025439999999999995,
+      "loss": 2.2733,
+      "step": 427
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 2.0781,
+      "step": 428
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002556,
+      "loss": 2.0981,
+      "step": 429
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002562,
+      "loss": 1.9903,
+      "step": 430
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00025679999999999995,
+      "loss": 1.9912,
+      "step": 431
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025739999999999997,
+      "loss": 2.0593,
+      "step": 432
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000258,
+      "loss": 1.9455,
+      "step": 433
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002586,
+      "loss": 2.1047,
+      "step": 434
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00025919999999999996,
+      "loss": 2.0198,
+      "step": 435
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00025979999999999997,
+      "loss": 2.083,
+      "step": 436
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002604,
+      "loss": 2.0199,
+      "step": 437
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.000261,
+      "loss": 1.9786,
+      "step": 438
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00026159999999999996,
+      "loss": 1.8767,
+      "step": 439
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002622,
+      "loss": 2.031,
+      "step": 440
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002628,
+      "loss": 1.9412,
+      "step": 441
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00026339999999999995,
+      "loss": 2.0339,
+      "step": 442
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 2.001,
+      "step": 443
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002646,
+      "loss": 1.8467,
+      "step": 444
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002652,
+      "loss": 1.9853,
+      "step": 445
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00026579999999999996,
+      "loss": 1.9902,
+      "step": 446
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00026639999999999997,
+      "loss": 1.8824,
+      "step": 447
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.000267,
+      "loss": 1.993,
+      "step": 448
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002676,
+      "loss": 1.9494,
+      "step": 449
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00026819999999999996,
+      "loss": 1.9021,
+      "step": 450
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002688,
+      "loss": 1.9085,
+      "step": 451
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002694,
+      "loss": 1.8425,
+      "step": 452
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00027,
+      "loss": 1.886,
+      "step": 453
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00027059999999999996,
+      "loss": 1.9113,
+      "step": 454
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0002712,
+      "loss": 1.8431,
+      "step": 455
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002718,
+      "loss": 1.8845,
+      "step": 456
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002724,
+      "loss": 1.925,
+      "step": 457
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 1.8258,
+      "step": 458
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0002736,
+      "loss": 1.7457,
+      "step": 459
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002742,
+      "loss": 1.9006,
+      "step": 460
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002748,
+      "loss": 1.8802,
+      "step": 461
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00027539999999999997,
+      "loss": 1.8806,
+      "step": 462
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.000276,
+      "loss": 1.7432,
+      "step": 463
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.0002766,
+      "loss": 1.7444,
+      "step": 464
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002772,
+      "loss": 1.9312,
+      "step": 465
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002778,
+      "loss": 1.8729,
+      "step": 466
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002784,
+      "loss": 1.8268,
+      "step": 467
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.000279,
+      "loss": 1.9584,
+      "step": 468
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00027959999999999997,
+      "loss": 1.7953,
+      "step": 469
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0002802,
+      "loss": 1.8302,
+      "step": 470
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002808,
+      "loss": 1.715,
+      "step": 471
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00028139999999999996,
+      "loss": 1.8619,
+      "step": 472
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 1.7755,
+      "step": 473
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0002826,
+      "loss": 1.9204,
+      "step": 474
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00028319999999999994,
+      "loss": 1.8149,
+      "step": 475
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00028379999999999996,
+      "loss": 1.8075,
+      "step": 476
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0002844,
+      "loss": 1.763,
+      "step": 477
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.000285,
+      "loss": 1.7823,
+      "step": 478
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00028559999999999995,
+      "loss": 1.8212,
+      "step": 479
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00028619999999999996,
+      "loss": 1.8485,
+      "step": 480
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002868,
+      "loss": 1.7713,
+      "step": 481
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.00028739999999999994,
+      "loss": 1.7514,
+      "step": 482
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 1.8947,
+      "step": 483
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00028859999999999997,
+      "loss": 1.6895,
+      "step": 484
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.0002892,
+      "loss": 1.6898,
+      "step": 485
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00028979999999999994,
+      "loss": 1.9068,
+      "step": 486
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00029039999999999996,
+      "loss": 1.7711,
+      "step": 487
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 1.8673,
+      "step": 488
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002916,
+      "loss": 1.8592,
+      "step": 489
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00029219999999999995,
+      "loss": 1.8277,
+      "step": 490
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00029279999999999996,
+      "loss": 1.8501,
+      "step": 491
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002934,
+      "loss": 1.8055,
+      "step": 492
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.000294,
+      "loss": 1.7788,
+      "step": 493
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00029459999999999995,
+      "loss": 1.6908,
+      "step": 494
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00029519999999999997,
+      "loss": 1.6804,
+      "step": 495
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.0002958,
+      "loss": 1.853,
+      "step": 496
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.0002964,
+      "loss": 1.6991,
+      "step": 497
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 1.727,
+      "step": 498
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00029759999999999997,
+      "loss": 1.666,
+      "step": 499
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.0002982,
+      "loss": 1.6613,
+      "step": 500
+    },
+    {
+      "epoch": 4.59,
+      "eval_cer": 0.25037856639529826,
+      "eval_loss": 0.8079274296760559,
+      "eval_runtime": 13.9777,
+      "eval_samples_per_second": 117.83,
+      "eval_steps_per_second": 1.86,
+      "eval_wer": 0.8541517720355428,
+      "step": 500
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0002988,
+      "loss": 1.7248,
+      "step": 501
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00029939999999999996,
+      "loss": 1.7596,
+      "step": 502
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0003,
+      "loss": 1.8184,
+      "step": 503
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0002999711538461538,
+      "loss": 1.6763,
+      "step": 504
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00029994230769230765,
+      "loss": 1.7465,
+      "step": 505
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00029991346153846154,
+      "loss": 1.728,
+      "step": 506
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0002998846153846153,
+      "loss": 1.6949,
+      "step": 507
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.0002998557692307692,
+      "loss": 1.7534,
+      "step": 508
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00029982692307692306,
+      "loss": 1.7506,
+      "step": 509
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0002997980769230769,
+      "loss": 1.8375,
+      "step": 510
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00029976923076923073,
+      "loss": 1.7,
+      "step": 511
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.00029974038461538457,
+      "loss": 1.782,
+      "step": 512
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0002997115384615384,
+      "loss": 1.6627,
+      "step": 513
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0002996826923076923,
+      "loss": 1.674,
+      "step": 514
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00029965384615384614,
+      "loss": 1.7381,
+      "step": 515
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.000299625,
+      "loss": 1.6171,
+      "step": 516
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0002995961538461538,
+      "loss": 1.6509,
+      "step": 517
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.00029956730769230765,
+      "loss": 1.6922,
+      "step": 518
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00029953846153846155,
+      "loss": 1.6117,
+      "step": 519
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00029950961538461533,
+      "loss": 1.588,
+      "step": 520
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0002994807692307692,
+      "loss": 1.7229,
+      "step": 521
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.00029945192307692306,
+      "loss": 1.6337,
+      "step": 522
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002994230769230769,
+      "loss": 1.631,
+      "step": 523
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00029939423076923074,
+      "loss": 1.6407,
+      "step": 524
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002993653846153846,
+      "loss": 1.664,
+      "step": 525
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00029933653846153847,
+      "loss": 1.6584,
+      "step": 526
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00029930769230769225,
+      "loss": 1.6223,
+      "step": 527
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.00029927884615384614,
+      "loss": 1.7282,
+      "step": 528
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00029925,
+      "loss": 1.8116,
+      "step": 529
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0002992211538461538,
+      "loss": 1.6944,
+      "step": 530
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00029919230769230766,
+      "loss": 1.6638,
+      "step": 531
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0002991634615384615,
+      "loss": 1.6906,
+      "step": 532
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00029913461538461534,
+      "loss": 1.6907,
+      "step": 533
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00029910576923076923,
+      "loss": 1.6264,
+      "step": 534
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00029907692307692307,
+      "loss": 1.8128,
+      "step": 535
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.0002990480769230769,
+      "loss": 1.7341,
+      "step": 536
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00029901923076923074,
+      "loss": 1.6639,
+      "step": 537
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002989903846153846,
+      "loss": 1.5123,
+      "step": 538
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002989615384615384,
+      "loss": 1.5523,
+      "step": 539
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00029893269230769226,
+      "loss": 1.6357,
+      "step": 540
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00029890384615384615,
+      "loss": 1.5318,
+      "step": 541
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000298875,
+      "loss": 1.6705,
+      "step": 542
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.0002988461538461538,
+      "loss": 1.6369,
+      "step": 543
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00029881730769230766,
+      "loss": 1.5031,
+      "step": 544
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.0002987884615384615,
+      "loss": 1.553,
+      "step": 545
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002987596153846154,
+      "loss": 1.5862,
+      "step": 546
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.0002987307692307692,
+      "loss": 1.6002,
+      "step": 547
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00029870192307692307,
+      "loss": 1.3398,
+      "step": 548
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.0002986730769230769,
+      "loss": 1.6936,
+      "step": 549
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.00029864423076923075,
+      "loss": 1.6268,
+      "step": 550
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002986153846153846,
+      "loss": 1.5878,
+      "step": 551
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002985865384615384,
+      "loss": 1.656,
+      "step": 552
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00029855769230769226,
+      "loss": 1.6005,
+      "step": 553
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0002985288461538461,
+      "loss": 1.5496,
+      "step": 554
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002985,
+      "loss": 1.6623,
+      "step": 555
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00029847115384615383,
+      "loss": 1.671,
+      "step": 556
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00029844230769230767,
+      "loss": 1.5461,
+      "step": 557
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.0002984134615384615,
+      "loss": 1.5917,
+      "step": 558
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00029838461538461534,
+      "loss": 1.5766,
+      "step": 559
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002983557692307692,
+      "loss": 1.7117,
+      "step": 560
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.0002983269230769231,
+      "loss": 1.5372,
+      "step": 561
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.0002982980769230769,
+      "loss": 1.5497,
+      "step": 562
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00029826923076923075,
+      "loss": 1.5617,
+      "step": 563
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.0002982403846153846,
+      "loss": 1.5948,
+      "step": 564
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00029821153846153843,
+      "loss": 1.5579,
+      "step": 565
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0002981826923076923,
+      "loss": 1.3991,
+      "step": 566
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0002981538461538461,
+      "loss": 1.5382,
+      "step": 567
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.000298125,
+      "loss": 1.6259,
+      "step": 568
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002980961538461538,
+      "loss": 1.5213,
+      "step": 569
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00029806730769230767,
+      "loss": 1.6134,
+      "step": 570
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0002980384615384615,
+      "loss": 1.5463,
+      "step": 571
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00029800961538461535,
+      "loss": 1.5587,
+      "step": 572
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.0002979807692307692,
+      "loss": 1.5212,
+      "step": 573
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.000297951923076923,
+      "loss": 1.6195,
+      "step": 574
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.0002979230769230769,
+      "loss": 1.5182,
+      "step": 575
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00029789423076923076,
+      "loss": 1.6271,
+      "step": 576
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0002978653846153846,
+      "loss": 1.6523,
+      "step": 577
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00029783653846153843,
+      "loss": 1.4375,
+      "step": 578
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00029780769230769227,
+      "loss": 1.4557,
+      "step": 579
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0002977788461538461,
+      "loss": 1.5103,
+      "step": 580
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00029775,
+      "loss": 1.487,
+      "step": 581
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.00029772115384615384,
+      "loss": 1.5735,
+      "step": 582
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0002976923076923077,
+      "loss": 1.5524,
+      "step": 583
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.0002976634615384615,
+      "loss": 1.6388,
+      "step": 584
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00029763461538461535,
+      "loss": 1.6018,
+      "step": 585
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00029760576923076925,
+      "loss": 1.5986,
+      "step": 586
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00029757692307692303,
+      "loss": 1.5045,
+      "step": 587
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002975480769230769,
+      "loss": 1.5155,
+      "step": 588
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.0002975192307692307,
+      "loss": 1.5342,
+      "step": 589
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.0002974903846153846,
+      "loss": 1.414,
+      "step": 590
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00029746153846153844,
+      "loss": 1.4848,
+      "step": 591
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0002974326923076923,
+      "loss": 1.4012,
+      "step": 592
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0002974038461538461,
+      "loss": 1.4671,
+      "step": 593
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00029737499999999995,
+      "loss": 1.5003,
+      "step": 594
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00029734615384615384,
+      "loss": 1.5382,
+      "step": 595
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002973173076923077,
+      "loss": 1.6466,
+      "step": 596
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.0002972884615384615,
+      "loss": 1.459,
+      "step": 597
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00029725961538461536,
+      "loss": 1.5353,
+      "step": 598
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002972307692307692,
+      "loss": 1.6523,
+      "step": 599
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00029720192307692303,
+      "loss": 1.525,
+      "step": 600
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00029717307692307687,
+      "loss": 1.5508,
+      "step": 601
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00029714423076923076,
+      "loss": 1.5993,
+      "step": 602
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0002971153846153846,
+      "loss": 1.5074,
+      "step": 603
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00029708653846153844,
+      "loss": 1.401,
+      "step": 604
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0002970576923076923,
+      "loss": 1.356,
+      "step": 605
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.0002970288461538461,
+      "loss": 1.5773,
+      "step": 606
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 1.6396,
+      "step": 607
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00029697115384615385,
+      "loss": 1.5741,
+      "step": 608
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00029694230769230763,
+      "loss": 1.6334,
+      "step": 609
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002969134615384615,
+      "loss": 1.5045,
+      "step": 610
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00029688461538461536,
+      "loss": 1.5342,
+      "step": 611
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0002968557692307692,
+      "loss": 1.5812,
+      "step": 612
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00029682692307692304,
+      "loss": 1.4745,
+      "step": 613
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0002967980769230769,
+      "loss": 1.4984,
+      "step": 614
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00029676923076923077,
+      "loss": 1.6008,
+      "step": 615
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00029674038461538455,
+      "loss": 1.5021,
+      "step": 616
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00029671153846153844,
+      "loss": 1.464,
+      "step": 617
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0002966826923076923,
+      "loss": 1.5561,
+      "step": 618
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0002966538461538461,
+      "loss": 1.5405,
+      "step": 619
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00029662499999999996,
+      "loss": 1.5319,
+      "step": 620
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0002965961538461538,
+      "loss": 1.4298,
+      "step": 621
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0002965673076923077,
+      "loss": 1.4598,
+      "step": 622
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00029653846153846153,
+      "loss": 1.4005,
+      "step": 623
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00029650961538461537,
+      "loss": 1.5438,
+      "step": 624
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0002964807692307692,
+      "loss": 1.4887,
+      "step": 625
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00029645192307692304,
+      "loss": 1.5223,
+      "step": 626
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002964230769230769,
+      "loss": 1.4489,
+      "step": 627
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.0002963942307692308,
+      "loss": 1.4061,
+      "step": 628
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.00029636538461538456,
+      "loss": 1.4445,
+      "step": 629
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00029633653846153845,
+      "loss": 1.6177,
+      "step": 630
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.0002963076923076923,
+      "loss": 1.4087,
+      "step": 631
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002962788461538461,
+      "loss": 1.4712,
+      "step": 632
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00029624999999999996,
+      "loss": 1.4733,
+      "step": 633
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0002962211538461538,
+      "loss": 1.5777,
+      "step": 634
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0002961923076923077,
+      "loss": 1.4219,
+      "step": 635
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0002961634615384615,
+      "loss": 1.5192,
+      "step": 636
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00029613461538461537,
+      "loss": 1.5435,
+      "step": 637
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.0002961057692307692,
+      "loss": 1.4818,
+      "step": 638
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00029607692307692305,
+      "loss": 1.4425,
+      "step": 639
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0002960480769230769,
+      "loss": 1.4976,
+      "step": 640
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0002960192307692307,
+      "loss": 1.4925,
+      "step": 641
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0002959903846153846,
+      "loss": 1.4853,
+      "step": 642
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.00029596153846153845,
+      "loss": 1.4376,
+      "step": 643
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0002959326923076923,
+      "loss": 1.5,
+      "step": 644
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00029590384615384613,
+      "loss": 1.5179,
+      "step": 645
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00029587499999999997,
+      "loss": 1.5648,
+      "step": 646
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002958461538461538,
+      "loss": 1.4738,
+      "step": 647
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002958173076923077,
+      "loss": 1.5028,
+      "step": 648
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.0002957884615384615,
+      "loss": 1.4813,
+      "step": 649
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002957596153846154,
+      "loss": 1.4424,
+      "step": 650
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.0002957307692307692,
+      "loss": 1.6667,
+      "step": 651
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00029570192307692305,
+      "loss": 1.4151,
+      "step": 652
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002956730769230769,
+      "loss": 1.4646,
+      "step": 653
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00029564423076923073,
+      "loss": 1.363,
+      "step": 654
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002956153846153846,
+      "loss": 1.4045,
+      "step": 655
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.0002955865384615384,
+      "loss": 1.4366,
+      "step": 656
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0002955576923076923,
+      "loss": 1.4319,
+      "step": 657
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.00029552884615384613,
+      "loss": 1.4812,
+      "step": 658
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00029549999999999997,
+      "loss": 1.5544,
+      "step": 659
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0002954711538461538,
+      "loss": 1.4726,
+      "step": 660
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00029544230769230765,
+      "loss": 1.5147,
+      "step": 661
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.00029541346153846154,
+      "loss": 1.4198,
+      "step": 662
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0002953846153846154,
+      "loss": 1.4731,
+      "step": 663
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0002953557692307692,
+      "loss": 1.4646,
+      "step": 664
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00029532692307692306,
+      "loss": 1.3274,
+      "step": 665
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.0002952980769230769,
+      "loss": 1.4136,
+      "step": 666
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00029526923076923073,
+      "loss": 1.2391,
+      "step": 667
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00029524038461538457,
+      "loss": 1.3782,
+      "step": 668
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002952115384615384,
+      "loss": 1.4194,
+      "step": 669
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0002951826923076923,
+      "loss": 1.4359,
+      "step": 670
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00029515384615384614,
+      "loss": 1.4706,
+      "step": 671
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.000295125,
+      "loss": 1.5051,
+      "step": 672
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002950961538461538,
+      "loss": 1.458,
+      "step": 673
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00029506730769230765,
+      "loss": 1.4158,
+      "step": 674
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00029503846153846155,
+      "loss": 1.39,
+      "step": 675
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00029500961538461533,
+      "loss": 1.4344,
+      "step": 676
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0002949807692307692,
+      "loss": 1.4228,
+      "step": 677
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00029495192307692306,
+      "loss": 1.2667,
+      "step": 678
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.0002949230769230769,
+      "loss": 1.4726,
+      "step": 679
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00029489423076923074,
+      "loss": 1.4229,
+      "step": 680
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0002948653846153846,
+      "loss": 1.3265,
+      "step": 681
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00029483653846153847,
+      "loss": 1.3744,
+      "step": 682
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00029480769230769225,
+      "loss": 1.3742,
+      "step": 683
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00029477884615384614,
+      "loss": 1.3577,
+      "step": 684
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00029475,
+      "loss": 1.3256,
+      "step": 685
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.0002947211538461538,
+      "loss": 1.509,
+      "step": 686
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00029469230769230766,
+      "loss": 1.3775,
+      "step": 687
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0002946634615384615,
+      "loss": 1.3985,
+      "step": 688
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00029463461538461533,
+      "loss": 1.5359,
+      "step": 689
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.0002946057692307692,
+      "loss": 1.4706,
+      "step": 690
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00029457692307692306,
+      "loss": 1.4827,
+      "step": 691
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0002945480769230769,
+      "loss": 1.4194,
+      "step": 692
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00029451923076923074,
+      "loss": 1.3666,
+      "step": 693
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0002944903846153846,
+      "loss": 1.2871,
+      "step": 694
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00029446153846153847,
+      "loss": 1.4429,
+      "step": 695
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00029443269230769226,
+      "loss": 1.4626,
+      "step": 696
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00029440384615384615,
+      "loss": 1.4164,
+      "step": 697
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.000294375,
+      "loss": 1.2492,
+      "step": 698
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0002943461538461538,
+      "loss": 1.3619,
+      "step": 699
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00029431730769230766,
+      "loss": 1.3744,
+      "step": 700
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002942884615384615,
+      "loss": 1.29,
+      "step": 701
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.0002942596153846154,
+      "loss": 1.4156,
+      "step": 702
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0002942307692307692,
+      "loss": 1.4037,
+      "step": 703
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00029420192307692307,
+      "loss": 1.3669,
+      "step": 704
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0002941730769230769,
+      "loss": 1.4027,
+      "step": 705
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00029414423076923075,
+      "loss": 1.4605,
+      "step": 706
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.0002941153846153846,
+      "loss": 1.3869,
+      "step": 707
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0002940865384615384,
+      "loss": 1.3774,
+      "step": 708
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00029405769230769226,
+      "loss": 1.425,
+      "step": 709
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00029402884615384615,
+      "loss": 1.3854,
+      "step": 710
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.000294,
+      "loss": 1.4403,
+      "step": 711
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00029397115384615383,
+      "loss": 1.363,
+      "step": 712
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00029394230769230767,
+      "loss": 1.4532,
+      "step": 713
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.0002939134615384615,
+      "loss": 1.4032,
+      "step": 714
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002938846153846154,
+      "loss": 1.6822,
+      "step": 715
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0002938557692307692,
+      "loss": 1.4484,
+      "step": 716
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002938269230769231,
+      "loss": 1.5052,
+      "step": 717
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.0002937980769230769,
+      "loss": 1.4604,
+      "step": 718
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00029376923076923075,
+      "loss": 1.4176,
+      "step": 719
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0002937403846153846,
+      "loss": 1.4905,
+      "step": 720
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0002937115384615384,
+      "loss": 1.3835,
+      "step": 721
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00029368269230769226,
+      "loss": 1.3405,
+      "step": 722
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0002936538461538461,
+      "loss": 1.4821,
+      "step": 723
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.000293625,
+      "loss": 1.3292,
+      "step": 724
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00029359615384615383,
+      "loss": 1.2677,
+      "step": 725
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00029356730769230767,
+      "loss": 1.2731,
+      "step": 726
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.0002935384615384615,
+      "loss": 1.4173,
+      "step": 727
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00029350961538461535,
+      "loss": 1.4918,
+      "step": 728
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0002934807692307692,
+      "loss": 1.2431,
+      "step": 729
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.0002934519230769231,
+      "loss": 1.5084,
+      "step": 730
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002934230769230769,
+      "loss": 1.3269,
+      "step": 731
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00029339423076923075,
+      "loss": 1.3175,
+      "step": 732
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0002933653846153846,
+      "loss": 1.3583,
+      "step": 733
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00029333653846153843,
+      "loss": 1.4524,
+      "step": 734
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00029330769230769227,
+      "loss": 1.4095,
+      "step": 735
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.0002932788461538461,
+      "loss": 1.4698,
+      "step": 736
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00029325,
+      "loss": 1.3867,
+      "step": 737
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00029322115384615384,
+      "loss": 1.3593,
+      "step": 738
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0002931923076923077,
+      "loss": 1.2626,
+      "step": 739
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0002931634615384615,
+      "loss": 1.4668,
+      "step": 740
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00029313461538461535,
+      "loss": 1.4897,
+      "step": 741
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0002931057692307692,
+      "loss": 1.4201,
+      "step": 742
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00029307692307692303,
+      "loss": 1.4401,
+      "step": 743
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0002930480769230769,
+      "loss": 1.2818,
+      "step": 744
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0002930192307692307,
+      "loss": 1.3727,
+      "step": 745
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002929903846153846,
+      "loss": 1.4991,
+      "step": 746
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00029296153846153844,
+      "loss": 1.3188,
+      "step": 747
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0002929326923076923,
+      "loss": 1.3824,
+      "step": 748
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0002929038461538461,
+      "loss": 1.3692,
+      "step": 749
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00029287499999999995,
+      "loss": 1.2925,
+      "step": 750
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00029284615384615384,
+      "loss": 1.406,
+      "step": 751
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0002928173076923077,
+      "loss": 1.4268,
+      "step": 752
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.0002927884615384615,
+      "loss": 1.4131,
+      "step": 753
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.00029275961538461536,
+      "loss": 1.4567,
+      "step": 754
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0002927307692307692,
+      "loss": 1.4445,
+      "step": 755
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00029270192307692303,
+      "loss": 1.3636,
+      "step": 756
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0002926730769230769,
+      "loss": 1.4568,
+      "step": 757
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00029264423076923076,
+      "loss": 1.2765,
+      "step": 758
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0002926153846153846,
+      "loss": 1.4719,
+      "step": 759
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00029258653846153844,
+      "loss": 1.3054,
+      "step": 760
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0002925576923076923,
+      "loss": 1.3302,
+      "step": 761
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002925288461538461,
+      "loss": 1.3211,
+      "step": 762
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00029249999999999995,
+      "loss": 1.3396,
+      "step": 763
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00029247115384615385,
+      "loss": 1.3485,
+      "step": 764
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00029244230769230763,
+      "loss": 1.2717,
+      "step": 765
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0002924134615384615,
+      "loss": 1.2487,
+      "step": 766
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00029238461538461536,
+      "loss": 1.3631,
+      "step": 767
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002923557692307692,
+      "loss": 1.4219,
+      "step": 768
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00029232692307692304,
+      "loss": 1.4171,
+      "step": 769
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0002922980769230769,
+      "loss": 1.3779,
+      "step": 770
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00029226923076923077,
+      "loss": 1.3191,
+      "step": 771
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0002922403846153846,
+      "loss": 1.2096,
+      "step": 772
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00029221153846153844,
+      "loss": 1.3383,
+      "step": 773
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.0002921826923076923,
+      "loss": 1.2846,
+      "step": 774
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.0002921538461538461,
+      "loss": 1.248,
+      "step": 775
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00029212499999999996,
+      "loss": 1.3903,
+      "step": 776
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00029209615384615385,
+      "loss": 1.2978,
+      "step": 777
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00029206730769230763,
+      "loss": 1.2456,
+      "step": 778
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00029203846153846153,
+      "loss": 1.3515,
+      "step": 779
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00029200961538461537,
+      "loss": 1.3162,
+      "step": 780
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0002919807692307692,
+      "loss": 1.2678,
+      "step": 781
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00029195192307692304,
+      "loss": 1.4755,
+      "step": 782
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.0002919230769230769,
+      "loss": 1.3638,
+      "step": 783
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00029189423076923077,
+      "loss": 1.3672,
+      "step": 784
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00029186538461538456,
+      "loss": 1.3074,
+      "step": 785
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00029183653846153845,
+      "loss": 1.2821,
+      "step": 786
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002918076923076923,
+      "loss": 1.3018,
+      "step": 787
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0002917788461538461,
+      "loss": 1.3658,
+      "step": 788
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00029174999999999996,
+      "loss": 1.3253,
+      "step": 789
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.0002917211538461538,
+      "loss": 1.204,
+      "step": 790
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.0002916923076923077,
+      "loss": 1.4094,
+      "step": 791
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00029166346153846153,
+      "loss": 1.3959,
+      "step": 792
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00029163461538461537,
+      "loss": 1.3628,
+      "step": 793
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0002916057692307692,
+      "loss": 1.3127,
+      "step": 794
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00029157692307692305,
+      "loss": 1.178,
+      "step": 795
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.0002915480769230769,
+      "loss": 1.3338,
+      "step": 796
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.0002915192307692307,
+      "loss": 1.292,
+      "step": 797
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00029149038461538456,
+      "loss": 1.5238,
+      "step": 798
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00029146153846153845,
+      "loss": 1.4193,
+      "step": 799
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.0002914326923076923,
+      "loss": 1.3347,
+      "step": 800
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00029140384615384613,
+      "loss": 1.2462,
+      "step": 801
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00029137499999999997,
+      "loss": 1.1992,
+      "step": 802
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0002913461538461538,
+      "loss": 1.2626,
+      "step": 803
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0002913173076923077,
+      "loss": 1.3837,
+      "step": 804
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0002912884615384615,
+      "loss": 1.3277,
+      "step": 805
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0002912596153846154,
+      "loss": 1.3608,
+      "step": 806
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.0002912307692307692,
+      "loss": 1.3721,
+      "step": 807
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00029120192307692305,
+      "loss": 1.4504,
+      "step": 808
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.0002911730769230769,
+      "loss": 1.3557,
+      "step": 809
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0002911442307692307,
+      "loss": 1.3381,
+      "step": 810
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.0002911153846153846,
+      "loss": 1.3528,
+      "step": 811
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.0002910865384615384,
+      "loss": 1.4158,
+      "step": 812
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0002910576923076923,
+      "loss": 1.3033,
+      "step": 813
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00029102884615384613,
+      "loss": 1.3297,
+      "step": 814
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 1.3106,
+      "step": 815
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.0002909711538461538,
+      "loss": 1.3534,
+      "step": 816
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00029094230769230765,
+      "loss": 1.2909,
+      "step": 817
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0002909134615384615,
+      "loss": 1.3281,
+      "step": 818
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.0002908846153846154,
+      "loss": 1.3572,
+      "step": 819
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.0002908557692307692,
+      "loss": 1.36,
+      "step": 820
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00029082692307692305,
+      "loss": 1.3212,
+      "step": 821
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.0002907980769230769,
+      "loss": 1.2619,
+      "step": 822
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00029076923076923073,
+      "loss": 1.3053,
+      "step": 823
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0002907403846153846,
+      "loss": 1.2908,
+      "step": 824
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.0002907115384615384,
+      "loss": 1.3345,
+      "step": 825
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.0002906826923076923,
+      "loss": 1.3565,
+      "step": 826
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00029065384615384614,
+      "loss": 1.2767,
+      "step": 827
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.000290625,
+      "loss": 1.3935,
+      "step": 828
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0002905961538461538,
+      "loss": 1.2747,
+      "step": 829
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00029056730769230765,
+      "loss": 1.3831,
+      "step": 830
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00029053846153846154,
+      "loss": 1.3785,
+      "step": 831
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00029050961538461533,
+      "loss": 1.2502,
+      "step": 832
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.0002904807692307692,
+      "loss": 1.3271,
+      "step": 833
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00029045192307692306,
+      "loss": 1.2771,
+      "step": 834
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0002904230769230769,
+      "loss": 1.2073,
+      "step": 835
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00029039423076923074,
+      "loss": 1.3386,
+      "step": 836
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.0002903653846153846,
+      "loss": 1.3454,
+      "step": 837
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0002903365384615384,
+      "loss": 1.4137,
+      "step": 838
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.0002903076923076923,
+      "loss": 1.2745,
+      "step": 839
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00029027884615384614,
+      "loss": 1.3313,
+      "step": 840
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00029025,
+      "loss": 1.1767,
+      "step": 841
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.0002902211538461538,
+      "loss": 1.3337,
+      "step": 842
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00029019230769230766,
+      "loss": 1.3239,
+      "step": 843
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00029016346153846155,
+      "loss": 1.2736,
+      "step": 844
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00029013461538461533,
+      "loss": 1.304,
+      "step": 845
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0002901057692307692,
+      "loss": 1.3404,
+      "step": 846
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00029007692307692306,
+      "loss": 1.2731,
+      "step": 847
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0002900480769230769,
+      "loss": 1.3777,
+      "step": 848
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00029001923076923074,
+      "loss": 1.4657,
+      "step": 849
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.0002899903846153846,
+      "loss": 1.1288,
+      "step": 850
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00028996153846153847,
+      "loss": 1.3401,
+      "step": 851
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00028993269230769225,
+      "loss": 1.2973,
+      "step": 852
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00028990384615384615,
+      "loss": 1.2779,
+      "step": 853
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.000289875,
+      "loss": 1.2831,
+      "step": 854
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0002898461538461538,
+      "loss": 1.1656,
+      "step": 855
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00028981730769230766,
+      "loss": 1.2714,
+      "step": 856
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0002897884615384615,
+      "loss": 1.3658,
+      "step": 857
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00028975961538461534,
+      "loss": 1.2487,
+      "step": 858
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00028973076923076923,
+      "loss": 1.2387,
+      "step": 859
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00028970192307692307,
+      "loss": 1.2637,
+      "step": 860
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0002896730769230769,
+      "loss": 1.2868,
+      "step": 861
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00028964423076923074,
+      "loss": 1.4895,
+      "step": 862
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0002896153846153846,
+      "loss": 1.2146,
+      "step": 863
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0002895865384615384,
+      "loss": 1.3045,
+      "step": 864
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00028955769230769226,
+      "loss": 1.2251,
+      "step": 865
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00028952884615384615,
+      "loss": 1.3366,
+      "step": 866
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.0002895,
+      "loss": 1.4156,
+      "step": 867
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00028947115384615383,
+      "loss": 1.2672,
+      "step": 868
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00028944230769230767,
+      "loss": 1.4122,
+      "step": 869
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.0002894134615384615,
+      "loss": 1.2516,
+      "step": 870
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0002893846153846154,
+      "loss": 1.2625,
+      "step": 871
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0002893557692307692,
+      "loss": 1.4866,
+      "step": 872
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00028932692307692307,
+      "loss": 1.29,
+      "step": 873
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00028929807692307686,
+      "loss": 1.2644,
+      "step": 874
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00028926923076923075,
+      "loss": 1.1984,
+      "step": 875
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0002892403846153846,
+      "loss": 1.2012,
+      "step": 876
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0002892115384615384,
+      "loss": 1.3409,
+      "step": 877
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00028918269230769226,
+      "loss": 1.2791,
+      "step": 878
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0002891538461538461,
+      "loss": 1.3249,
+      "step": 879
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.000289125,
+      "loss": 1.1443,
+      "step": 880
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00028909615384615383,
+      "loss": 1.1543,
+      "step": 881
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00028906730769230767,
+      "loss": 1.2783,
+      "step": 882
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.0002890384615384615,
+      "loss": 1.2759,
+      "step": 883
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.00028900961538461535,
+      "loss": 1.4482,
+      "step": 884
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0002889807692307692,
+      "loss": 1.2248,
+      "step": 885
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0002889519230769231,
+      "loss": 1.2432,
+      "step": 886
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0002889230769230769,
+      "loss": 1.2341,
+      "step": 887
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00028889423076923075,
+      "loss": 1.2214,
+      "step": 888
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.0002888653846153846,
+      "loss": 1.2412,
+      "step": 889
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00028883653846153843,
+      "loss": 1.4034,
+      "step": 890
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.0002888076923076923,
+      "loss": 1.3318,
+      "step": 891
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0002887788461538461,
+      "loss": 1.2337,
+      "step": 892
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00028875,
+      "loss": 1.2571,
+      "step": 893
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0002887211538461538,
+      "loss": 1.2379,
+      "step": 894
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.0002886923076923077,
+      "loss": 1.3764,
+      "step": 895
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.0002886634615384615,
+      "loss": 1.3149,
+      "step": 896
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00028863461538461535,
+      "loss": 1.2621,
+      "step": 897
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.0002886057692307692,
+      "loss": 1.2069,
+      "step": 898
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00028857692307692303,
+      "loss": 1.2142,
+      "step": 899
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0002885480769230769,
+      "loss": 1.2412,
+      "step": 900
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00028851923076923076,
+      "loss": 1.2007,
+      "step": 901
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0002884903846153846,
+      "loss": 1.3044,
+      "step": 902
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00028846153846153843,
+      "loss": 1.3186,
+      "step": 903
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00028843269230769227,
+      "loss": 1.2032,
+      "step": 904
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0002884038461538461,
+      "loss": 1.2233,
+      "step": 905
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.000288375,
+      "loss": 1.2558,
+      "step": 906
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00028834615384615384,
+      "loss": 1.2192,
+      "step": 907
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0002883173076923077,
+      "loss": 1.1428,
+      "step": 908
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0002882884615384615,
+      "loss": 1.1871,
+      "step": 909
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00028825961538461536,
+      "loss": 1.2998,
+      "step": 910
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.0002882307692307692,
+      "loss": 1.1298,
+      "step": 911
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00028820192307692303,
+      "loss": 1.2907,
+      "step": 912
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.0002881730769230769,
+      "loss": 1.1928,
+      "step": 913
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0002881442307692307,
+      "loss": 1.3972,
+      "step": 914
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0002881153846153846,
+      "loss": 1.2928,
+      "step": 915
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00028808653846153844,
+      "loss": 1.2991,
+      "step": 916
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0002880576923076923,
+      "loss": 1.458,
+      "step": 917
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.0002880288461538461,
+      "loss": 1.278,
+      "step": 918
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 1.2706,
+      "step": 919
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00028797115384615385,
+      "loss": 1.2624,
+      "step": 920
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.0002879423076923077,
+      "loss": 1.3111,
+      "step": 921
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0002879134615384615,
+      "loss": 1.1485,
+      "step": 922
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00028788461538461536,
+      "loss": 1.2443,
+      "step": 923
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0002878557692307692,
+      "loss": 1.1955,
+      "step": 924
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00028782692307692304,
+      "loss": 1.1974,
+      "step": 925
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0002877980769230769,
+      "loss": 1.214,
+      "step": 926
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00028776923076923077,
+      "loss": 1.3636,
+      "step": 927
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0002877403846153846,
+      "loss": 1.1999,
+      "step": 928
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00028771153846153844,
+      "loss": 1.3478,
+      "step": 929
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.0002876826923076923,
+      "loss": 1.2962,
+      "step": 930
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.0002876538461538461,
+      "loss": 1.3828,
+      "step": 931
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00028762499999999996,
+      "loss": 1.3609,
+      "step": 932
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00028759615384615385,
+      "loss": 1.2781,
+      "step": 933
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00028756730769230763,
+      "loss": 1.1763,
+      "step": 934
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0002875384615384615,
+      "loss": 1.3765,
+      "step": 935
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00028750961538461536,
+      "loss": 1.2783,
+      "step": 936
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0002874807692307692,
+      "loss": 1.1851,
+      "step": 937
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00028745192307692304,
+      "loss": 1.1783,
+      "step": 938
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0002874230769230769,
+      "loss": 1.2622,
+      "step": 939
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00028739423076923077,
+      "loss": 1.2226,
+      "step": 940
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00028736538461538456,
+      "loss": 1.2676,
+      "step": 941
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00028733653846153845,
+      "loss": 1.3196,
+      "step": 942
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.0002873076923076923,
+      "loss": 1.1977,
+      "step": 943
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.0002872788461538461,
+      "loss": 1.167,
+      "step": 944
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00028724999999999996,
+      "loss": 1.1849,
+      "step": 945
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.0002872211538461538,
+      "loss": 1.2489,
+      "step": 946
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0002871923076923077,
+      "loss": 1.419,
+      "step": 947
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00028716346153846153,
+      "loss": 1.2238,
+      "step": 948
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00028713461538461537,
+      "loss": 1.3082,
+      "step": 949
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0002871057692307692,
+      "loss": 1.2786,
+      "step": 950
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00028707692307692305,
+      "loss": 1.3634,
+      "step": 951
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.0002870480769230769,
+      "loss": 1.2459,
+      "step": 952
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0002870192307692308,
+      "loss": 1.3965,
+      "step": 953
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00028699038461538456,
+      "loss": 1.152,
+      "step": 954
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00028696153846153845,
+      "loss": 1.3515,
+      "step": 955
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0002869326923076923,
+      "loss": 1.2741,
+      "step": 956
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00028690384615384613,
+      "loss": 1.1794,
+      "step": 957
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00028687499999999997,
+      "loss": 1.3279,
+      "step": 958
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0002868461538461538,
+      "loss": 1.4241,
+      "step": 959
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0002868173076923077,
+      "loss": 1.305,
+      "step": 960
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0002867884615384615,
+      "loss": 1.2673,
+      "step": 961
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0002867596153846154,
+      "loss": 1.2169,
+      "step": 962
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0002867307692307692,
+      "loss": 1.3008,
+      "step": 963
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.00028670192307692305,
+      "loss": 1.2258,
+      "step": 964
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.0002866730769230769,
+      "loss": 1.2281,
+      "step": 965
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0002866442307692307,
+      "loss": 1.4098,
+      "step": 966
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.0002866153846153846,
+      "loss": 1.2768,
+      "step": 967
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00028658653846153846,
+      "loss": 1.1391,
+      "step": 968
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.0002865576923076923,
+      "loss": 1.4337,
+      "step": 969
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00028652884615384613,
+      "loss": 1.3497,
+      "step": 970
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00028649999999999997,
+      "loss": 1.3592,
+      "step": 971
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0002864711538461538,
+      "loss": 1.1369,
+      "step": 972
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.0002864423076923077,
+      "loss": 1.1792,
+      "step": 973
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0002864134615384615,
+      "loss": 1.3223,
+      "step": 974
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0002863846153846154,
+      "loss": 1.2912,
+      "step": 975
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0002863557692307692,
+      "loss": 1.3231,
+      "step": 976
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00028632692307692305,
+      "loss": 1.3137,
+      "step": 977
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.0002862980769230769,
+      "loss": 1.3661,
+      "step": 978
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00028626923076923073,
+      "loss": 1.2376,
+      "step": 979
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.0002862403846153846,
+      "loss": 1.2077,
+      "step": 980
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0002862115384615384,
+      "loss": 1.2522,
+      "step": 981
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.0002861826923076923,
+      "loss": 1.2712,
+      "step": 982
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00028615384615384614,
+      "loss": 1.1538,
+      "step": 983
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.000286125,
+      "loss": 1.1327,
+      "step": 984
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.0002860961538461538,
+      "loss": 1.0596,
+      "step": 985
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00028606730769230765,
+      "loss": 1.2553,
+      "step": 986
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0002860384615384615,
+      "loss": 1.3157,
+      "step": 987
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0002860096153846154,
+      "loss": 1.2561,
+      "step": 988
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.0002859807692307692,
+      "loss": 1.3399,
+      "step": 989
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00028595192307692306,
+      "loss": 1.0923,
+      "step": 990
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0002859230769230769,
+      "loss": 1.1767,
+      "step": 991
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00028589423076923073,
+      "loss": 1.1207,
+      "step": 992
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00028586538461538457,
+      "loss": 1.2896,
+      "step": 993
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.0002858365384615384,
+      "loss": 1.2776,
+      "step": 994
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0002858076923076923,
+      "loss": 1.2262,
+      "step": 995
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00028577884615384614,
+      "loss": 1.2506,
+      "step": 996
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00028575,
+      "loss": 1.2396,
+      "step": 997
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0002857211538461538,
+      "loss": 1.4892,
+      "step": 998
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00028569230769230766,
+      "loss": 1.3448,
+      "step": 999
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00028566346153846155,
+      "loss": 1.3496,
+      "step": 1000
+    },
+    {
+      "epoch": 9.17,
+      "eval_cer": 0.15183136178953208,
+      "eval_loss": 0.4729197323322296,
+      "eval_runtime": 13.9151,
+      "eval_samples_per_second": 118.361,
+      "eval_steps_per_second": 1.868,
+      "eval_wer": 0.5967725462159126,
+      "step": 1000
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00028563461538461533,
+      "loss": 1.3301,
+      "step": 1001
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0002856057692307692,
+      "loss": 1.2001,
+      "step": 1002
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00028557692307692306,
+      "loss": 1.2599,
+      "step": 1003
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.0002855480769230769,
+      "loss": 1.2045,
+      "step": 1004
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00028551923076923074,
+      "loss": 1.3028,
+      "step": 1005
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.0002854903846153846,
+      "loss": 1.2509,
+      "step": 1006
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0002854615384615384,
+      "loss": 1.2102,
+      "step": 1007
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00028543269230769225,
+      "loss": 1.2812,
+      "step": 1008
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00028540384615384615,
+      "loss": 1.2977,
+      "step": 1009
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.000285375,
+      "loss": 1.2507,
+      "step": 1010
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0002853461538461538,
+      "loss": 1.2743,
+      "step": 1011
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00028531730769230766,
+      "loss": 1.3867,
+      "step": 1012
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.0002852884615384615,
+      "loss": 1.3349,
+      "step": 1013
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00028525961538461534,
+      "loss": 1.2383,
+      "step": 1014
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00028523076923076923,
+      "loss": 1.1887,
+      "step": 1015
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00028520192307692307,
+      "loss": 1.2339,
+      "step": 1016
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.0002851730769230769,
+      "loss": 1.3013,
+      "step": 1017
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00028514423076923074,
+      "loss": 1.3717,
+      "step": 1018
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0002851153846153846,
+      "loss": 1.2454,
+      "step": 1019
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0002850865384615385,
+      "loss": 1.267,
+      "step": 1020
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00028505769230769226,
+      "loss": 1.1671,
+      "step": 1021
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00028502884615384615,
+      "loss": 1.2524,
+      "step": 1022
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.000285,
+      "loss": 1.2313,
+      "step": 1023
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0002849711538461538,
+      "loss": 1.3258,
+      "step": 1024
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00028494230769230766,
+      "loss": 1.2777,
+      "step": 1025
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0002849134615384615,
+      "loss": 1.2684,
+      "step": 1026
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00028488461538461534,
+      "loss": 1.3478,
+      "step": 1027
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.0002848557692307692,
+      "loss": 1.1854,
+      "step": 1028
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00028482692307692307,
+      "loss": 1.275,
+      "step": 1029
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.0002847980769230769,
+      "loss": 1.2289,
+      "step": 1030
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00028476923076923075,
+      "loss": 1.1886,
+      "step": 1031
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0002847403846153846,
+      "loss": 1.2139,
+      "step": 1032
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0002847115384615384,
+      "loss": 1.1703,
+      "step": 1033
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00028468269230769226,
+      "loss": 1.1887,
+      "step": 1034
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00028465384615384615,
+      "loss": 1.2108,
+      "step": 1035
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.000284625,
+      "loss": 1.3025,
+      "step": 1036
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00028459615384615383,
+      "loss": 1.4425,
+      "step": 1037
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00028456730769230767,
+      "loss": 1.2713,
+      "step": 1038
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0002845384615384615,
+      "loss": 1.1865,
+      "step": 1039
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.0002845096153846154,
+      "loss": 1.1847,
+      "step": 1040
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0002844807692307692,
+      "loss": 1.2571,
+      "step": 1041
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0002844519230769231,
+      "loss": 1.2435,
+      "step": 1042
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00028442307692307686,
+      "loss": 1.2977,
+      "step": 1043
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00028439423076923075,
+      "loss": 1.2277,
+      "step": 1044
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.0002843653846153846,
+      "loss": 1.3295,
+      "step": 1045
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00028433653846153843,
+      "loss": 1.1866,
+      "step": 1046
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00028430769230769227,
+      "loss": 1.2195,
+      "step": 1047
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0002842788461538461,
+      "loss": 1.2514,
+      "step": 1048
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00028425,
+      "loss": 1.1495,
+      "step": 1049
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00028422115384615384,
+      "loss": 1.3396,
+      "step": 1050
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0002841923076923077,
+      "loss": 1.1963,
+      "step": 1051
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.0002841634615384615,
+      "loss": 1.1978,
+      "step": 1052
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00028413461538461535,
+      "loss": 1.1479,
+      "step": 1053
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0002841057692307692,
+      "loss": 1.0622,
+      "step": 1054
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0002840769230769231,
+      "loss": 1.3145,
+      "step": 1055
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0002840480769230769,
+      "loss": 1.3506,
+      "step": 1056
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00028401923076923076,
+      "loss": 1.1392,
+      "step": 1057
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.0002839903846153846,
+      "loss": 1.3099,
+      "step": 1058
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00028396153846153843,
+      "loss": 1.3252,
+      "step": 1059
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00028393269230769227,
+      "loss": 1.1711,
+      "step": 1060
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.0002839038461538461,
+      "loss": 1.267,
+      "step": 1061
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.000283875,
+      "loss": 1.2163,
+      "step": 1062
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0002838461538461538,
+      "loss": 1.2111,
+      "step": 1063
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.0002838173076923077,
+      "loss": 1.1299,
+      "step": 1064
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0002837884615384615,
+      "loss": 1.3006,
+      "step": 1065
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00028375961538461535,
+      "loss": 1.3204,
+      "step": 1066
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.0002837307692307692,
+      "loss": 1.3015,
+      "step": 1067
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00028370192307692303,
+      "loss": 1.2675,
+      "step": 1068
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0002836730769230769,
+      "loss": 1.3065,
+      "step": 1069
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.0002836442307692307,
+      "loss": 1.2402,
+      "step": 1070
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0002836153846153846,
+      "loss": 1.1932,
+      "step": 1071
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00028358653846153844,
+      "loss": 1.3055,
+      "step": 1072
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0002835576923076923,
+      "loss": 1.1873,
+      "step": 1073
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0002835288461538461,
+      "loss": 1.1402,
+      "step": 1074
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00028349999999999995,
+      "loss": 1.3334,
+      "step": 1075
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00028347115384615384,
+      "loss": 1.2055,
+      "step": 1076
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0002834423076923077,
+      "loss": 1.2306,
+      "step": 1077
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.0002834134615384615,
+      "loss": 1.1293,
+      "step": 1078
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00028338461538461536,
+      "loss": 1.3256,
+      "step": 1079
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.0002833557692307692,
+      "loss": 1.2696,
+      "step": 1080
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00028332692307692304,
+      "loss": 1.3599,
+      "step": 1081
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00028329807692307693,
+      "loss": 1.2807,
+      "step": 1082
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0002832692307692307,
+      "loss": 1.3258,
+      "step": 1083
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0002832403846153846,
+      "loss": 1.2322,
+      "step": 1084
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00028321153846153844,
+      "loss": 1.2839,
+      "step": 1085
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0002831826923076923,
+      "loss": 1.2638,
+      "step": 1086
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.0002831538461538461,
+      "loss": 1.2511,
+      "step": 1087
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00028312499999999996,
+      "loss": 1.1554,
+      "step": 1088
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00028309615384615385,
+      "loss": 1.1554,
+      "step": 1089
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00028306730769230763,
+      "loss": 1.1883,
+      "step": 1090
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.0002830384615384615,
+      "loss": 1.186,
+      "step": 1091
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00028300961538461536,
+      "loss": 1.1227,
+      "step": 1092
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0002829807692307692,
+      "loss": 1.0908,
+      "step": 1093
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00028295192307692304,
+      "loss": 1.0399,
+      "step": 1094
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.0002829230769230769,
+      "loss": 1.2858,
+      "step": 1095
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00028289423076923077,
+      "loss": 1.1572,
+      "step": 1096
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0002828653846153846,
+      "loss": 1.2022,
+      "step": 1097
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00028283653846153845,
+      "loss": 1.3498,
+      "step": 1098
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.0002828076923076923,
+      "loss": 1.1944,
+      "step": 1099
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0002827788461538461,
+      "loss": 1.2493,
+      "step": 1100
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00028274999999999996,
+      "loss": 1.2221,
+      "step": 1101
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00028272115384615385,
+      "loss": 1.1961,
+      "step": 1102
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00028269230769230764,
+      "loss": 1.2412,
+      "step": 1103
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00028266346153846153,
+      "loss": 1.0861,
+      "step": 1104
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00028263461538461537,
+      "loss": 1.2181,
+      "step": 1105
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.0002826057692307692,
+      "loss": 1.2085,
+      "step": 1106
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00028257692307692304,
+      "loss": 1.1816,
+      "step": 1107
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0002825480769230769,
+      "loss": 1.2594,
+      "step": 1108
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0002825192307692308,
+      "loss": 1.2341,
+      "step": 1109
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00028249038461538456,
+      "loss": 1.1538,
+      "step": 1110
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00028246153846153845,
+      "loss": 1.0999,
+      "step": 1111
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0002824326923076923,
+      "loss": 1.19,
+      "step": 1112
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00028240384615384613,
+      "loss": 1.1618,
+      "step": 1113
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00028237499999999997,
+      "loss": 1.1559,
+      "step": 1114
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.0002823461538461538,
+      "loss": 1.2953,
+      "step": 1115
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0002823173076923077,
+      "loss": 1.3,
+      "step": 1116
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00028228846153846153,
+      "loss": 1.1969,
+      "step": 1117
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00028225961538461537,
+      "loss": 1.1942,
+      "step": 1118
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0002822307692307692,
+      "loss": 1.2549,
+      "step": 1119
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00028220192307692305,
+      "loss": 1.0386,
+      "step": 1120
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.0002821730769230769,
+      "loss": 1.5231,
+      "step": 1121
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0002821442307692307,
+      "loss": 1.3051,
+      "step": 1122
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00028211538461538456,
+      "loss": 1.1821,
+      "step": 1123
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00028208653846153846,
+      "loss": 1.2187,
+      "step": 1124
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0002820576923076923,
+      "loss": 1.132,
+      "step": 1125
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00028202884615384613,
+      "loss": 1.2078,
+      "step": 1126
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 1.1682,
+      "step": 1127
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.0002819711538461538,
+      "loss": 1.1968,
+      "step": 1128
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.0002819423076923077,
+      "loss": 1.2488,
+      "step": 1129
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0002819134615384615,
+      "loss": 1.0097,
+      "step": 1130
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.0002818846153846154,
+      "loss": 1.1679,
+      "step": 1131
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0002818557692307692,
+      "loss": 1.306,
+      "step": 1132
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00028182692307692305,
+      "loss": 1.1634,
+      "step": 1133
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0002817980769230769,
+      "loss": 1.2251,
+      "step": 1134
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00028176923076923073,
+      "loss": 1.1642,
+      "step": 1135
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0002817403846153846,
+      "loss": 1.1862,
+      "step": 1136
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.0002817115384615384,
+      "loss": 1.1732,
+      "step": 1137
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.0002816826923076923,
+      "loss": 1.2858,
+      "step": 1138
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00028165384615384614,
+      "loss": 1.1104,
+      "step": 1139
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.000281625,
+      "loss": 1.3046,
+      "step": 1140
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0002815961538461538,
+      "loss": 1.2306,
+      "step": 1141
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00028156730769230765,
+      "loss": 1.1982,
+      "step": 1142
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.0002815384615384615,
+      "loss": 1.2378,
+      "step": 1143
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0002815096153846154,
+      "loss": 1.1192,
+      "step": 1144
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0002814807692307692,
+      "loss": 1.115,
+      "step": 1145
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00028145192307692306,
+      "loss": 1.2076,
+      "step": 1146
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0002814230769230769,
+      "loss": 1.1275,
+      "step": 1147
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00028139423076923073,
+      "loss": 1.2669,
+      "step": 1148
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0002813653846153846,
+      "loss": 1.255,
+      "step": 1149
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0002813365384615384,
+      "loss": 1.0762,
+      "step": 1150
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.0002813076923076923,
+      "loss": 1.2402,
+      "step": 1151
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00028127884615384614,
+      "loss": 1.2675,
+      "step": 1152
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00028125,
+      "loss": 1.2499,
+      "step": 1153
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0002812211538461538,
+      "loss": 1.207,
+      "step": 1154
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00028119230769230766,
+      "loss": 1.1824,
+      "step": 1155
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00028116346153846155,
+      "loss": 1.1416,
+      "step": 1156
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00028113461538461533,
+      "loss": 1.2161,
+      "step": 1157
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0002811057692307692,
+      "loss": 1.3032,
+      "step": 1158
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00028107692307692306,
+      "loss": 1.1688,
+      "step": 1159
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.0002810480769230769,
+      "loss": 1.3109,
+      "step": 1160
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00028101923076923074,
+      "loss": 1.2551,
+      "step": 1161
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0002809903846153846,
+      "loss": 1.0786,
+      "step": 1162
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.0002809615384615384,
+      "loss": 1.3196,
+      "step": 1163
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.0002809326923076923,
+      "loss": 1.2473,
+      "step": 1164
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00028090384615384615,
+      "loss": 1.271,
+      "step": 1165
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.000280875,
+      "loss": 1.1296,
+      "step": 1166
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.0002808461538461538,
+      "loss": 1.1351,
+      "step": 1167
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00028081730769230766,
+      "loss": 1.1604,
+      "step": 1168
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00028078846153846155,
+      "loss": 1.2374,
+      "step": 1169
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00028075961538461534,
+      "loss": 1.1653,
+      "step": 1170
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00028073076923076923,
+      "loss": 1.4771,
+      "step": 1171
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00028070192307692307,
+      "loss": 1.2399,
+      "step": 1172
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.0002806730769230769,
+      "loss": 1.263,
+      "step": 1173
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00028064423076923074,
+      "loss": 1.1997,
+      "step": 1174
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.0002806153846153846,
+      "loss": 1.1041,
+      "step": 1175
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0002805865384615384,
+      "loss": 1.1089,
+      "step": 1176
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00028055769230769226,
+      "loss": 1.1477,
+      "step": 1177
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00028052884615384615,
+      "loss": 1.2502,
+      "step": 1178
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.0002805,
+      "loss": 1.1077,
+      "step": 1179
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0002804711538461538,
+      "loss": 1.0809,
+      "step": 1180
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00028044230769230766,
+      "loss": 1.1358,
+      "step": 1181
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0002804134615384615,
+      "loss": 1.2055,
+      "step": 1182
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00028038461538461534,
+      "loss": 1.2268,
+      "step": 1183
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00028035576923076923,
+      "loss": 1.2393,
+      "step": 1184
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00028032692307692307,
+      "loss": 1.2169,
+      "step": 1185
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0002802980769230769,
+      "loss": 1.1082,
+      "step": 1186
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00028026923076923075,
+      "loss": 1.233,
+      "step": 1187
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0002802403846153846,
+      "loss": 1.2751,
+      "step": 1188
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0002802115384615384,
+      "loss": 1.2755,
+      "step": 1189
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00028018269230769226,
+      "loss": 1.204,
+      "step": 1190
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00028015384615384615,
+      "loss": 1.1956,
+      "step": 1191
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.000280125,
+      "loss": 1.0832,
+      "step": 1192
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00028009615384615383,
+      "loss": 1.1517,
+      "step": 1193
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00028006730769230767,
+      "loss": 1.1084,
+      "step": 1194
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0002800384615384615,
+      "loss": 1.0822,
+      "step": 1195
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00028000961538461534,
+      "loss": 1.3157,
+      "step": 1196
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0002799807692307692,
+      "loss": 1.1852,
+      "step": 1197
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.0002799519230769231,
+      "loss": 1.383,
+      "step": 1198
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00027992307692307686,
+      "loss": 1.163,
+      "step": 1199
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00027989423076923075,
+      "loss": 1.2253,
+      "step": 1200
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0002798653846153846,
+      "loss": 1.15,
+      "step": 1201
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00027983653846153843,
+      "loss": 1.1218,
+      "step": 1202
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00027980769230769227,
+      "loss": 1.1439,
+      "step": 1203
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0002797788461538461,
+      "loss": 1.1716,
+      "step": 1204
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00027975,
+      "loss": 1.1998,
+      "step": 1205
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00027972115384615383,
+      "loss": 1.0032,
+      "step": 1206
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00027969230769230767,
+      "loss": 1.0598,
+      "step": 1207
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0002796634615384615,
+      "loss": 1.2737,
+      "step": 1208
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00027963461538461535,
+      "loss": 1.097,
+      "step": 1209
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.0002796057692307692,
+      "loss": 1.1647,
+      "step": 1210
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0002795769230769231,
+      "loss": 1.2657,
+      "step": 1211
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0002795480769230769,
+      "loss": 1.1407,
+      "step": 1212
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00027951923076923076,
+      "loss": 1.2446,
+      "step": 1213
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.0002794903846153846,
+      "loss": 1.2358,
+      "step": 1214
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00027946153846153843,
+      "loss": 1.1304,
+      "step": 1215
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00027943269230769227,
+      "loss": 1.0532,
+      "step": 1216
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0002794038461538461,
+      "loss": 1.1913,
+      "step": 1217
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.000279375,
+      "loss": 1.1897,
+      "step": 1218
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0002793461538461538,
+      "loss": 1.0879,
+      "step": 1219
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0002793173076923077,
+      "loss": 1.191,
+      "step": 1220
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.0002792884615384615,
+      "loss": 1.2127,
+      "step": 1221
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00027925961538461535,
+      "loss": 1.1426,
+      "step": 1222
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0002792307692307692,
+      "loss": 1.1591,
+      "step": 1223
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.00027920192307692303,
+      "loss": 1.1473,
+      "step": 1224
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0002791730769230769,
+      "loss": 1.2332,
+      "step": 1225
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00027914423076923076,
+      "loss": 1.1478,
+      "step": 1226
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0002791153846153846,
+      "loss": 1.0235,
+      "step": 1227
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00027908653846153844,
+      "loss": 1.1175,
+      "step": 1228
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0002790576923076923,
+      "loss": 1.2938,
+      "step": 1229
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0002790288461538461,
+      "loss": 1.2276,
+      "step": 1230
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.000279,
+      "loss": 1.192,
+      "step": 1231
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00027897115384615384,
+      "loss": 1.0843,
+      "step": 1232
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0002789423076923077,
+      "loss": 1.1374,
+      "step": 1233
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0002789134615384615,
+      "loss": 1.1798,
+      "step": 1234
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00027888461538461536,
+      "loss": 1.2542,
+      "step": 1235
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.0002788557692307692,
+      "loss": 1.2396,
+      "step": 1236
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00027882692307692303,
+      "loss": 1.1909,
+      "step": 1237
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0002787980769230769,
+      "loss": 1.1811,
+      "step": 1238
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0002787692307692307,
+      "loss": 1.0793,
+      "step": 1239
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0002787403846153846,
+      "loss": 1.1589,
+      "step": 1240
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00027871153846153844,
+      "loss": 1.2749,
+      "step": 1241
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0002786826923076923,
+      "loss": 1.3591,
+      "step": 1242
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.0002786538461538461,
+      "loss": 1.3127,
+      "step": 1243
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00027862499999999996,
+      "loss": 1.1059,
+      "step": 1244
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00027859615384615385,
+      "loss": 1.0991,
+      "step": 1245
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0002785673076923077,
+      "loss": 1.0599,
+      "step": 1246
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.0002785384615384615,
+      "loss": 1.2035,
+      "step": 1247
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00027850961538461536,
+      "loss": 1.3735,
+      "step": 1248
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0002784807692307692,
+      "loss": 1.3442,
+      "step": 1249
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00027845192307692304,
+      "loss": 1.0947,
+      "step": 1250
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.0002784230769230769,
+      "loss": 1.0881,
+      "step": 1251
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.0002783942307692307,
+      "loss": 1.2519,
+      "step": 1252
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0002783653846153846,
+      "loss": 1.2046,
+      "step": 1253
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00027833653846153845,
+      "loss": 1.1791,
+      "step": 1254
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.0002783076923076923,
+      "loss": 1.2302,
+      "step": 1255
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0002782788461538461,
+      "loss": 1.1257,
+      "step": 1256
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00027824999999999996,
+      "loss": 1.2165,
+      "step": 1257
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00027822115384615385,
+      "loss": 1.1526,
+      "step": 1258
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00027819230769230764,
+      "loss": 1.1991,
+      "step": 1259
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.00027816346153846153,
+      "loss": 1.2184,
+      "step": 1260
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00027813461538461537,
+      "loss": 1.1224,
+      "step": 1261
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0002781057692307692,
+      "loss": 1.1536,
+      "step": 1262
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00027807692307692304,
+      "loss": 1.1416,
+      "step": 1263
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0002780480769230769,
+      "loss": 1.1655,
+      "step": 1264
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0002780192307692308,
+      "loss": 1.2656,
+      "step": 1265
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00027799038461538456,
+      "loss": 1.0293,
+      "step": 1266
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00027796153846153845,
+      "loss": 1.2719,
+      "step": 1267
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.0002779326923076923,
+      "loss": 1.3421,
+      "step": 1268
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0002779038461538461,
+      "loss": 1.1874,
+      "step": 1269
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00027787499999999996,
+      "loss": 1.0847,
+      "step": 1270
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.0002778461538461538,
+      "loss": 1.0193,
+      "step": 1271
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00027781730769230764,
+      "loss": 1.2128,
+      "step": 1272
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00027778846153846153,
+      "loss": 1.127,
+      "step": 1273
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.00027775961538461537,
+      "loss": 1.125,
+      "step": 1274
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.0002777307692307692,
+      "loss": 1.1958,
+      "step": 1275
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00027770192307692305,
+      "loss": 1.18,
+      "step": 1276
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0002776730769230769,
+      "loss": 1.0689,
+      "step": 1277
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0002776442307692308,
+      "loss": 1.3012,
+      "step": 1278
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00027761538461538456,
+      "loss": 1.1331,
+      "step": 1279
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00027758653846153845,
+      "loss": 1.2968,
+      "step": 1280
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0002775576923076923,
+      "loss": 1.1244,
+      "step": 1281
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00027752884615384613,
+      "loss": 1.1224,
+      "step": 1282
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00027749999999999997,
+      "loss": 1.1193,
+      "step": 1283
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.0002774711538461538,
+      "loss": 1.1523,
+      "step": 1284
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0002774423076923077,
+      "loss": 1.2005,
+      "step": 1285
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0002774134615384615,
+      "loss": 1.1926,
+      "step": 1286
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.0002773846153846154,
+      "loss": 1.3036,
+      "step": 1287
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0002773557692307692,
+      "loss": 1.0813,
+      "step": 1288
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00027732692307692305,
+      "loss": 1.0717,
+      "step": 1289
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.0002772980769230769,
+      "loss": 1.0762,
+      "step": 1290
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00027726923076923073,
+      "loss": 1.2448,
+      "step": 1291
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.00027724038461538457,
+      "loss": 1.1696,
+      "step": 1292
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00027721153846153846,
+      "loss": 1.1776,
+      "step": 1293
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.0002771826923076923,
+      "loss": 1.1503,
+      "step": 1294
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00027715384615384614,
+      "loss": 1.1458,
+      "step": 1295
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.000277125,
+      "loss": 1.2879,
+      "step": 1296
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0002770961538461538,
+      "loss": 1.2141,
+      "step": 1297
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.0002770673076923077,
+      "loss": 1.1552,
+      "step": 1298
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.0002770384615384615,
+      "loss": 1.2169,
+      "step": 1299
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0002770096153846154,
+      "loss": 1.1666,
+      "step": 1300
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0002769807692307692,
+      "loss": 1.1629,
+      "step": 1301
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00027695192307692306,
+      "loss": 1.1789,
+      "step": 1302
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0002769230769230769,
+      "loss": 1.1994,
+      "step": 1303
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.00027689423076923073,
+      "loss": 1.2173,
+      "step": 1304
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.0002768653846153846,
+      "loss": 1.2243,
+      "step": 1305
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.0002768365384615384,
+      "loss": 1.2556,
+      "step": 1306
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.0002768076923076923,
+      "loss": 1.2683,
+      "step": 1307
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00027677884615384614,
+      "loss": 1.2549,
+      "step": 1308
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00027675,
+      "loss": 1.1741,
+      "step": 1309
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0002767211538461538,
+      "loss": 1.1129,
+      "step": 1310
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00027669230769230765,
+      "loss": 1.1058,
+      "step": 1311
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0002766634615384615,
+      "loss": 1.2437,
+      "step": 1312
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.0002766346153846154,
+      "loss": 1.2597,
+      "step": 1313
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.0002766057692307692,
+      "loss": 1.1523,
+      "step": 1314
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00027657692307692306,
+      "loss": 1.2897,
+      "step": 1315
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.0002765480769230769,
+      "loss": 1.1657,
+      "step": 1316
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00027651923076923074,
+      "loss": 0.9994,
+      "step": 1317
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0002764903846153846,
+      "loss": 1.1329,
+      "step": 1318
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0002764615384615384,
+      "loss": 1.1921,
+      "step": 1319
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0002764326923076923,
+      "loss": 1.1862,
+      "step": 1320
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00027640384615384614,
+      "loss": 1.1714,
+      "step": 1321
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.000276375,
+      "loss": 1.1228,
+      "step": 1322
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.0002763461538461538,
+      "loss": 1.0575,
+      "step": 1323
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00027631730769230766,
+      "loss": 1.2445,
+      "step": 1324
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00027628846153846155,
+      "loss": 1.1309,
+      "step": 1325
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00027625961538461533,
+      "loss": 1.2147,
+      "step": 1326
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00027623076923076923,
+      "loss": 1.2544,
+      "step": 1327
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.000276201923076923,
+      "loss": 1.262,
+      "step": 1328
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.0002761730769230769,
+      "loss": 1.0444,
+      "step": 1329
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00027614423076923074,
+      "loss": 0.9495,
+      "step": 1330
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0002761153846153846,
+      "loss": 1.0768,
+      "step": 1331
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0002760865384615384,
+      "loss": 1.1317,
+      "step": 1332
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00027605769230769226,
+      "loss": 1.2106,
+      "step": 1333
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00027602884615384615,
+      "loss": 1.1688,
+      "step": 1334
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.000276,
+      "loss": 1.1234,
+      "step": 1335
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.0002759711538461538,
+      "loss": 1.1428,
+      "step": 1336
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00027594230769230766,
+      "loss": 1.0946,
+      "step": 1337
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.0002759134615384615,
+      "loss": 1.2571,
+      "step": 1338
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00027588461538461534,
+      "loss": 1.1408,
+      "step": 1339
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.00027585576923076923,
+      "loss": 1.0818,
+      "step": 1340
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00027582692307692307,
+      "loss": 1.1408,
+      "step": 1341
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.0002757980769230769,
+      "loss": 1.1199,
+      "step": 1342
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00027576923076923075,
+      "loss": 1.116,
+      "step": 1343
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.0002757403846153846,
+      "loss": 1.1167,
+      "step": 1344
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.0002757115384615385,
+      "loss": 1.1997,
+      "step": 1345
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00027568269230769226,
+      "loss": 1.3034,
+      "step": 1346
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00027565384615384615,
+      "loss": 1.1998,
+      "step": 1347
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00027562499999999994,
+      "loss": 1.0628,
+      "step": 1348
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00027559615384615383,
+      "loss": 1.1094,
+      "step": 1349
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00027556730769230767,
+      "loss": 1.191,
+      "step": 1350
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.0002755384615384615,
+      "loss": 1.2202,
+      "step": 1351
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00027550961538461534,
+      "loss": 1.1871,
+      "step": 1352
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.0002754807692307692,
+      "loss": 1.2377,
+      "step": 1353
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.0002754519230769231,
+      "loss": 1.1028,
+      "step": 1354
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0002754230769230769,
+      "loss": 1.2526,
+      "step": 1355
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00027539423076923075,
+      "loss": 1.1911,
+      "step": 1356
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0002753653846153846,
+      "loss": 1.2679,
+      "step": 1357
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.0002753365384615384,
+      "loss": 1.2879,
+      "step": 1358
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00027530769230769227,
+      "loss": 1.1664,
+      "step": 1359
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00027527884615384616,
+      "loss": 1.1736,
+      "step": 1360
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00027525,
+      "loss": 1.1225,
+      "step": 1361
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00027522115384615383,
+      "loss": 1.2684,
+      "step": 1362
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00027519230769230767,
+      "loss": 1.0729,
+      "step": 1363
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.0002751634615384615,
+      "loss": 1.1446,
+      "step": 1364
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.0002751346153846154,
+      "loss": 1.1737,
+      "step": 1365
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.0002751057692307692,
+      "loss": 1.0807,
+      "step": 1366
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.0002750769230769231,
+      "loss": 1.1476,
+      "step": 1367
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00027504807692307686,
+      "loss": 1.1812,
+      "step": 1368
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00027501923076923075,
+      "loss": 1.1649,
+      "step": 1369
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.0002749903846153846,
+      "loss": 1.156,
+      "step": 1370
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00027496153846153843,
+      "loss": 1.1429,
+      "step": 1371
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00027493269230769227,
+      "loss": 1.0685,
+      "step": 1372
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.0002749038461538461,
+      "loss": 1.1782,
+      "step": 1373
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.000274875,
+      "loss": 1.1798,
+      "step": 1374
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00027484615384615384,
+      "loss": 1.0673,
+      "step": 1375
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.0002748173076923077,
+      "loss": 1.2214,
+      "step": 1376
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.0002747884615384615,
+      "loss": 1.2096,
+      "step": 1377
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00027475961538461535,
+      "loss": 0.9912,
+      "step": 1378
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.0002747307692307692,
+      "loss": 1.1982,
+      "step": 1379
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.0002747019230769231,
+      "loss": 1.105,
+      "step": 1380
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.0002746730769230769,
+      "loss": 1.1628,
+      "step": 1381
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00027464423076923076,
+      "loss": 1.2529,
+      "step": 1382
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0002746153846153846,
+      "loss": 1.2359,
+      "step": 1383
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00027458653846153844,
+      "loss": 1.2195,
+      "step": 1384
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.0002745576923076923,
+      "loss": 1.3499,
+      "step": 1385
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0002745288461538461,
+      "loss": 1.0706,
+      "step": 1386
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0002745,
+      "loss": 1.2262,
+      "step": 1387
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.0002744711538461538,
+      "loss": 1.1608,
+      "step": 1388
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.0002744423076923077,
+      "loss": 1.2781,
+      "step": 1389
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0002744134615384615,
+      "loss": 1.1076,
+      "step": 1390
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00027438461538461536,
+      "loss": 1.2019,
+      "step": 1391
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0002743557692307692,
+      "loss": 1.1009,
+      "step": 1392
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00027432692307692303,
+      "loss": 1.0777,
+      "step": 1393
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.0002742980769230769,
+      "loss": 1.2536,
+      "step": 1394
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.0002742692307692307,
+      "loss": 1.1329,
+      "step": 1395
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.0002742403846153846,
+      "loss": 1.0979,
+      "step": 1396
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00027421153846153844,
+      "loss": 1.122,
+      "step": 1397
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0002741826923076923,
+      "loss": 1.0845,
+      "step": 1398
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0002741538461538461,
+      "loss": 1.153,
+      "step": 1399
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00027412499999999995,
+      "loss": 1.2159,
+      "step": 1400
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.00027409615384615385,
+      "loss": 1.1878,
+      "step": 1401
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0002740673076923077,
+      "loss": 1.2022,
+      "step": 1402
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0002740384615384615,
+      "loss": 1.0803,
+      "step": 1403
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00027400961538461536,
+      "loss": 1.1606,
+      "step": 1404
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.0002739807692307692,
+      "loss": 1.1489,
+      "step": 1405
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00027395192307692304,
+      "loss": 1.1816,
+      "step": 1406
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00027392307692307693,
+      "loss": 1.2122,
+      "step": 1407
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.0002738942307692307,
+      "loss": 1.2605,
+      "step": 1408
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.0002738653846153846,
+      "loss": 1.1614,
+      "step": 1409
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00027383653846153844,
+      "loss": 1.171,
+      "step": 1410
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.0002738076923076923,
+      "loss": 1.1254,
+      "step": 1411
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.0002737788461538461,
+      "loss": 1.1885,
+      "step": 1412
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00027374999999999996,
+      "loss": 1.2559,
+      "step": 1413
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00027372115384615385,
+      "loss": 1.165,
+      "step": 1414
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00027369230769230764,
+      "loss": 1.1875,
+      "step": 1415
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00027366346153846153,
+      "loss": 1.1315,
+      "step": 1416
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00027363461538461537,
+      "loss": 1.1178,
+      "step": 1417
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.0002736057692307692,
+      "loss": 1.1037,
+      "step": 1418
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00027357692307692304,
+      "loss": 1.0889,
+      "step": 1419
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.0002735480769230769,
+      "loss": 1.2063,
+      "step": 1420
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00027351923076923077,
+      "loss": 1.0455,
+      "step": 1421
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.0002734903846153846,
+      "loss": 1.1411,
+      "step": 1422
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0002734903846153846,
+      "loss": 1.1524,
+      "step": 1423
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00027346153846153845,
+      "loss": 1.0737,
+      "step": 1424
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.0002734326923076923,
+      "loss": 1.0867,
+      "step": 1425
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.0002734038461538461,
+      "loss": 1.1343,
+      "step": 1426
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00027337499999999996,
+      "loss": 1.1524,
+      "step": 1427
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00027334615384615386,
+      "loss": 1.1203,
+      "step": 1428
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00027331730769230764,
+      "loss": 1.1345,
+      "step": 1429
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00027328846153846153,
+      "loss": 1.2476,
+      "step": 1430
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00027325961538461537,
+      "loss": 1.0594,
+      "step": 1431
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.0002732307692307692,
+      "loss": 1.0273,
+      "step": 1432
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00027320192307692305,
+      "loss": 1.2336,
+      "step": 1433
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.0002731730769230769,
+      "loss": 1.0302,
+      "step": 1434
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.0002731442307692308,
+      "loss": 1.1564,
+      "step": 1435
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00027311538461538456,
+      "loss": 1.0305,
+      "step": 1436
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00027308653846153845,
+      "loss": 1.1498,
+      "step": 1437
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.0002730576923076923,
+      "loss": 1.0518,
+      "step": 1438
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00027302884615384613,
+      "loss": 1.0156,
+      "step": 1439
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 1.2044,
+      "step": 1440
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.0002729711538461538,
+      "loss": 1.0884,
+      "step": 1441
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00027294230769230764,
+      "loss": 1.0809,
+      "step": 1442
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00027291346153846154,
+      "loss": 1.1039,
+      "step": 1443
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.0002728846153846154,
+      "loss": 1.2211,
+      "step": 1444
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.0002728557692307692,
+      "loss": 1.0613,
+      "step": 1445
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00027282692307692305,
+      "loss": 1.1357,
+      "step": 1446
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0002727980769230769,
+      "loss": 1.1676,
+      "step": 1447
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00027276923076923073,
+      "loss": 1.1244,
+      "step": 1448
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00027274038461538457,
+      "loss": 1.0561,
+      "step": 1449
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00027271153846153846,
+      "loss": 1.165,
+      "step": 1450
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.0002726826923076923,
+      "loss": 1.0872,
+      "step": 1451
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00027265384615384613,
+      "loss": 1.1538,
+      "step": 1452
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00027262499999999997,
+      "loss": 1.3029,
+      "step": 1453
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.0002725961538461538,
+      "loss": 1.1304,
+      "step": 1454
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0002725673076923077,
+      "loss": 1.0794,
+      "step": 1455
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.0002725384615384615,
+      "loss": 1.1283,
+      "step": 1456
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.0002725096153846154,
+      "loss": 1.1886,
+      "step": 1457
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.0002724807692307692,
+      "loss": 1.0293,
+      "step": 1458
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00027245192307692306,
+      "loss": 1.1207,
+      "step": 1459
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.0002724230769230769,
+      "loss": 1.1664,
+      "step": 1460
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00027239423076923073,
+      "loss": 1.3356,
+      "step": 1461
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00027236538461538457,
+      "loss": 1.1331,
+      "step": 1462
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.0002723365384615384,
+      "loss": 1.1649,
+      "step": 1463
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.0002723076923076923,
+      "loss": 1.1641,
+      "step": 1464
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00027227884615384614,
+      "loss": 1.1376,
+      "step": 1465
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00027225,
+      "loss": 1.188,
+      "step": 1466
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.0002722211538461538,
+      "loss": 1.1333,
+      "step": 1467
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00027219230769230765,
+      "loss": 1.1276,
+      "step": 1468
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.0002721634615384615,
+      "loss": 1.0392,
+      "step": 1469
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.0002721346153846154,
+      "loss": 1.1215,
+      "step": 1470
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0002721057692307692,
+      "loss": 1.2266,
+      "step": 1471
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00027207692307692306,
+      "loss": 1.0964,
+      "step": 1472
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.0002720480769230769,
+      "loss": 1.2801,
+      "step": 1473
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00027201923076923074,
+      "loss": 1.1573,
+      "step": 1474
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00027199038461538463,
+      "loss": 1.0791,
+      "step": 1475
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.0002719615384615384,
+      "loss": 1.0688,
+      "step": 1476
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.0002719326923076923,
+      "loss": 1.0866,
+      "step": 1477
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00027190384615384614,
+      "loss": 1.1334,
+      "step": 1478
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.000271875,
+      "loss": 1.2033,
+      "step": 1479
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.0002718461538461538,
+      "loss": 1.192,
+      "step": 1480
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00027181730769230766,
+      "loss": 1.1304,
+      "step": 1481
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.0002717884615384615,
+      "loss": 1.2075,
+      "step": 1482
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00027175961538461533,
+      "loss": 1.0361,
+      "step": 1483
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0002717307692307692,
+      "loss": 1.1487,
+      "step": 1484
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00027170192307692306,
+      "loss": 1.2372,
+      "step": 1485
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.0002716730769230769,
+      "loss": 1.1962,
+      "step": 1486
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00027164423076923074,
+      "loss": 1.1213,
+      "step": 1487
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0002716153846153846,
+      "loss": 1.1205,
+      "step": 1488
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.0002715865384615384,
+      "loss": 1.1569,
+      "step": 1489
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 0.0002715576923076923,
+      "loss": 1.1254,
+      "step": 1490
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.00027152884615384615,
+      "loss": 1.2436,
+      "step": 1491
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.0002715,
+      "loss": 1.2952,
+      "step": 1492
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 0.0002714711538461538,
+      "loss": 1.0937,
+      "step": 1493
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 0.00027144230769230766,
+      "loss": 1.1514,
+      "step": 1494
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.00027141346153846155,
+      "loss": 1.0813,
+      "step": 1495
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.00027138461538461534,
+      "loss": 1.1984,
+      "step": 1496
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 0.00027135576923076923,
+      "loss": 1.1638,
+      "step": 1497
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 0.00027132692307692307,
+      "loss": 1.2124,
+      "step": 1498
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.0002712980769230769,
+      "loss": 1.1007,
+      "step": 1499
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.00027126923076923075,
+      "loss": 1.1003,
+      "step": 1500
+    },
+    {
+      "epoch": 13.76,
+      "eval_cer": 0.135654207709261,
+      "eval_loss": 0.4106036126613617,
+      "eval_runtime": 13.8868,
+      "eval_samples_per_second": 118.602,
+      "eval_steps_per_second": 1.872,
+      "eval_wer": 0.5225206822592177,
+      "step": 1500
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 0.0002712403846153846,
+      "loss": 1.1401,
+      "step": 1501
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.0002712115384615384,
+      "loss": 0.9885,
+      "step": 1502
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 0.00027118269230769226,
+      "loss": 1.1127,
+      "step": 1503
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 0.00027115384615384615,
+      "loss": 1.2584,
+      "step": 1504
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.000271125,
+      "loss": 1.0434,
+      "step": 1505
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.00027109615384615383,
+      "loss": 1.1498,
+      "step": 1506
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.00027106730769230767,
+      "loss": 1.0426,
+      "step": 1507
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.0002710384615384615,
+      "loss": 1.1749,
+      "step": 1508
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.00027100961538461534,
+      "loss": 1.1459,
+      "step": 1509
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 0.00027098076923076924,
+      "loss": 1.2047,
+      "step": 1510
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.0002709519230769231,
+      "loss": 1.1953,
+      "step": 1511
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 0.0002709230769230769,
+      "loss": 0.996,
+      "step": 1512
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.00027089423076923075,
+      "loss": 1.1462,
+      "step": 1513
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.0002708653846153846,
+      "loss": 1.0823,
+      "step": 1514
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 0.0002708365384615384,
+      "loss": 1.1356,
+      "step": 1515
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.00027080769230769226,
+      "loss": 1.1636,
+      "step": 1516
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.00027077884615384616,
+      "loss": 1.1841,
+      "step": 1517
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 0.00027074999999999994,
+      "loss": 1.109,
+      "step": 1518
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00027072115384615383,
+      "loss": 1.1006,
+      "step": 1519
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00027069230769230767,
+      "loss": 1.08,
+      "step": 1520
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.0002706634615384615,
+      "loss": 1.0332,
+      "step": 1521
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 0.00027063461538461535,
+      "loss": 1.2198,
+      "step": 1522
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 0.0002706057692307692,
+      "loss": 1.1618,
+      "step": 1523
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 0.0002705769230769231,
+      "loss": 1.1977,
+      "step": 1524
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.00027054807692307686,
+      "loss": 1.0896,
+      "step": 1525
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.00027051923076923075,
+      "loss": 1.1629,
+      "step": 1526
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.0002704903846153846,
+      "loss": 1.0743,
+      "step": 1527
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.00027046153846153843,
+      "loss": 1.0712,
+      "step": 1528
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 0.00027043269230769227,
+      "loss": 1.0685,
+      "step": 1529
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 0.0002704038461538461,
+      "loss": 1.0773,
+      "step": 1530
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.000270375,
+      "loss": 1.1038,
+      "step": 1531
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.00027034615384615384,
+      "loss": 1.0812,
+      "step": 1532
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.0002703173076923077,
+      "loss": 1.1335,
+      "step": 1533
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 0.0002702884615384615,
+      "loss": 1.1376,
+      "step": 1534
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.00027025961538461535,
+      "loss": 1.0341,
+      "step": 1535
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 0.0002702307692307692,
+      "loss": 1.0777,
+      "step": 1536
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 0.0002702019230769231,
+      "loss": 0.9807,
+      "step": 1537
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 0.00027017307692307687,
+      "loss": 1.1058,
+      "step": 1538
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 0.00027014423076923076,
+      "loss": 1.0442,
+      "step": 1539
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 0.0002701153846153846,
+      "loss": 1.0861,
+      "step": 1540
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.00027008653846153843,
+      "loss": 1.1365,
+      "step": 1541
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 0.0002700576923076923,
+      "loss": 1.0193,
+      "step": 1542
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 0.0002700288461538461,
+      "loss": 1.0034,
+      "step": 1543
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.00027,
+      "loss": 1.1445,
+      "step": 1544
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.0002699711538461538,
+      "loss": 1.1088,
+      "step": 1545
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 0.0002699423076923077,
+      "loss": 1.0988,
+      "step": 1546
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 0.0002699134615384615,
+      "loss": 1.1426,
+      "step": 1547
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 0.00026988461538461536,
+      "loss": 1.1537,
+      "step": 1548
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0002698557692307692,
+      "loss": 1.0861,
+      "step": 1549
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.00026982692307692303,
+      "loss": 1.2524,
+      "step": 1550
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 0.0002697980769230769,
+      "loss": 1.1595,
+      "step": 1551
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.00026976923076923076,
+      "loss": 1.1879,
+      "step": 1552
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 0.0002697403846153846,
+      "loss": 1.1267,
+      "step": 1553
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 0.00026971153846153844,
+      "loss": 1.1853,
+      "step": 1554
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.0002696826923076923,
+      "loss": 1.0831,
+      "step": 1555
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.0002696538461538461,
+      "loss": 1.1988,
+      "step": 1556
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.000269625,
+      "loss": 1.1245,
+      "step": 1557
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 0.0002695961538461538,
+      "loss": 1.0919,
+      "step": 1558
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.0002695673076923077,
+      "loss": 1.1338,
+      "step": 1559
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 0.0002695384615384615,
+      "loss": 1.0885,
+      "step": 1560
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.00026950961538461536,
+      "loss": 1.0448,
+      "step": 1561
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 0.0002694807692307692,
+      "loss": 1.106,
+      "step": 1562
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.00026945192307692304,
+      "loss": 1.1729,
+      "step": 1563
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.00026942307692307693,
+      "loss": 1.1718,
+      "step": 1564
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 0.0002693942307692307,
+      "loss": 1.0443,
+      "step": 1565
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 0.0002693653846153846,
+      "loss": 1.0297,
+      "step": 1566
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 0.00026933653846153844,
+      "loss": 1.0891,
+      "step": 1567
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.0002693076923076923,
+      "loss": 1.0954,
+      "step": 1568
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.0002692788461538461,
+      "loss": 1.3179,
+      "step": 1569
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.00026924999999999996,
+      "loss": 1.1623,
+      "step": 1570
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.00026922115384615385,
+      "loss": 1.0222,
+      "step": 1571
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 0.0002691923076923077,
+      "loss": 1.0809,
+      "step": 1572
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.0002691634615384615,
+      "loss": 1.1652,
+      "step": 1573
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.00026913461538461536,
+      "loss": 1.0351,
+      "step": 1574
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 0.0002691057692307692,
+      "loss": 1.171,
+      "step": 1575
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 0.00026907692307692304,
+      "loss": 1.2782,
+      "step": 1576
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0002690480769230769,
+      "loss": 1.283,
+      "step": 1577
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 0.0002690192307692307,
+      "loss": 1.1352,
+      "step": 1578
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 0.0002689903846153846,
+      "loss": 1.0455,
+      "step": 1579
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.00026896153846153845,
+      "loss": 1.1516,
+      "step": 1580
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.0002689326923076923,
+      "loss": 1.2542,
+      "step": 1581
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.0002689038461538461,
+      "loss": 1.0717,
+      "step": 1582
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 0.00026887499999999996,
+      "loss": 1.1811,
+      "step": 1583
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 0.00026884615384615385,
+      "loss": 1.1034,
+      "step": 1584
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.00026881730769230764,
+      "loss": 1.0141,
+      "step": 1585
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.00026878846153846153,
+      "loss": 1.0995,
+      "step": 1586
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 0.00026875961538461537,
+      "loss": 1.1096,
+      "step": 1587
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.0002687307692307692,
+      "loss": 1.2255,
+      "step": 1588
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.00026870192307692305,
+      "loss": 1.1082,
+      "step": 1589
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 0.0002686730769230769,
+      "loss": 1.0609,
+      "step": 1590
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 0.0002686442307692308,
+      "loss": 1.1201,
+      "step": 1591
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.00026861538461538456,
+      "loss": 1.1689,
+      "step": 1592
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.00026858653846153845,
+      "loss": 1.192,
+      "step": 1593
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.0002685576923076923,
+      "loss": 1.1178,
+      "step": 1594
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 0.00026852884615384613,
+      "loss": 0.9871,
+      "step": 1595
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 0.00026849999999999997,
+      "loss": 1.1232,
+      "step": 1596
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 0.0002684711538461538,
+      "loss": 1.0108,
+      "step": 1597
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 0.00026844230769230764,
+      "loss": 1.0507,
+      "step": 1598
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 0.00026841346153846154,
+      "loss": 1.0386,
+      "step": 1599
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.0002683846153846154,
+      "loss": 1.185,
+      "step": 1600
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 0.0002683557692307692,
+      "loss": 1.3129,
+      "step": 1601
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.00026832692307692305,
+      "loss": 1.1663,
+      "step": 1602
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 0.0002682980769230769,
+      "loss": 1.0434,
+      "step": 1603
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.0002682692307692308,
+      "loss": 0.9928,
+      "step": 1604
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.00026824038461538456,
+      "loss": 1.1422,
+      "step": 1605
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.00026821153846153846,
+      "loss": 1.1334,
+      "step": 1606
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 0.0002681826923076923,
+      "loss": 1.1913,
+      "step": 1607
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 0.00026815384615384613,
+      "loss": 1.1018,
+      "step": 1608
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.00026812499999999997,
+      "loss": 1.1647,
+      "step": 1609
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.0002680961538461538,
+      "loss": 1.0574,
+      "step": 1610
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.0002680673076923077,
+      "loss": 1.096,
+      "step": 1611
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 0.0002680384615384615,
+      "loss": 1.073,
+      "step": 1612
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 0.0002680096153846154,
+      "loss": 1.1866,
+      "step": 1613
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.0002679807692307692,
+      "loss": 1.0567,
+      "step": 1614
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 0.00026795192307692305,
+      "loss": 1.1096,
+      "step": 1615
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.0002679230769230769,
+      "loss": 1.1928,
+      "step": 1616
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.00026789423076923073,
+      "loss": 1.084,
+      "step": 1617
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.00026786538461538457,
+      "loss": 1.1381,
+      "step": 1618
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 0.00026783653846153846,
+      "loss": 1.0253,
+      "step": 1619
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.0002678076923076923,
+      "loss": 1.1017,
+      "step": 1620
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 0.00026777884615384614,
+      "loss": 1.1494,
+      "step": 1621
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 0.00026775,
+      "loss": 1.074,
+      "step": 1622
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.0002677211538461538,
+      "loss": 1.1145,
+      "step": 1623
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 0.0002676923076923077,
+      "loss": 1.1366,
+      "step": 1624
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 0.0002676634615384615,
+      "loss": 1.2155,
+      "step": 1625
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.0002676346153846154,
+      "loss": 1.0617,
+      "step": 1626
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.0002676057692307692,
+      "loss": 1.0398,
+      "step": 1627
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.00026757692307692306,
+      "loss": 1.0293,
+      "step": 1628
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.0002675480769230769,
+      "loss": 1.0746,
+      "step": 1629
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 0.00026751923076923074,
+      "loss": 1.1262,
+      "step": 1630
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.00026749038461538463,
+      "loss": 1.1311,
+      "step": 1631
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.0002674615384615384,
+      "loss": 1.1841,
+      "step": 1632
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 0.0002674326923076923,
+      "loss": 1.0549,
+      "step": 1633
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 0.00026740384615384614,
+      "loss": 1.1956,
+      "step": 1634
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.000267375,
+      "loss": 1.1263,
+      "step": 1635
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 0.0002673461538461538,
+      "loss": 1.0872,
+      "step": 1636
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 0.00026731730769230766,
+      "loss": 1.0721,
+      "step": 1637
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.0002672884615384615,
+      "loss": 1.1142,
+      "step": 1638
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 0.0002672596153846154,
+      "loss": 1.0182,
+      "step": 1639
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 0.0002672307692307692,
+      "loss": 1.2242,
+      "step": 1640
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.00026720192307692306,
+      "loss": 1.0953,
+      "step": 1641
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.0002671730769230769,
+      "loss": 1.0538,
+      "step": 1642
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 0.00026714423076923074,
+      "loss": 1.0046,
+      "step": 1643
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 0.0002671153846153846,
+      "loss": 1.004,
+      "step": 1644
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 0.0002670865384615384,
+      "loss": 1.0934,
+      "step": 1645
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 0.0002670576923076923,
+      "loss": 1.1129,
+      "step": 1646
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.00026702884615384615,
+      "loss": 1.1381,
+      "step": 1647
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 0.000267,
+      "loss": 1.0923,
+      "step": 1648
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 0.0002669711538461538,
+      "loss": 1.0041,
+      "step": 1649
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.00026694230769230766,
+      "loss": 1.0283,
+      "step": 1650
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 0.0002669134615384615,
+      "loss": 1.0361,
+      "step": 1651
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.00026688461538461534,
+      "loss": 1.0941,
+      "step": 1652
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.00026685576923076923,
+      "loss": 1.091,
+      "step": 1653
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.000266826923076923,
+      "loss": 1.0226,
+      "step": 1654
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 0.0002667980769230769,
+      "loss": 1.0521,
+      "step": 1655
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.00026676923076923074,
+      "loss": 0.97,
+      "step": 1656
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 0.0002667403846153846,
+      "loss": 1.0531,
+      "step": 1657
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 0.0002667115384615384,
+      "loss": 0.9774,
+      "step": 1658
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.00026668269230769226,
+      "loss": 1.1102,
+      "step": 1659
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 0.00026665384615384615,
+      "loss": 1.0144,
+      "step": 1660
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.000266625,
+      "loss": 1.1359,
+      "step": 1661
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 0.00026659615384615383,
+      "loss": 1.0677,
+      "step": 1662
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.00026656730769230767,
+      "loss": 1.1297,
+      "step": 1663
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.0002665384615384615,
+      "loss": 1.2493,
+      "step": 1664
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.00026650961538461534,
+      "loss": 1.0643,
+      "step": 1665
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.00026648076923076923,
+      "loss": 1.2714,
+      "step": 1666
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 0.00026645192307692307,
+      "loss": 0.9796,
+      "step": 1667
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.0002664230769230769,
+      "loss": 1.0188,
+      "step": 1668
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 0.00026639423076923075,
+      "loss": 0.9624,
+      "step": 1669
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.0002663653846153846,
+      "loss": 1.1127,
+      "step": 1670
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.0002663365384615384,
+      "loss": 1.2697,
+      "step": 1671
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 0.00026630769230769226,
+      "loss": 1.0204,
+      "step": 1672
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 0.00026627884615384616,
+      "loss": 1.2538,
+      "step": 1673
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 0.00026624999999999994,
+      "loss": 1.1339,
+      "step": 1674
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.00026622115384615383,
+      "loss": 1.0855,
+      "step": 1675
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.00026619230769230767,
+      "loss": 0.983,
+      "step": 1676
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.0002661634615384615,
+      "loss": 1.1864,
+      "step": 1677
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.00026613461538461535,
+      "loss": 1.1467,
+      "step": 1678
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 0.0002661057692307692,
+      "loss": 1.0533,
+      "step": 1679
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.0002660769230769231,
+      "loss": 1.1012,
+      "step": 1680
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 0.0002660480769230769,
+      "loss": 1.0245,
+      "step": 1681
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.00026601923076923075,
+      "loss": 1.1404,
+      "step": 1682
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 0.0002659903846153846,
+      "loss": 1.1565,
+      "step": 1683
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 0.00026596153846153843,
+      "loss": 1.0746,
+      "step": 1684
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.00026593269230769227,
+      "loss": 1.2976,
+      "step": 1685
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 0.00026590384615384616,
+      "loss": 0.9856,
+      "step": 1686
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 0.000265875,
+      "loss": 1.0818,
+      "step": 1687
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.00026584615384615384,
+      "loss": 1.0826,
+      "step": 1688
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.0002658173076923077,
+      "loss": 1.0184,
+      "step": 1689
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.0002657884615384615,
+      "loss": 1.1158,
+      "step": 1690
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 0.00026575961538461535,
+      "loss": 1.2632,
+      "step": 1691
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.0002657307692307692,
+      "loss": 1.1219,
+      "step": 1692
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 0.0002657019230769231,
+      "loss": 1.0534,
+      "step": 1693
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 0.00026567307692307687,
+      "loss": 0.9824,
+      "step": 1694
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 0.00026564423076923076,
+      "loss": 1.0696,
+      "step": 1695
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.0002656153846153846,
+      "loss": 1.1289,
+      "step": 1696
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.00026558653846153843,
+      "loss": 1.1059,
+      "step": 1697
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 0.00026555769230769227,
+      "loss": 1.0054,
+      "step": 1698
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.0002655288461538461,
+      "loss": 1.106,
+      "step": 1699
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 0.0002655,
+      "loss": 1.0047,
+      "step": 1700
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.00026547115384615384,
+      "loss": 1.1314,
+      "step": 1701
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.0002654423076923077,
+      "loss": 1.1981,
+      "step": 1702
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 0.0002654134615384615,
+      "loss": 1.1076,
+      "step": 1703
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.00026538461538461536,
+      "loss": 1.1153,
+      "step": 1704
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 0.0002653557692307692,
+      "loss": 1.1508,
+      "step": 1705
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.0002653269230769231,
+      "loss": 1.1044,
+      "step": 1706
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.00026529807692307687,
+      "loss": 1.0892,
+      "step": 1707
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 0.00026526923076923076,
+      "loss": 1.1836,
+      "step": 1708
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.0002652403846153846,
+      "loss": 1.2626,
+      "step": 1709
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 0.00026521153846153844,
+      "loss": 1.1008,
+      "step": 1710
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.0002651826923076923,
+      "loss": 1.0512,
+      "step": 1711
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 0.0002651538461538461,
+      "loss": 1.1561,
+      "step": 1712
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.000265125,
+      "loss": 1.1338,
+      "step": 1713
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.0002650961538461538,
+      "loss": 1.0021,
+      "step": 1714
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.0002650673076923077,
+      "loss": 0.9853,
+      "step": 1715
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 0.0002650384615384615,
+      "loss": 1.1302,
+      "step": 1716
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 0.00026500961538461536,
+      "loss": 1.0497,
+      "step": 1717
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.0002649807692307692,
+      "loss": 1.0914,
+      "step": 1718
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 0.00026495192307692304,
+      "loss": 1.1253,
+      "step": 1719
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.00026492307692307693,
+      "loss": 1.199,
+      "step": 1720
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.0002648942307692307,
+      "loss": 1.1933,
+      "step": 1721
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 0.0002648653846153846,
+      "loss": 1.1097,
+      "step": 1722
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 0.00026483653846153844,
+      "loss": 1.1101,
+      "step": 1723
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.0002648076923076923,
+      "loss": 1.0546,
+      "step": 1724
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.0002647788461538461,
+      "loss": 1.0641,
+      "step": 1725
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.00026474999999999996,
+      "loss": 1.1667,
+      "step": 1726
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 0.0002647211538461538,
+      "loss": 1.098,
+      "step": 1727
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 0.0002646923076923077,
+      "loss": 1.1839,
+      "step": 1728
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.0002646634615384615,
+      "loss": 1.1686,
+      "step": 1729
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 0.00026463461538461536,
+      "loss": 1.1289,
+      "step": 1730
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 0.0002646057692307692,
+      "loss": 1.0536,
+      "step": 1731
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.00026457692307692304,
+      "loss": 1.1647,
+      "step": 1732
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 0.00026454807692307693,
+      "loss": 1.0984,
+      "step": 1733
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 0.0002645192307692307,
+      "loss": 1.0749,
+      "step": 1734
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.0002644903846153846,
+      "loss": 1.2883,
+      "step": 1735
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.00026446153846153845,
+      "loss": 1.193,
+      "step": 1736
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.0002644326923076923,
+      "loss": 1.1419,
+      "step": 1737
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.0002644038461538461,
+      "loss": 1.1158,
+      "step": 1738
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.00026437499999999996,
+      "loss": 1.1403,
+      "step": 1739
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 0.00026434615384615385,
+      "loss": 1.1103,
+      "step": 1740
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.00026431730769230764,
+      "loss": 1.1064,
+      "step": 1741
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 0.00026428846153846153,
+      "loss": 1.0477,
+      "step": 1742
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.00026425961538461537,
+      "loss": 1.1005,
+      "step": 1743
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 0.0002642307692307692,
+      "loss": 1.2071,
+      "step": 1744
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.00026420192307692304,
+      "loss": 1.1087,
+      "step": 1745
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 0.0002641730769230769,
+      "loss": 1.1223,
+      "step": 1746
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 0.0002641442307692307,
+      "loss": 1.0624,
+      "step": 1747
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 0.0002641153846153846,
+      "loss": 1.0101,
+      "step": 1748
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.00026408653846153845,
+      "loss": 1.1228,
+      "step": 1749
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.0002640576923076923,
+      "loss": 1.1713,
+      "step": 1750
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.00026402884615384613,
+      "loss": 1.133,
+      "step": 1751
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 1.106,
+      "step": 1752
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.00026397115384615386,
+      "loss": 0.9752,
+      "step": 1753
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 0.00026394230769230764,
+      "loss": 1.0479,
+      "step": 1754
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 0.00026391346153846153,
+      "loss": 1.1398,
+      "step": 1755
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.00026388461538461537,
+      "loss": 1.0564,
+      "step": 1756
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.0002638557692307692,
+      "loss": 0.9976,
+      "step": 1757
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 0.00026382692307692305,
+      "loss": 1.1508,
+      "step": 1758
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.0002637980769230769,
+      "loss": 1.0761,
+      "step": 1759
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 0.0002637692307692308,
+      "loss": 0.9689,
+      "step": 1760
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.00026374038461538456,
+      "loss": 1.0433,
+      "step": 1761
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.00026371153846153846,
+      "loss": 1.1666,
+      "step": 1762
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.0002636826923076923,
+      "loss": 1.1981,
+      "step": 1763
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.00026365384615384613,
+      "loss": 1.181,
+      "step": 1764
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.00026362499999999997,
+      "loss": 1.0355,
+      "step": 1765
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 0.0002635961538461538,
+      "loss": 1.1107,
+      "step": 1766
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 0.00026356730769230765,
+      "loss": 1.0696,
+      "step": 1767
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.00026353846153846154,
+      "loss": 1.1363,
+      "step": 1768
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 0.0002635096153846154,
+      "loss": 1.2918,
+      "step": 1769
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.0002634807692307692,
+      "loss": 1.0729,
+      "step": 1770
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.00026345192307692305,
+      "loss": 1.0058,
+      "step": 1771
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 0.0002634230769230769,
+      "loss": 1.0899,
+      "step": 1772
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.00026339423076923073,
+      "loss": 1.0737,
+      "step": 1773
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.00026336538461538457,
+      "loss": 1.0774,
+      "step": 1774
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.00026333653846153846,
+      "loss": 1.0725,
+      "step": 1775
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 0.0002633076923076923,
+      "loss": 1.2133,
+      "step": 1776
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 0.00026327884615384614,
+      "loss": 1.0573,
+      "step": 1777
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.00026325,
+      "loss": 1.018,
+      "step": 1778
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.0002632211538461538,
+      "loss": 1.1126,
+      "step": 1779
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 0.0002631923076923077,
+      "loss": 1.1052,
+      "step": 1780
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.0002631634615384615,
+      "loss": 1.081,
+      "step": 1781
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.0002631346153846154,
+      "loss": 0.9922,
+      "step": 1782
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 0.00026310576923076917,
+      "loss": 0.8907,
+      "step": 1783
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 0.00026307692307692306,
+      "loss": 0.9776,
+      "step": 1784
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.0002630480769230769,
+      "loss": 1.1562,
+      "step": 1785
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.00026301923076923073,
+      "loss": 1.0875,
+      "step": 1786
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.00026299038461538457,
+      "loss": 1.1667,
+      "step": 1787
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 0.0002629615384615384,
+      "loss": 1.0876,
+      "step": 1788
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.0002629326923076923,
+      "loss": 1.132,
+      "step": 1789
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.00026290384615384614,
+      "loss": 0.9458,
+      "step": 1790
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 0.000262875,
+      "loss": 1.0129,
+      "step": 1791
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 0.0002628461538461538,
+      "loss": 1.0533,
+      "step": 1792
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.00026281730769230766,
+      "loss": 1.0916,
+      "step": 1793
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 0.0002627884615384615,
+      "loss": 1.034,
+      "step": 1794
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 0.0002627596153846154,
+      "loss": 1.0423,
+      "step": 1795
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 0.0002627307692307692,
+      "loss": 0.9357,
+      "step": 1796
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.00026270192307692306,
+      "loss": 0.98,
+      "step": 1797
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.0002626730769230769,
+      "loss": 1.1382,
+      "step": 1798
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.00026264423076923074,
+      "loss": 1.0617,
+      "step": 1799
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.00026261538461538463,
+      "loss": 1.1888,
+      "step": 1800
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 0.0002625865384615384,
+      "loss": 1.1505,
+      "step": 1801
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 0.0002625576923076923,
+      "loss": 1.1689,
+      "step": 1802
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.0002625288461538461,
+      "loss": 1.0916,
+      "step": 1803
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 0.0002625,
+      "loss": 1.172,
+      "step": 1804
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 0.0002624711538461538,
+      "loss": 0.9824,
+      "step": 1805
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.00026244230769230766,
+      "loss": 0.9937,
+      "step": 1806
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 0.0002624134615384615,
+      "loss": 1.1626,
+      "step": 1807
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.00026238461538461534,
+      "loss": 1.061,
+      "step": 1808
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 0.00026235576923076923,
+      "loss": 1.0077,
+      "step": 1809
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.00026232692307692307,
+      "loss": 1.0787,
+      "step": 1810
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.0002622980769230769,
+      "loss": 1.163,
+      "step": 1811
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.00026226923076923074,
+      "loss": 1.1116,
+      "step": 1812
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 0.0002622403846153846,
+      "loss": 1.1456,
+      "step": 1813
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.0002622115384615384,
+      "loss": 1.1263,
+      "step": 1814
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.0002621826923076923,
+      "loss": 1.0352,
+      "step": 1815
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 0.00026215384615384615,
+      "loss": 1.0384,
+      "step": 1816
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 0.000262125,
+      "loss": 1.1247,
+      "step": 1817
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.0002620961538461538,
+      "loss": 1.1148,
+      "step": 1818
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 0.00026206730769230766,
+      "loss": 1.2799,
+      "step": 1819
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.00026203846153846156,
+      "loss": 1.0978,
+      "step": 1820
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 0.00026200961538461534,
+      "loss": 1.0941,
+      "step": 1821
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.00026198076923076923,
+      "loss": 1.0245,
+      "step": 1822
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.000261951923076923,
+      "loss": 0.9983,
+      "step": 1823
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 0.0002619230769230769,
+      "loss": 1.0867,
+      "step": 1824
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 0.00026189423076923075,
+      "loss": 1.2837,
+      "step": 1825
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.0002618653846153846,
+      "loss": 1.0752,
+      "step": 1826
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 0.0002618365384615384,
+      "loss": 1.0657,
+      "step": 1827
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 0.00026180769230769226,
+      "loss": 0.9216,
+      "step": 1828
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.00026177884615384615,
+      "loss": 1.1634,
+      "step": 1829
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.00026175,
+      "loss": 1.1016,
+      "step": 1830
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 0.00026172115384615383,
+      "loss": 1.0691,
+      "step": 1831
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.00026169230769230767,
+      "loss": 1.07,
+      "step": 1832
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 0.0002616634615384615,
+      "loss": 1.0912,
+      "step": 1833
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00026163461538461535,
+      "loss": 0.9722,
+      "step": 1834
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00026160576923076924,
+      "loss": 1.0394,
+      "step": 1835
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.0002615769230769231,
+      "loss": 1.0121,
+      "step": 1836
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 0.0002615480769230769,
+      "loss": 1.1554,
+      "step": 1837
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 0.00026151923076923075,
+      "loss": 1.2538,
+      "step": 1838
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.0002614903846153846,
+      "loss": 1.0202,
+      "step": 1839
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 0.00026146153846153843,
+      "loss": 1.0345,
+      "step": 1840
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 0.00026143269230769227,
+      "loss": 1.1186,
+      "step": 1841
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.00026140384615384616,
+      "loss": 1.1962,
+      "step": 1842
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 0.00026137499999999994,
+      "loss": 1.0499,
+      "step": 1843
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 0.00026134615384615384,
+      "loss": 1.1186,
+      "step": 1844
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 0.0002613173076923077,
+      "loss": 1.0396,
+      "step": 1845
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.0002612884615384615,
+      "loss": 1.1591,
+      "step": 1846
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.00026125961538461535,
+      "loss": 1.0513,
+      "step": 1847
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.0002612307692307692,
+      "loss": 0.9729,
+      "step": 1848
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 0.0002612019230769231,
+      "loss": 1.1129,
+      "step": 1849
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 0.00026117307692307686,
+      "loss": 1.1907,
+      "step": 1850
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 0.00026114423076923076,
+      "loss": 1.0121,
+      "step": 1851
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 0.0002611153846153846,
+      "loss": 1.0802,
+      "step": 1852
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 0.00026108653846153843,
+      "loss": 1.1779,
+      "step": 1853
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 0.00026105769230769227,
+      "loss": 0.9921,
+      "step": 1854
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 0.0002610288461538461,
+      "loss": 1.0464,
+      "step": 1855
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 0.000261,
+      "loss": 1.0295,
+      "step": 1856
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 0.00026097115384615384,
+      "loss": 1.0894,
+      "step": 1857
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.0002609423076923077,
+      "loss": 1.2458,
+      "step": 1858
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 0.0002609134615384615,
+      "loss": 1.087,
+      "step": 1859
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 0.00026088461538461535,
+      "loss": 1.0662,
+      "step": 1860
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 0.0002608557692307692,
+      "loss": 1.1646,
+      "step": 1861
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 0.0002608269230769231,
+      "loss": 1.2471,
+      "step": 1862
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 0.00026079807692307687,
+      "loss": 1.2024,
+      "step": 1863
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 0.00026076923076923076,
+      "loss": 1.0521,
+      "step": 1864
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 0.0002607403846153846,
+      "loss": 1.0128,
+      "step": 1865
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 0.00026071153846153844,
+      "loss": 1.0734,
+      "step": 1866
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 0.0002606826923076923,
+      "loss": 0.9704,
+      "step": 1867
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 0.0002606538461538461,
+      "loss": 0.9872,
+      "step": 1868
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 0.000260625,
+      "loss": 1.0556,
+      "step": 1869
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 0.0002605961538461538,
+      "loss": 0.9767,
+      "step": 1870
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.0002605673076923077,
+      "loss": 1.1437,
+      "step": 1871
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.0002605384615384615,
+      "loss": 1.184,
+      "step": 1872
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 0.00026050961538461536,
+      "loss": 1.1155,
+      "step": 1873
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 0.0002604807692307692,
+      "loss": 1.0071,
+      "step": 1874
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 0.00026045192307692303,
+      "loss": 1.0873,
+      "step": 1875
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 0.00026042307692307693,
+      "loss": 1.0298,
+      "step": 1876
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 0.00026039423076923077,
+      "loss": 1.1046,
+      "step": 1877
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 0.0002603653846153846,
+      "loss": 0.9434,
+      "step": 1878
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00026033653846153844,
+      "loss": 1.0591,
+      "step": 1879
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 0.0002603076923076923,
+      "loss": 1.0844,
+      "step": 1880
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 0.0002602788461538461,
+      "loss": 1.0927,
+      "step": 1881
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 0.00026025,
+      "loss": 1.137,
+      "step": 1882
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.0002602211538461538,
+      "loss": 1.0211,
+      "step": 1883
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.0002601923076923077,
+      "loss": 1.0877,
+      "step": 1884
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 0.0002601634615384615,
+      "loss": 1.0527,
+      "step": 1885
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 0.00026013461538461536,
+      "loss": 1.0591,
+      "step": 1886
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.0002601057692307692,
+      "loss": 0.8612,
+      "step": 1887
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 0.00026007692307692304,
+      "loss": 1.0477,
+      "step": 1888
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 0.00026004807692307693,
+      "loss": 1.1617,
+      "step": 1889
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 0.0002600192307692307,
+      "loss": 1.0493,
+      "step": 1890
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 0.0002599903846153846,
+      "loss": 1.1363,
+      "step": 1891
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 0.00025996153846153845,
+      "loss": 1.0454,
+      "step": 1892
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 0.0002599326923076923,
+      "loss": 1.1451,
+      "step": 1893
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 0.0002599038461538461,
+      "loss": 0.9511,
+      "step": 1894
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.00025987499999999996,
+      "loss": 1.1151,
+      "step": 1895
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.00025984615384615385,
+      "loss": 1.0239,
+      "step": 1896
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 0.0002598173076923077,
+      "loss": 1.0905,
+      "step": 1897
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 0.00025978846153846153,
+      "loss": 1.0859,
+      "step": 1898
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 0.00025975961538461537,
+      "loss": 1.0785,
+      "step": 1899
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.0002597307692307692,
+      "loss": 1.0271,
+      "step": 1900
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 0.00025970192307692304,
+      "loss": 0.9972,
+      "step": 1901
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 0.0002596730769230769,
+      "loss": 1.1333,
+      "step": 1902
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.0002596442307692307,
+      "loss": 0.9637,
+      "step": 1903
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 0.0002596153846153846,
+      "loss": 0.948,
+      "step": 1904
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 0.00025958653846153845,
+      "loss": 0.9961,
+      "step": 1905
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 0.0002595576923076923,
+      "loss": 1.057,
+      "step": 1906
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.0002595288461538461,
+      "loss": 1.0843,
+      "step": 1907
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.00025949999999999997,
+      "loss": 1.1042,
+      "step": 1908
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 0.00025947115384615386,
+      "loss": 1.1081,
+      "step": 1909
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 0.00025944230769230764,
+      "loss": 1.1205,
+      "step": 1910
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 0.00025941346153846153,
+      "loss": 1.0539,
+      "step": 1911
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.00025938461538461537,
+      "loss": 0.9988,
+      "step": 1912
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 0.0002593557692307692,
+      "loss": 1.124,
+      "step": 1913
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 0.00025932692307692305,
+      "loss": 1.0489,
+      "step": 1914
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 0.0002592980769230769,
+      "loss": 1.1463,
+      "step": 1915
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 0.0002592692307692307,
+      "loss": 1.0743,
+      "step": 1916
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 0.00025924038461538456,
+      "loss": 1.103,
+      "step": 1917
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 0.00025921153846153846,
+      "loss": 1.1273,
+      "step": 1918
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.0002591826923076923,
+      "loss": 0.994,
+      "step": 1919
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.00025915384615384613,
+      "loss": 1.0639,
+      "step": 1920
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 0.00025912499999999997,
+      "loss": 0.9569,
+      "step": 1921
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 0.0002590961538461538,
+      "loss": 1.1433,
+      "step": 1922
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 0.00025906730769230765,
+      "loss": 1.0343,
+      "step": 1923
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 0.00025903846153846154,
+      "loss": 1.064,
+      "step": 1924
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 0.0002590096153846154,
+      "loss": 1.0203,
+      "step": 1925
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 0.0002589807692307692,
+      "loss": 1.2711,
+      "step": 1926
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 0.00025895192307692305,
+      "loss": 1.2644,
+      "step": 1927
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 0.0002589230769230769,
+      "loss": 1.1352,
+      "step": 1928
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.0002588942307692308,
+      "loss": 0.9594,
+      "step": 1929
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 0.00025886538461538457,
+      "loss": 1.054,
+      "step": 1930
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.00025883653846153846,
+      "loss": 0.9826,
+      "step": 1931
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.0002588076923076923,
+      "loss": 1.0344,
+      "step": 1932
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.00025877884615384614,
+      "loss": 1.1072,
+      "step": 1933
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 0.00025875,
+      "loss": 1.148,
+      "step": 1934
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 0.0002587211538461538,
+      "loss": 1.0525,
+      "step": 1935
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.00025869230769230765,
+      "loss": 1.1441,
+      "step": 1936
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 0.0002586634615384615,
+      "loss": 0.9204,
+      "step": 1937
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 0.0002586346153846154,
+      "loss": 0.966,
+      "step": 1938
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 0.0002586057692307692,
+      "loss": 0.9399,
+      "step": 1939
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 0.00025857692307692306,
+      "loss": 1.1371,
+      "step": 1940
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 0.0002585480769230769,
+      "loss": 1.0456,
+      "step": 1941
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 0.00025851923076923073,
+      "loss": 1.0395,
+      "step": 1942
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.00025849038461538457,
+      "loss": 1.0735,
+      "step": 1943
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.00025846153846153846,
+      "loss": 1.0237,
+      "step": 1944
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.0002584326923076923,
+      "loss": 1.0979,
+      "step": 1945
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 0.00025840384615384614,
+      "loss": 1.1539,
+      "step": 1946
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 0.000258375,
+      "loss": 1.1105,
+      "step": 1947
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 0.0002583461538461538,
+      "loss": 0.9988,
+      "step": 1948
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 0.0002583173076923077,
+      "loss": 1.0376,
+      "step": 1949
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.0002582884615384615,
+      "loss": 1.0347,
+      "step": 1950
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 0.0002582596153846154,
+      "loss": 1.1115,
+      "step": 1951
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 0.0002582307692307692,
+      "loss": 1.0361,
+      "step": 1952
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 0.00025820192307692306,
+      "loss": 1.224,
+      "step": 1953
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 0.0002581730769230769,
+      "loss": 1.1304,
+      "step": 1954
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 0.00025814423076923074,
+      "loss": 1.0181,
+      "step": 1955
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 0.0002581153846153846,
+      "loss": 1.1691,
+      "step": 1956
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 0.0002580865384615384,
+      "loss": 1.1348,
+      "step": 1957
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 0.0002580576923076923,
+      "loss": 1.1316,
+      "step": 1958
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 0.00025802884615384614,
+      "loss": 1.1778,
+      "step": 1959
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 0.000258,
+      "loss": 1.1065,
+      "step": 1960
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 0.0002579711538461538,
+      "loss": 1.0395,
+      "step": 1961
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 0.00025794230769230766,
+      "loss": 1.1678,
+      "step": 1962
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 0.0002579134615384615,
+      "loss": 0.9825,
+      "step": 1963
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 0.0002578846153846154,
+      "loss": 0.8588,
+      "step": 1964
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.00025785576923076923,
+      "loss": 1.0864,
+      "step": 1965
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 0.00025782692307692307,
+      "loss": 1.0267,
+      "step": 1966
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 0.0002577980769230769,
+      "loss": 1.2579,
+      "step": 1967
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.00025776923076923074,
+      "loss": 1.0909,
+      "step": 1968
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.0002577403846153846,
+      "loss": 1.0774,
+      "step": 1969
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 0.0002577115384615384,
+      "loss": 1.064,
+      "step": 1970
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 0.0002576826923076923,
+      "loss": 0.9298,
+      "step": 1971
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 0.0002576538461538461,
+      "loss": 0.9129,
+      "step": 1972
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 0.000257625,
+      "loss": 0.979,
+      "step": 1973
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 0.0002575961538461538,
+      "loss": 1.0344,
+      "step": 1974
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 0.00025756730769230766,
+      "loss": 1.0662,
+      "step": 1975
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 0.0002575384615384615,
+      "loss": 1.0192,
+      "step": 1976
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 0.00025750961538461534,
+      "loss": 1.0103,
+      "step": 1977
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 0.00025748076923076923,
+      "loss": 1.0814,
+      "step": 1978
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.000257451923076923,
+      "loss": 0.9907,
+      "step": 1979
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.0002574230769230769,
+      "loss": 1.0722,
+      "step": 1980
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.00025739423076923075,
+      "loss": 1.1023,
+      "step": 1981
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 0.0002573653846153846,
+      "loss": 0.9553,
+      "step": 1982
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 0.0002573365384615384,
+      "loss": 1.104,
+      "step": 1983
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 0.00025730769230769226,
+      "loss": 1.0535,
+      "step": 1984
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 0.00025727884615384615,
+      "loss": 1.0851,
+      "step": 1985
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 0.00025725,
+      "loss": 1.0023,
+      "step": 1986
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 0.00025722115384615383,
+      "loss": 0.9845,
+      "step": 1987
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 0.00025719230769230767,
+      "loss": 1.084,
+      "step": 1988
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 0.0002571634615384615,
+      "loss": 1.13,
+      "step": 1989
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 0.00025713461538461534,
+      "loss": 0.9302,
+      "step": 1990
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 0.00025710576923076924,
+      "loss": 0.9495,
+      "step": 1991
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.000257076923076923,
+      "loss": 1.0355,
+      "step": 1992
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.0002570480769230769,
+      "loss": 1.0798,
+      "step": 1993
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.00025701923076923075,
+      "loss": 1.0708,
+      "step": 1994
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 0.0002569903846153846,
+      "loss": 1.0238,
+      "step": 1995
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 0.00025696153846153843,
+      "loss": 1.0003,
+      "step": 1996
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.00025693269230769227,
+      "loss": 1.0591,
+      "step": 1997
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 0.00025690384615384616,
+      "loss": 1.09,
+      "step": 1998
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 0.00025687499999999994,
+      "loss": 1.0959,
+      "step": 1999
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 0.00025684615384615383,
+      "loss": 1.1532,
+      "step": 2000
+    },
+    {
+      "epoch": 18.35,
+      "eval_cer": 0.12562407232492243,
+      "eval_loss": 0.3956577181816101,
+      "eval_runtime": 14.0272,
+      "eval_samples_per_second": 117.415,
+      "eval_steps_per_second": 1.854,
+      "eval_wer": 0.4978041058114595,
+      "step": 2000
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 0.00025681730769230767,
+      "loss": 0.9703,
+      "step": 2001
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 0.0002567884615384615,
+      "loss": 0.9783,
+      "step": 2002
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 0.00025675961538461535,
+      "loss": 0.9757,
+      "step": 2003
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.0002567307692307692,
+      "loss": 0.9851,
+      "step": 2004
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.0002567019230769231,
+      "loss": 1.1141,
+      "step": 2005
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 0.0002566730769230769,
+      "loss": 1.1051,
+      "step": 2006
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 0.00025664423076923076,
+      "loss": 1.0741,
+      "step": 2007
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 0.0002566153846153846,
+      "loss": 1.0939,
+      "step": 2008
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 0.00025658653846153843,
+      "loss": 1.058,
+      "step": 2009
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 0.00025655769230769227,
+      "loss": 1.1365,
+      "step": 2010
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 0.00025652884615384616,
+      "loss": 1.1621,
+      "step": 2011
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 0.00025649999999999995,
+      "loss": 1.0612,
+      "step": 2012
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 0.00025647115384615384,
+      "loss": 1.0138,
+      "step": 2013
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 0.0002564423076923077,
+      "loss": 0.9584,
+      "step": 2014
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 0.0002564134615384615,
+      "loss": 1.0187,
+      "step": 2015
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.00025638461538461535,
+      "loss": 1.0508,
+      "step": 2016
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.0002563557692307692,
+      "loss": 0.9923,
+      "step": 2017
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.0002563269230769231,
+      "loss": 1.1305,
+      "step": 2018
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 0.00025629807692307687,
+      "loss": 1.0779,
+      "step": 2019
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 0.00025626923076923076,
+      "loss": 0.8873,
+      "step": 2020
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 0.0002562403846153846,
+      "loss": 1.0218,
+      "step": 2021
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 0.00025621153846153844,
+      "loss": 1.0477,
+      "step": 2022
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 0.0002561826923076923,
+      "loss": 0.9662,
+      "step": 2023
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 0.0002561538461538461,
+      "loss": 1.0579,
+      "step": 2024
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 0.000256125,
+      "loss": 1.1121,
+      "step": 2025
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.00025609615384615384,
+      "loss": 1.1835,
+      "step": 2026
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 0.0002560673076923077,
+      "loss": 0.9851,
+      "step": 2027
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.0002560384615384615,
+      "loss": 1.2287,
+      "step": 2028
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.00025600961538461536,
+      "loss": 1.018,
+      "step": 2029
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.0002559807692307692,
+      "loss": 1.1407,
+      "step": 2030
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 0.0002559519230769231,
+      "loss": 1.2022,
+      "step": 2031
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 0.00025592307692307687,
+      "loss": 1.031,
+      "step": 2032
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 0.00025589423076923076,
+      "loss": 1.1179,
+      "step": 2033
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.0002558653846153846,
+      "loss": 1.0409,
+      "step": 2034
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 0.00025583653846153844,
+      "loss": 1.007,
+      "step": 2035
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 0.0002558076923076923,
+      "loss": 1.0498,
+      "step": 2036
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 0.0002557788461538461,
+      "loss": 1.1112,
+      "step": 2037
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 0.00025575,
+      "loss": 1.107,
+      "step": 2038
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 0.0002557211538461538,
+      "loss": 1.1029,
+      "step": 2039
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.0002556923076923077,
+      "loss": 1.1112,
+      "step": 2040
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.0002556634615384615,
+      "loss": 1.033,
+      "step": 2041
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 0.00025563461538461536,
+      "loss": 1.1153,
+      "step": 2042
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 0.0002556057692307692,
+      "loss": 1.0074,
+      "step": 2043
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 0.00025557692307692304,
+      "loss": 1.0649,
+      "step": 2044
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 0.00025554807692307693,
+      "loss": 1.0584,
+      "step": 2045
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 0.0002555192307692307,
+      "loss": 1.1195,
+      "step": 2046
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 0.0002554903846153846,
+      "loss": 1.037,
+      "step": 2047
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 0.00025546153846153845,
+      "loss": 0.8903,
+      "step": 2048
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 0.0002554326923076923,
+      "loss": 1.1047,
+      "step": 2049
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.0002554038461538461,
+      "loss": 1.0635,
+      "step": 2050
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 0.00025537499999999996,
+      "loss": 1.1636,
+      "step": 2051
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 0.0002553461538461538,
+      "loss": 1.0848,
+      "step": 2052
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 0.0002553173076923077,
+      "loss": 1.0613,
+      "step": 2053
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 0.00025528846153846153,
+      "loss": 1.1201,
+      "step": 2054
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 0.00025525961538461537,
+      "loss": 0.9547,
+      "step": 2055
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 0.0002552307692307692,
+      "loss": 1.1083,
+      "step": 2056
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 0.00025520192307692304,
+      "loss": 1.0189,
+      "step": 2057
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 0.00025517307692307694,
+      "loss": 0.9503,
+      "step": 2058
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.0002551442307692307,
+      "loss": 1.0372,
+      "step": 2059
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 0.0002551153846153846,
+      "loss": 1.0757,
+      "step": 2060
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 0.00025508653846153845,
+      "loss": 1.1562,
+      "step": 2061
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.0002550576923076923,
+      "loss": 1.0793,
+      "step": 2062
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 0.0002550288461538461,
+      "loss": 1.0973,
+      "step": 2063
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 1.1126,
+      "step": 2064
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.00025497115384615386,
+      "loss": 1.0121,
+      "step": 2065
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 0.00025494230769230764,
+      "loss": 1.1293,
+      "step": 2066
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 0.00025491346153846153,
+      "loss": 1.0485,
+      "step": 2067
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 0.00025488461538461537,
+      "loss": 1.0277,
+      "step": 2068
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 0.0002548557692307692,
+      "loss": 1.062,
+      "step": 2069
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 0.00025482692307692305,
+      "loss": 1.0084,
+      "step": 2070
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 0.0002547980769230769,
+      "loss": 0.9881,
+      "step": 2071
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 0.0002547692307692307,
+      "loss": 1.0186,
+      "step": 2072
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 0.0002547403846153846,
+      "loss": 1.1073,
+      "step": 2073
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 0.00025471153846153845,
+      "loss": 1.154,
+      "step": 2074
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 0.0002546826923076923,
+      "loss": 1.085,
+      "step": 2075
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 0.00025465384615384613,
+      "loss": 1.0468,
+      "step": 2076
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 0.00025462499999999997,
+      "loss": 1.1881,
+      "step": 2077
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 0.00025459615384615386,
+      "loss": 1.0856,
+      "step": 2078
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 0.00025456730769230764,
+      "loss": 1.0259,
+      "step": 2079
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 0.00025453846153846154,
+      "loss": 0.9425,
+      "step": 2080
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 0.0002545096153846154,
+      "loss": 0.9728,
+      "step": 2081
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 0.0002544807692307692,
+      "loss": 0.971,
+      "step": 2082
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.00025445192307692305,
+      "loss": 1.0376,
+      "step": 2083
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 0.0002544230769230769,
+      "loss": 0.954,
+      "step": 2084
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 0.0002543942307692308,
+      "loss": 0.9672,
+      "step": 2085
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 0.00025436538461538457,
+      "loss": 1.0163,
+      "step": 2086
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 0.00025433653846153846,
+      "loss": 0.9258,
+      "step": 2087
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 0.0002543076923076923,
+      "loss": 1.0388,
+      "step": 2088
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00025427884615384613,
+      "loss": 1.0928,
+      "step": 2089
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00025425,
+      "loss": 0.9599,
+      "step": 2090
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 0.0002542211538461538,
+      "loss": 1.0151,
+      "step": 2091
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 0.00025419230769230765,
+      "loss": 0.9777,
+      "step": 2092
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 0.00025416346153846154,
+      "loss": 1.0121,
+      "step": 2093
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 0.0002541346153846154,
+      "loss": 1.1512,
+      "step": 2094
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.0002541057692307692,
+      "loss": 1.0851,
+      "step": 2095
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 0.00025407692307692306,
+      "loss": 1.0644,
+      "step": 2096
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 0.0002540480769230769,
+      "loss": 0.9613,
+      "step": 2097
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 0.00025401923076923073,
+      "loss": 1.087,
+      "step": 2098
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 0.00025399038461538457,
+      "loss": 1.0778,
+      "step": 2099
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 0.00025396153846153846,
+      "loss": 0.9762,
+      "step": 2100
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.0002539326923076923,
+      "loss": 1.0896,
+      "step": 2101
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.00025390384615384614,
+      "loss": 1.0925,
+      "step": 2102
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 0.000253875,
+      "loss": 0.9813,
+      "step": 2103
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.0002538461538461538,
+      "loss": 1.0354,
+      "step": 2104
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 0.0002538173076923077,
+      "loss": 0.9593,
+      "step": 2105
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 0.0002537884615384615,
+      "loss": 1.0467,
+      "step": 2106
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 0.0002537596153846154,
+      "loss": 1.0698,
+      "step": 2107
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 0.00025373076923076917,
+      "loss": 1.0887,
+      "step": 2108
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 0.00025370192307692306,
+      "loss": 1.0331,
+      "step": 2109
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 0.0002536730769230769,
+      "loss": 1.0898,
+      "step": 2110
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 0.00025364423076923074,
+      "loss": 1.0897,
+      "step": 2111
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 0.0002536153846153846,
+      "loss": 1.0122,
+      "step": 2112
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 0.0002535865384615384,
+      "loss": 1.0567,
+      "step": 2113
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 0.0002535576923076923,
+      "loss": 1.0508,
+      "step": 2114
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 0.00025352884615384614,
+      "loss": 0.9558,
+      "step": 2115
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.0002535,
+      "loss": 0.8954,
+      "step": 2116
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 0.0002534711538461538,
+      "loss": 1.0178,
+      "step": 2117
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 0.00025344230769230766,
+      "loss": 1.0386,
+      "step": 2118
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 0.0002534134615384615,
+      "loss": 1.0077,
+      "step": 2119
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 0.0002533846153846154,
+      "loss": 1.0925,
+      "step": 2120
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 0.0002533557692307692,
+      "loss": 1.0702,
+      "step": 2121
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 0.00025332692307692306,
+      "loss": 1.073,
+      "step": 2122
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 0.0002532980769230769,
+      "loss": 0.967,
+      "step": 2123
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 0.00025326923076923074,
+      "loss": 1.0996,
+      "step": 2124
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.0002532403846153846,
+      "loss": 0.9936,
+      "step": 2125
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.0002532115384615384,
+      "loss": 1.1718,
+      "step": 2126
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 0.0002531826923076923,
+      "loss": 1.1424,
+      "step": 2127
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 0.0002531538461538461,
+      "loss": 1.0656,
+      "step": 2128
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 0.000253125,
+      "loss": 0.925,
+      "step": 2129
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 0.0002530961538461538,
+      "loss": 1.1236,
+      "step": 2130
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 0.00025306730769230766,
+      "loss": 0.9576,
+      "step": 2131
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 0.0002530384615384615,
+      "loss": 1.0386,
+      "step": 2132
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 0.00025300961538461534,
+      "loss": 1.0474,
+      "step": 2133
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 0.00025298076923076923,
+      "loss": 0.9848,
+      "step": 2134
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 0.00025295192307692307,
+      "loss": 1.0427,
+      "step": 2135
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 0.0002529230769230769,
+      "loss": 0.97,
+      "step": 2136
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.00025289423076923075,
+      "loss": 1.0434,
+      "step": 2137
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.0002528653846153846,
+      "loss": 1.1548,
+      "step": 2138
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 0.0002528365384615384,
+      "loss": 1.1979,
+      "step": 2139
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 0.0002528076923076923,
+      "loss": 0.9508,
+      "step": 2140
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 0.00025277884615384615,
+      "loss": 1.1082,
+      "step": 2141
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 0.00025275,
+      "loss": 1.0382,
+      "step": 2142
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 0.00025272115384615383,
+      "loss": 0.9539,
+      "step": 2143
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.00025269230769230767,
+      "loss": 1.171,
+      "step": 2144
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 0.0002526634615384615,
+      "loss": 1.0507,
+      "step": 2145
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 0.00025263461538461534,
+      "loss": 1.1148,
+      "step": 2146
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 0.00025260576923076924,
+      "loss": 0.9653,
+      "step": 2147
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 0.000252576923076923,
+      "loss": 1.0383,
+      "step": 2148
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 0.0002525480769230769,
+      "loss": 1.0085,
+      "step": 2149
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 0.00025251923076923075,
+      "loss": 1.0447,
+      "step": 2150
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 0.0002524903846153846,
+      "loss": 1.0114,
+      "step": 2151
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 0.0002524615384615384,
+      "loss": 1.2098,
+      "step": 2152
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 0.00025243269230769226,
+      "loss": 1.0541,
+      "step": 2153
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 0.00025240384615384616,
+      "loss": 1.0,
+      "step": 2154
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 0.000252375,
+      "loss": 1.0451,
+      "step": 2155
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.00025234615384615383,
+      "loss": 1.0867,
+      "step": 2156
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 0.00025231730769230767,
+      "loss": 1.0787,
+      "step": 2157
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 0.0002522884615384615,
+      "loss": 1.0777,
+      "step": 2158
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 0.00025225961538461535,
+      "loss": 0.8751,
+      "step": 2159
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 0.00025223076923076924,
+      "loss": 0.9201,
+      "step": 2160
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.0002522019230769231,
+      "loss": 1.1334,
+      "step": 2161
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.0002521730769230769,
+      "loss": 1.1007,
+      "step": 2162
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 0.00025214423076923075,
+      "loss": 1.0639,
+      "step": 2163
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 0.0002521153846153846,
+      "loss": 0.8941,
+      "step": 2164
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 0.00025208653846153843,
+      "loss": 1.1408,
+      "step": 2165
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 0.00025205769230769227,
+      "loss": 1.1391,
+      "step": 2166
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 0.00025202884615384616,
+      "loss": 1.067,
+      "step": 2167
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 0.9291,
+      "step": 2168
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 0.00025197115384615384,
+      "loss": 1.0838,
+      "step": 2169
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 0.0002519423076923077,
+      "loss": 1.14,
+      "step": 2170
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 0.0002519134615384615,
+      "loss": 1.2505,
+      "step": 2171
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.00025188461538461535,
+      "loss": 0.9432,
+      "step": 2172
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.0002518557692307692,
+      "loss": 0.9933,
+      "step": 2173
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.0002518269230769231,
+      "loss": 1.0047,
+      "step": 2174
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 0.00025179807692307687,
+      "loss": 1.1222,
+      "step": 2175
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 0.00025176923076923076,
+      "loss": 1.1224,
+      "step": 2176
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.0002517403846153846,
+      "loss": 1.061,
+      "step": 2177
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 0.00025171153846153844,
+      "loss": 0.966,
+      "step": 2178
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 0.0002516826923076923,
+      "loss": 1.0342,
+      "step": 2179
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.0002516538461538461,
+      "loss": 1.1469,
+      "step": 2180
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 0.00025162499999999995,
+      "loss": 1.0495,
+      "step": 2181
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 0.00025159615384615384,
+      "loss": 0.997,
+      "step": 2182
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 0.0002515673076923077,
+      "loss": 0.9632,
+      "step": 2183
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 0.0002515384615384615,
+      "loss": 1.0417,
+      "step": 2184
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 0.00025150961538461536,
+      "loss": 1.2122,
+      "step": 2185
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.0002514807692307692,
+      "loss": 1.1547,
+      "step": 2186
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.0002514519230769231,
+      "loss": 1.0236,
+      "step": 2187
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 0.00025142307692307687,
+      "loss": 1.0997,
+      "step": 2188
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 0.00025139423076923076,
+      "loss": 1.0335,
+      "step": 2189
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 0.0002513653846153846,
+      "loss": 1.0485,
+      "step": 2190
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 0.00025133653846153844,
+      "loss": 1.111,
+      "step": 2191
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 0.0002513076923076923,
+      "loss": 1.1596,
+      "step": 2192
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 0.0002512788461538461,
+      "loss": 0.9675,
+      "step": 2193
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 0.00025125,
+      "loss": 1.0628,
+      "step": 2194
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 0.0002512211538461538,
+      "loss": 0.9323,
+      "step": 2195
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 0.0002511923076923077,
+      "loss": 1.008,
+      "step": 2196
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.0002511634615384615,
+      "loss": 1.0209,
+      "step": 2197
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.00025113461538461536,
+      "loss": 1.0282,
+      "step": 2198
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.0002511057692307692,
+      "loss": 1.2196,
+      "step": 2199
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 0.00025107692307692304,
+      "loss": 0.9254,
+      "step": 2200
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.0002510480769230769,
+      "loss": 1.0454,
+      "step": 2201
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 0.00025101923076923077,
+      "loss": 0.9107,
+      "step": 2202
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 0.0002509903846153846,
+      "loss": 1.0814,
+      "step": 2203
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 0.00025096153846153844,
+      "loss": 0.9912,
+      "step": 2204
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 0.0002509326923076923,
+      "loss": 0.9584,
+      "step": 2205
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 0.0002509038461538461,
+      "loss": 1.1658,
+      "step": 2206
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 0.000250875,
+      "loss": 0.9101,
+      "step": 2207
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 0.0002508461538461538,
+      "loss": 1.0919,
+      "step": 2208
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 0.0002508173076923077,
+      "loss": 1.1028,
+      "step": 2209
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 0.00025078846153846153,
+      "loss": 1.0411,
+      "step": 2210
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 0.00025075961538461537,
+      "loss": 1.1668,
+      "step": 2211
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 0.0002507307692307692,
+      "loss": 0.9067,
+      "step": 2212
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 0.00025070192307692304,
+      "loss": 0.946,
+      "step": 2213
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 0.00025067307692307693,
+      "loss": 0.9829,
+      "step": 2214
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 0.0002506442307692307,
+      "loss": 1.0911,
+      "step": 2215
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 0.0002506153846153846,
+      "loss": 0.9884,
+      "step": 2216
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 0.00025058653846153845,
+      "loss": 1.0796,
+      "step": 2217
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 0.0002505576923076923,
+      "loss": 1.1911,
+      "step": 2218
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 0.0002505288461538461,
+      "loss": 1.0274,
+      "step": 2219
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 0.00025049999999999996,
+      "loss": 0.9811,
+      "step": 2220
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 0.0002504711538461538,
+      "loss": 1.0571,
+      "step": 2221
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 0.0002504423076923077,
+      "loss": 0.9689,
+      "step": 2222
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 0.00025041346153846153,
+      "loss": 1.0487,
+      "step": 2223
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 0.00025038461538461537,
+      "loss": 1.2162,
+      "step": 2224
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 0.0002503557692307692,
+      "loss": 0.9064,
+      "step": 2225
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 0.00025032692307692305,
+      "loss": 1.0082,
+      "step": 2226
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 0.0002502980769230769,
+      "loss": 1.0547,
+      "step": 2227
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 0.0002502692307692307,
+      "loss": 1.091,
+      "step": 2228
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 0.0002502403846153846,
+      "loss": 1.0892,
+      "step": 2229
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 0.00025021153846153845,
+      "loss": 1.1565,
+      "step": 2230
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 0.0002501826923076923,
+      "loss": 1.0765,
+      "step": 2231
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 0.00025015384615384613,
+      "loss": 1.0928,
+      "step": 2232
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 0.00025012499999999997,
+      "loss": 0.9036,
+      "step": 2233
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 0.00025009615384615386,
+      "loss": 0.9123,
+      "step": 2234
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 0.00025006730769230764,
+      "loss": 1.0183,
+      "step": 2235
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 0.00025003846153846154,
+      "loss": 0.9834,
+      "step": 2236
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 0.0002500096153846153,
+      "loss": 0.9987,
+      "step": 2237
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 0.0002499807692307692,
+      "loss": 0.9926,
+      "step": 2238
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 0.00024995192307692305,
+      "loss": 0.9416,
+      "step": 2239
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 0.0002499230769230769,
+      "loss": 1.0584,
+      "step": 2240
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 0.0002498942307692307,
+      "loss": 1.184,
+      "step": 2241
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 0.00024986538461538457,
+      "loss": 1.0967,
+      "step": 2242
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 0.00024983653846153846,
+      "loss": 1.0339,
+      "step": 2243
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 0.0002498076923076923,
+      "loss": 0.9646,
+      "step": 2244
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 0.00024977884615384613,
+      "loss": 0.9119,
+      "step": 2245
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.00024974999999999997,
+      "loss": 0.9079,
+      "step": 2246
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.0002497211538461538,
+      "loss": 1.1815,
+      "step": 2247
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 0.00024969230769230765,
+      "loss": 0.9657,
+      "step": 2248
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 0.00024966346153846154,
+      "loss": 1.1919,
+      "step": 2249
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 0.0002496346153846154,
+      "loss": 0.9703,
+      "step": 2250
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 0.0002496057692307692,
+      "loss": 1.0874,
+      "step": 2251
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 0.00024957692307692306,
+      "loss": 0.9546,
+      "step": 2252
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 0.0002495480769230769,
+      "loss": 0.944,
+      "step": 2253
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 0.0002495192307692308,
+      "loss": 0.9925,
+      "step": 2254
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 0.00024949038461538457,
+      "loss": 1.1004,
+      "step": 2255
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 0.00024946153846153846,
+      "loss": 0.9923,
+      "step": 2256
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 0.00024943269230769225,
+      "loss": 0.9423,
+      "step": 2257
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.00024940384615384614,
+      "loss": 0.9638,
+      "step": 2258
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.000249375,
+      "loss": 1.0756,
+      "step": 2259
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 0.0002493461538461538,
+      "loss": 1.0498,
+      "step": 2260
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 0.00024931730769230765,
+      "loss": 1.0227,
+      "step": 2261
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 0.0002492884615384615,
+      "loss": 1.0502,
+      "step": 2262
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 0.0002492596153846154,
+      "loss": 1.0247,
+      "step": 2263
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 0.0002492307692307692,
+      "loss": 1.0153,
+      "step": 2264
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 0.00024920192307692306,
+      "loss": 1.0978,
+      "step": 2265
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 0.0002491730769230769,
+      "loss": 1.0462,
+      "step": 2266
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 0.00024914423076923074,
+      "loss": 1.0771,
+      "step": 2267
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 0.0002491153846153846,
+      "loss": 0.9672,
+      "step": 2268
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 0.00024908653846153847,
+      "loss": 1.0011,
+      "step": 2269
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.0002490576923076923,
+      "loss": 0.9798,
+      "step": 2270
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.00024902884615384614,
+      "loss": 1.0174,
+      "step": 2271
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 0.000249,
+      "loss": 0.9587,
+      "step": 2272
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 0.0002489711538461538,
+      "loss": 0.9924,
+      "step": 2273
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 0.0002489423076923077,
+      "loss": 0.9783,
+      "step": 2274
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 0.0002489134615384615,
+      "loss": 0.9852,
+      "step": 2275
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 0.0002488846153846154,
+      "loss": 1.0007,
+      "step": 2276
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 0.00024885576923076917,
+      "loss": 0.9956,
+      "step": 2277
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 0.00024882692307692306,
+      "loss": 1.0386,
+      "step": 2278
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 0.0002487980769230769,
+      "loss": 1.0274,
+      "step": 2279
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 0.00024876923076923074,
+      "loss": 1.109,
+      "step": 2280
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 0.0002487403846153846,
+      "loss": 1.0757,
+      "step": 2281
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.0002487115384615384,
+      "loss": 1.0642,
+      "step": 2282
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.0002486826923076923,
+      "loss": 0.9984,
+      "step": 2283
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 0.00024865384615384615,
+      "loss": 1.0263,
+      "step": 2284
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 0.000248625,
+      "loss": 0.9915,
+      "step": 2285
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 0.0002485961538461538,
+      "loss": 1.0354,
+      "step": 2286
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 0.00024856730769230766,
+      "loss": 0.9999,
+      "step": 2287
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 0.0002485384615384615,
+      "loss": 0.9398,
+      "step": 2288
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 0.0002485096153846154,
+      "loss": 1.0857,
+      "step": 2289
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 0.00024848076923076923,
+      "loss": 1.0468,
+      "step": 2290
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 0.00024845192307692307,
+      "loss": 0.9552,
+      "step": 2291
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 0.0002484230769230769,
+      "loss": 1.0016,
+      "step": 2292
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 0.00024839423076923074,
+      "loss": 0.8649,
+      "step": 2293
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 0.0002483653846153846,
+      "loss": 1.1533,
+      "step": 2294
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 0.0002483365384615384,
+      "loss": 1.1247,
+      "step": 2295
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 0.0002483076923076923,
+      "loss": 1.1274,
+      "step": 2296
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 0.0002482788461538461,
+      "loss": 1.0029,
+      "step": 2297
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 0.00024825,
+      "loss": 0.9945,
+      "step": 2298
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 0.00024822115384615383,
+      "loss": 0.9896,
+      "step": 2299
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 0.00024819230769230767,
+      "loss": 0.984,
+      "step": 2300
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 0.0002481634615384615,
+      "loss": 1.0717,
+      "step": 2301
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 0.00024813461538461534,
+      "loss": 1.1171,
+      "step": 2302
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 0.00024810576923076923,
+      "loss": 1.0317,
+      "step": 2303
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 0.000248076923076923,
+      "loss": 1.0215,
+      "step": 2304
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 0.0002480480769230769,
+      "loss": 0.9327,
+      "step": 2305
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 0.00024801923076923075,
+      "loss": 1.0022,
+      "step": 2306
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 0.0002479903846153846,
+      "loss": 1.0626,
+      "step": 2307
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 0.0002479615384615384,
+      "loss": 1.1625,
+      "step": 2308
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 0.00024793269230769226,
+      "loss": 1.0253,
+      "step": 2309
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 0.00024790384615384616,
+      "loss": 0.9541,
+      "step": 2310
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 0.000247875,
+      "loss": 0.9247,
+      "step": 2311
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 0.00024784615384615383,
+      "loss": 0.9765,
+      "step": 2312
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 0.00024781730769230767,
+      "loss": 1.0285,
+      "step": 2313
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 0.0002477884615384615,
+      "loss": 0.7955,
+      "step": 2314
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 0.00024775961538461535,
+      "loss": 1.0744,
+      "step": 2315
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 0.00024773076923076924,
+      "loss": 1.0363,
+      "step": 2316
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 0.000247701923076923,
+      "loss": 0.8729,
+      "step": 2317
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 0.0002476730769230769,
+      "loss": 0.9094,
+      "step": 2318
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 0.00024764423076923075,
+      "loss": 1.1221,
+      "step": 2319
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 0.0002476153846153846,
+      "loss": 0.8997,
+      "step": 2320
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 0.00024758653846153843,
+      "loss": 0.8927,
+      "step": 2321
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 0.00024755769230769227,
+      "loss": 1.0777,
+      "step": 2322
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 0.00024752884615384616,
+      "loss": 1.038,
+      "step": 2323
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 0.00024749999999999994,
+      "loss": 1.0306,
+      "step": 2324
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 0.00024747115384615384,
+      "loss": 1.0436,
+      "step": 2325
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 0.0002474423076923077,
+      "loss": 1.0319,
+      "step": 2326
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 0.0002474134615384615,
+      "loss": 1.146,
+      "step": 2327
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 0.00024738461538461535,
+      "loss": 0.9436,
+      "step": 2328
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 0.0002473557692307692,
+      "loss": 1.0461,
+      "step": 2329
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 0.0002473269230769231,
+      "loss": 1.112,
+      "step": 2330
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 0.0002472980769230769,
+      "loss": 0.9625,
+      "step": 2331
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 0.00024726923076923076,
+      "loss": 1.0989,
+      "step": 2332
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 0.0002472403846153846,
+      "loss": 1.0553,
+      "step": 2333
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 0.00024721153846153843,
+      "loss": 0.967,
+      "step": 2334
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 0.00024718269230769227,
+      "loss": 0.9735,
+      "step": 2335
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 0.00024715384615384616,
+      "loss": 0.9804,
+      "step": 2336
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 0.00024712499999999995,
+      "loss": 1.004,
+      "step": 2337
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 0.00024709615384615384,
+      "loss": 1.0495,
+      "step": 2338
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 0.0002470673076923077,
+      "loss": 1.0427,
+      "step": 2339
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 0.0002470384615384615,
+      "loss": 1.095,
+      "step": 2340
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 0.00024700961538461536,
+      "loss": 0.8996,
+      "step": 2341
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 0.0002469807692307692,
+      "loss": 1.0172,
+      "step": 2342
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 0.0002469519230769231,
+      "loss": 1.0325,
+      "step": 2343
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 0.00024692307692307687,
+      "loss": 1.1055,
+      "step": 2344
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 0.00024689423076923076,
+      "loss": 0.9461,
+      "step": 2345
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 0.0002468653846153846,
+      "loss": 0.8207,
+      "step": 2346
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 0.00024683653846153844,
+      "loss": 0.9104,
+      "step": 2347
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 0.0002468076923076923,
+      "loss": 0.9176,
+      "step": 2348
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 0.0002467788461538461,
+      "loss": 0.9786,
+      "step": 2349
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 0.00024675,
+      "loss": 1.0939,
+      "step": 2350
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 0.00024672115384615385,
+      "loss": 1.0773,
+      "step": 2351
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 0.0002466923076923077,
+      "loss": 0.9992,
+      "step": 2352
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 0.0002466634615384615,
+      "loss": 0.9204,
+      "step": 2353
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 0.00024663461538461536,
+      "loss": 0.9966,
+      "step": 2354
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 0.0002466057692307692,
+      "loss": 0.973,
+      "step": 2355
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 0.0002465769230769231,
+      "loss": 1.0416,
+      "step": 2356
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 0.0002465480769230769,
+      "loss": 1.0817,
+      "step": 2357
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 0.00024651923076923077,
+      "loss": 1.1516,
+      "step": 2358
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 0.0002464903846153846,
+      "loss": 0.9971,
+      "step": 2359
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 0.00024646153846153844,
+      "loss": 1.0192,
+      "step": 2360
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 0.0002464326923076923,
+      "loss": 0.9631,
+      "step": 2361
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 0.0002464038461538461,
+      "loss": 1.0526,
+      "step": 2362
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 0.000246375,
+      "loss": 1.1358,
+      "step": 2363
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 0.0002463461538461538,
+      "loss": 0.9341,
+      "step": 2364
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 0.0002463173076923077,
+      "loss": 0.9851,
+      "step": 2365
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 0.0002462884615384615,
+      "loss": 0.9223,
+      "step": 2366
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 0.00024625961538461536,
+      "loss": 0.9225,
+      "step": 2367
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 0.0002462307692307692,
+      "loss": 0.9916,
+      "step": 2368
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 0.00024620192307692304,
+      "loss": 0.9075,
+      "step": 2369
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 0.00024617307692307693,
+      "loss": 1.0716,
+      "step": 2370
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 0.0002461442307692307,
+      "loss": 0.9622,
+      "step": 2371
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 0.0002461153846153846,
+      "loss": 1.0533,
+      "step": 2372
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 0.00024608653846153845,
+      "loss": 0.9304,
+      "step": 2373
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 0.0002460576923076923,
+      "loss": 0.9603,
+      "step": 2374
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 0.0002460288461538461,
+      "loss": 1.0125,
+      "step": 2375
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 0.9514,
+      "step": 2376
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 0.0002459711538461538,
+      "loss": 1.0132,
+      "step": 2377
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 0.0002459423076923077,
+      "loss": 1.028,
+      "step": 2378
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 0.00024591346153846153,
+      "loss": 0.9281,
+      "step": 2379
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 0.00024588461538461537,
+      "loss": 1.0633,
+      "step": 2380
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 0.0002458557692307692,
+      "loss": 1.0499,
+      "step": 2381
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 0.00024582692307692305,
+      "loss": 1.0397,
+      "step": 2382
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 0.00024579807692307694,
+      "loss": 1.1488,
+      "step": 2383
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 0.0002457692307692307,
+      "loss": 0.816,
+      "step": 2384
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 0.0002457403846153846,
+      "loss": 1.0981,
+      "step": 2385
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 0.00024571153846153845,
+      "loss": 0.9742,
+      "step": 2386
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 0.0002456826923076923,
+      "loss": 0.9765,
+      "step": 2387
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 0.00024565384615384613,
+      "loss": 0.9621,
+      "step": 2388
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 0.00024562499999999997,
+      "loss": 1.0141,
+      "step": 2389
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 0.0002455961538461538,
+      "loss": 1.0574,
+      "step": 2390
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 0.00024556730769230764,
+      "loss": 0.9589,
+      "step": 2391
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 0.00024553846153846154,
+      "loss": 0.898,
+      "step": 2392
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 0.0002455096153846154,
+      "loss": 1.0181,
+      "step": 2393
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 0.0002454807692307692,
+      "loss": 1.0224,
+      "step": 2394
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 0.00024545192307692305,
+      "loss": 1.0483,
+      "step": 2395
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 0.0002454230769230769,
+      "loss": 0.9467,
+      "step": 2396
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 0.0002453942307692307,
+      "loss": 0.9929,
+      "step": 2397
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 0.0002453653846153846,
+      "loss": 1.1621,
+      "step": 2398
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 0.00024533653846153846,
+      "loss": 0.9726,
+      "step": 2399
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 0.0002453076923076923,
+      "loss": 0.8964,
+      "step": 2400
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 0.00024527884615384613,
+      "loss": 0.9516,
+      "step": 2401
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 0.00024524999999999997,
+      "loss": 1.0199,
+      "step": 2402
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 0.00024522115384615386,
+      "loss": 1.098,
+      "step": 2403
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 0.00024519230769230765,
+      "loss": 1.0478,
+      "step": 2404
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 0.00024516346153846154,
+      "loss": 0.9676,
+      "step": 2405
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 0.0002451346153846154,
+      "loss": 1.0075,
+      "step": 2406
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 0.0002451057692307692,
+      "loss": 1.0927,
+      "step": 2407
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 0.00024507692307692305,
+      "loss": 1.0315,
+      "step": 2408
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 0.0002450480769230769,
+      "loss": 0.9808,
+      "step": 2409
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 0.00024501923076923073,
+      "loss": 0.937,
+      "step": 2410
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 0.00024499038461538457,
+      "loss": 0.9526,
+      "step": 2411
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 0.00024496153846153846,
+      "loss": 0.9909,
+      "step": 2412
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 0.0002449326923076923,
+      "loss": 0.8961,
+      "step": 2413
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 0.00024490384615384614,
+      "loss": 1.0613,
+      "step": 2414
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 0.000244875,
+      "loss": 1.0552,
+      "step": 2415
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 0.0002448461538461538,
+      "loss": 1.0221,
+      "step": 2416
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 0.00024481730769230765,
+      "loss": 1.0374,
+      "step": 2417
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 0.00024478846153846154,
+      "loss": 1.0817,
+      "step": 2418
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 0.0002447596153846154,
+      "loss": 1.0685,
+      "step": 2419
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 0.0002447307692307692,
+      "loss": 0.9469,
+      "step": 2420
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 0.00024470192307692306,
+      "loss": 1.0008,
+      "step": 2421
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 0.0002446730769230769,
+      "loss": 1.0193,
+      "step": 2422
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 0.00024464423076923073,
+      "loss": 1.1084,
+      "step": 2423
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 0.0002446153846153846,
+      "loss": 0.8162,
+      "step": 2424
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 0.00024458653846153847,
+      "loss": 0.9486,
+      "step": 2425
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 0.0002445576923076923,
+      "loss": 0.7643,
+      "step": 2426
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 0.00024452884615384614,
+      "loss": 1.0369,
+      "step": 2427
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 0.0002445,
+      "loss": 0.9099,
+      "step": 2428
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 0.0002444711538461538,
+      "loss": 0.9573,
+      "step": 2429
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 0.00024444230769230766,
+      "loss": 1.0971,
+      "step": 2430
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 0.0002444134615384615,
+      "loss": 1.0211,
+      "step": 2431
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 0.0002443846153846154,
+      "loss": 0.8894,
+      "step": 2432
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 0.00024435576923076917,
+      "loss": 0.9721,
+      "step": 2433
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 0.00024432692307692306,
+      "loss": 1.1158,
+      "step": 2434
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 0.0002442980769230769,
+      "loss": 1.0692,
+      "step": 2435
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 0.00024426923076923074,
+      "loss": 0.9908,
+      "step": 2436
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 0.0002442403846153846,
+      "loss": 0.8691,
+      "step": 2437
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 0.0002442115384615384,
+      "loss": 0.906,
+      "step": 2438
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 0.0002441826923076923,
+      "loss": 1.0143,
+      "step": 2439
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 0.00024415384615384615,
+      "loss": 0.9733,
+      "step": 2440
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 0.00024412499999999996,
+      "loss": 1.0197,
+      "step": 2441
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 0.00024409615384615382,
+      "loss": 1.0411,
+      "step": 2442
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 0.00024406730769230766,
+      "loss": 1.0066,
+      "step": 2443
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 0.00024403846153846153,
+      "loss": 1.015,
+      "step": 2444
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 0.00024400961538461536,
+      "loss": 0.9314,
+      "step": 2445
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 0.0002439807692307692,
+      "loss": 0.9733,
+      "step": 2446
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 0.00024395192307692307,
+      "loss": 1.0133,
+      "step": 2447
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 0.00024392307692307688,
+      "loss": 1.0567,
+      "step": 2448
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 0.00024389423076923074,
+      "loss": 0.939,
+      "step": 2449
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 0.0002438653846153846,
+      "loss": 0.9665,
+      "step": 2450
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 0.00024383653846153842,
+      "loss": 1.0669,
+      "step": 2451
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 0.00024380769230769229,
+      "loss": 1.061,
+      "step": 2452
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 0.00024377884615384612,
+      "loss": 1.0746,
+      "step": 2453
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 0.00024375,
+      "loss": 0.9868,
+      "step": 2454
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 0.00024372115384615383,
+      "loss": 1.0346,
+      "step": 2455
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 0.00024369230769230767,
+      "loss": 0.8872,
+      "step": 2456
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 0.00024366346153846153,
+      "loss": 0.9823,
+      "step": 2457
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 0.00024363461538461534,
+      "loss": 0.991,
+      "step": 2458
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 0.0002436057692307692,
+      "loss": 0.981,
+      "step": 2459
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 0.00024357692307692307,
+      "loss": 1.1096,
+      "step": 2460
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 0.00024354807692307688,
+      "loss": 0.983,
+      "step": 2461
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 0.00024351923076923075,
+      "loss": 0.9233,
+      "step": 2462
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 0.00024349038461538459,
+      "loss": 1.0197,
+      "step": 2463
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 0.00024346153846153845,
+      "loss": 0.9427,
+      "step": 2464
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 0.0002434326923076923,
+      "loss": 0.9564,
+      "step": 2465
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 0.00024340384615384613,
+      "loss": 1.0805,
+      "step": 2466
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 0.000243375,
+      "loss": 1.0746,
+      "step": 2467
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 0.0002433461538461538,
+      "loss": 0.9944,
+      "step": 2468
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 0.00024331730769230767,
+      "loss": 0.8372,
+      "step": 2469
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 0.00024328846153846153,
+      "loss": 1.0722,
+      "step": 2470
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 0.00024325961538461535,
+      "loss": 1.0046,
+      "step": 2471
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 0.0002432307692307692,
+      "loss": 0.998,
+      "step": 2472
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 0.00024320192307692305,
+      "loss": 1.0534,
+      "step": 2473
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 0.00024317307692307691,
+      "loss": 1.0629,
+      "step": 2474
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 0.00024314423076923075,
+      "loss": 1.0305,
+      "step": 2475
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 0.0002431153846153846,
+      "loss": 0.9799,
+      "step": 2476
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 0.00024308653846153846,
+      "loss": 1.0048,
+      "step": 2477
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 0.00024305769230769227,
+      "loss": 1.0339,
+      "step": 2478
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 0.00024302884615384613,
+      "loss": 1.0877,
+      "step": 2479
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 0.000243,
+      "loss": 1.019,
+      "step": 2480
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 0.0002429711538461538,
+      "loss": 0.9782,
+      "step": 2481
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 0.00024294230769230767,
+      "loss": 1.0962,
+      "step": 2482
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 0.0002429134615384615,
+      "loss": 0.9783,
+      "step": 2483
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 0.00024288461538461538,
+      "loss": 1.1296,
+      "step": 2484
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 0.00024285576923076922,
+      "loss": 1.0535,
+      "step": 2485
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 0.00024282692307692305,
+      "loss": 1.025,
+      "step": 2486
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 0.00024279807692307692,
+      "loss": 0.9938,
+      "step": 2487
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 0.00024276923076923073,
+      "loss": 1.016,
+      "step": 2488
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 0.0002427403846153846,
+      "loss": 0.9533,
+      "step": 2489
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 0.00024271153846153843,
+      "loss": 1.0553,
+      "step": 2490
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 0.00024268269230769227,
+      "loss": 0.8619,
+      "step": 2491
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 0.00024265384615384614,
+      "loss": 1.0653,
+      "step": 2492
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 0.00024262499999999998,
+      "loss": 1.0486,
+      "step": 2493
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 0.00024259615384615384,
+      "loss": 0.9191,
+      "step": 2494
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 0.00024256730769230765,
+      "loss": 1.0053,
+      "step": 2495
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 0.00024253846153846152,
+      "loss": 1.0735,
+      "step": 2496
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 0.00024250961538461538,
+      "loss": 1.0469,
+      "step": 2497
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 0.0002424807692307692,
+      "loss": 0.8749,
+      "step": 2498
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 0.00024245192307692306,
+      "loss": 1.0268,
+      "step": 2499
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 0.0002424230769230769,
+      "loss": 1.0305,
+      "step": 2500
+    },
+    {
+      "epoch": 22.94,
+      "eval_cer": 0.1290574071575286,
+      "eval_loss": 0.3763577938079834,
+      "eval_runtime": 13.9197,
+      "eval_samples_per_second": 118.322,
+      "eval_steps_per_second": 1.868,
+      "eval_wer": 0.5007660096006536,
+      "step": 2500
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 0.00024239423076923073,
+      "loss": 1.0169,
+      "step": 2501
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 0.0002423653846153846,
+      "loss": 1.0712,
+      "step": 2502
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 0.00024233653846153844,
+      "loss": 1.0134,
+      "step": 2503
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 0.0002423076923076923,
+      "loss": 1.0801,
+      "step": 2504
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 0.00024227884615384611,
+      "loss": 0.8579,
+      "step": 2505
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 0.00024224999999999998,
+      "loss": 0.9384,
+      "step": 2506
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 0.00024222115384615384,
+      "loss": 1.1154,
+      "step": 2507
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 0.00024219230769230766,
+      "loss": 1.0373,
+      "step": 2508
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 0.00024216346153846152,
+      "loss": 0.8013,
+      "step": 2509
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 0.00024213461538461536,
+      "loss": 0.8923,
+      "step": 2510
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 0.0002421057692307692,
+      "loss": 0.9321,
+      "step": 2511
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 0.00024207692307692306,
+      "loss": 0.9405,
+      "step": 2512
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 0.0002420480769230769,
+      "loss": 1.0994,
+      "step": 2513
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 0.00024201923076923077,
+      "loss": 0.9154,
+      "step": 2514
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 0.00024199038461538458,
+      "loss": 0.8906,
+      "step": 2515
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 0.00024196153846153844,
+      "loss": 1.036,
+      "step": 2516
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 0.0002419326923076923,
+      "loss": 1.0533,
+      "step": 2517
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 0.00024190384615384612,
+      "loss": 1.0259,
+      "step": 2518
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 0.00024187499999999998,
+      "loss": 0.9996,
+      "step": 2519
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 0.00024184615384615382,
+      "loss": 0.9283,
+      "step": 2520
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 0.00024181730769230766,
+      "loss": 1.0247,
+      "step": 2521
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 0.00024178846153846153,
+      "loss": 0.9203,
+      "step": 2522
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 0.00024175961538461536,
+      "loss": 1.0583,
+      "step": 2523
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 0.00024173076923076923,
+      "loss": 1.0519,
+      "step": 2524
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 0.00024170192307692304,
+      "loss": 1.0443,
+      "step": 2525
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 0.0002416730769230769,
+      "loss": 1.1138,
+      "step": 2526
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 0.00024164423076923077,
+      "loss": 0.9624,
+      "step": 2527
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 0.00024161538461538458,
+      "loss": 0.966,
+      "step": 2528
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 0.00024158653846153845,
+      "loss": 0.9319,
+      "step": 2529
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 0.00024155769230769228,
+      "loss": 1.0354,
+      "step": 2530
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 0.00024152884615384612,
+      "loss": 0.9019,
+      "step": 2531
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 0.0002415,
+      "loss": 1.0272,
+      "step": 2532
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 0.00024147115384615383,
+      "loss": 1.0886,
+      "step": 2533
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 0.0002414423076923077,
+      "loss": 0.9449,
+      "step": 2534
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 0.0002414134615384615,
+      "loss": 1.0288,
+      "step": 2535
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 0.00024138461538461537,
+      "loss": 1.1403,
+      "step": 2536
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 0.00024135576923076923,
+      "loss": 0.9779,
+      "step": 2537
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 0.00024132692307692304,
+      "loss": 0.9177,
+      "step": 2538
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 0.0002412980769230769,
+      "loss": 0.9604,
+      "step": 2539
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 0.00024126923076923075,
+      "loss": 0.9402,
+      "step": 2540
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 0.00024124038461538459,
+      "loss": 1.0191,
+      "step": 2541
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 0.00024121153846153845,
+      "loss": 1.0426,
+      "step": 2542
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 0.0002411826923076923,
+      "loss": 1.026,
+      "step": 2543
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 0.00024115384615384615,
+      "loss": 1.1089,
+      "step": 2544
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 0.00024112499999999997,
+      "loss": 1.0679,
+      "step": 2545
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 0.00024109615384615383,
+      "loss": 0.9404,
+      "step": 2546
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 0.0002410673076923077,
+      "loss": 0.8838,
+      "step": 2547
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 0.0002410384615384615,
+      "loss": 0.9859,
+      "step": 2548
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 0.00024100961538461537,
+      "loss": 1.0199,
+      "step": 2549
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 0.0002409807692307692,
+      "loss": 1.1213,
+      "step": 2550
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 0.00024095192307692305,
+      "loss": 0.8465,
+      "step": 2551
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 0.0002409230769230769,
+      "loss": 0.9573,
+      "step": 2552
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 0.00024089423076923075,
+      "loss": 0.9833,
+      "step": 2553
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 0.00024086538461538462,
+      "loss": 1.002,
+      "step": 2554
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 0.00024083653846153843,
+      "loss": 1.0389,
+      "step": 2555
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 0.0002408076923076923,
+      "loss": 0.9287,
+      "step": 2556
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 0.0002407788461538461,
+      "loss": 0.8941,
+      "step": 2557
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 0.00024074999999999997,
+      "loss": 1.0039,
+      "step": 2558
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 0.00024072115384615384,
+      "loss": 1.007,
+      "step": 2559
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 0.00024069230769230767,
+      "loss": 1.0447,
+      "step": 2560
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 0.0002406634615384615,
+      "loss": 1.0056,
+      "step": 2561
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 0.00024063461538461535,
+      "loss": 1.053,
+      "step": 2562
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 0.00024060576923076922,
+      "loss": 0.8958,
+      "step": 2563
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 0.00024057692307692308,
+      "loss": 0.9761,
+      "step": 2564
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 0.0002405480769230769,
+      "loss": 0.9343,
+      "step": 2565
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 0.00024051923076923076,
+      "loss": 1.1249,
+      "step": 2566
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 0.00024049038461538457,
+      "loss": 0.8779,
+      "step": 2567
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 0.00024046153846153843,
+      "loss": 0.829,
+      "step": 2568
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 0.0002404326923076923,
+      "loss": 1.071,
+      "step": 2569
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 0.00024040384615384614,
+      "loss": 0.9896,
+      "step": 2570
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 0.00024037499999999997,
+      "loss": 0.8804,
+      "step": 2571
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 0.0002403461538461538,
+      "loss": 0.9863,
+      "step": 2572
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 0.00024031730769230768,
+      "loss": 1.0295,
+      "step": 2573
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 0.00024028846153846154,
+      "loss": 0.9566,
+      "step": 2574
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 0.00024025961538461535,
+      "loss": 0.8748,
+      "step": 2575
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 0.00024023076923076922,
+      "loss": 1.0898,
+      "step": 2576
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 0.00024020192307692303,
+      "loss": 1.0121,
+      "step": 2577
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 0.0002401730769230769,
+      "loss": 0.9366,
+      "step": 2578
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 0.00024014423076923076,
+      "loss": 0.9123,
+      "step": 2579
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 0.0002401153846153846,
+      "loss": 1.1958,
+      "step": 2580
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 0.00024008653846153844,
+      "loss": 1.0522,
+      "step": 2581
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 0.00024005769230769228,
+      "loss": 1.0121,
+      "step": 2582
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 0.00024002884615384614,
+      "loss": 1.0871,
+      "step": 2583
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 1.0457,
+      "step": 2584
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 0.00023997115384615382,
+      "loss": 0.8922,
+      "step": 2585
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 0.00023994230769230768,
+      "loss": 0.9461,
+      "step": 2586
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 0.0002399134615384615,
+      "loss": 0.894,
+      "step": 2587
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 0.00023988461538461536,
+      "loss": 0.9194,
+      "step": 2588
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 0.00023985576923076922,
+      "loss": 0.8221,
+      "step": 2589
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 0.00023982692307692306,
+      "loss": 0.8919,
+      "step": 2590
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 0.0002397980769230769,
+      "loss": 0.8744,
+      "step": 2591
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 0.00023976923076923074,
+      "loss": 0.9383,
+      "step": 2592
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 0.0002397403846153846,
+      "loss": 1.0766,
+      "step": 2593
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 0.00023971153846153844,
+      "loss": 0.9391,
+      "step": 2594
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 0.00023968269230769228,
+      "loss": 0.9948,
+      "step": 2595
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 0.00023965384615384615,
+      "loss": 0.9384,
+      "step": 2596
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 0.00023962499999999996,
+      "loss": 1.034,
+      "step": 2597
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 0.00023959615384615382,
+      "loss": 1.0524,
+      "step": 2598
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 0.0002395673076923077,
+      "loss": 1.0099,
+      "step": 2599
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 0.00023953846153846152,
+      "loss": 1.1377,
+      "step": 2600
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 0.00023950961538461536,
+      "loss": 1.0414,
+      "step": 2601
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 0.0002394807692307692,
+      "loss": 0.9964,
+      "step": 2602
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 0.00023945192307692307,
+      "loss": 0.9811,
+      "step": 2603
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 0.0002394230769230769,
+      "loss": 0.9164,
+      "step": 2604
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 0.00023939423076923074,
+      "loss": 1.0058,
+      "step": 2605
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 0.0002393653846153846,
+      "loss": 1.0143,
+      "step": 2606
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 0.00023933653846153842,
+      "loss": 1.0079,
+      "step": 2607
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 0.00023930769230769228,
+      "loss": 1.0616,
+      "step": 2608
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 0.00023927884615384615,
+      "loss": 1.1004,
+      "step": 2609
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 0.00023925,
+      "loss": 1.0394,
+      "step": 2610
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 0.00023922115384615383,
+      "loss": 1.0252,
+      "step": 2611
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 0.00023919230769230766,
+      "loss": 1.0073,
+      "step": 2612
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 0.00023916346153846153,
+      "loss": 0.8884,
+      "step": 2613
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 0.00023913461538461537,
+      "loss": 1.0081,
+      "step": 2614
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 0.0002391057692307692,
+      "loss": 1.0664,
+      "step": 2615
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 0.00023907692307692307,
+      "loss": 0.9502,
+      "step": 2616
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 0.00023904807692307688,
+      "loss": 0.9432,
+      "step": 2617
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 0.00023901923076923075,
+      "loss": 1.0395,
+      "step": 2618
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 0.00023899038461538459,
+      "loss": 0.902,
+      "step": 2619
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 0.00023896153846153845,
+      "loss": 0.9133,
+      "step": 2620
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 0.0002389326923076923,
+      "loss": 0.8401,
+      "step": 2621
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 0.00023890384615384613,
+      "loss": 0.99,
+      "step": 2622
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 0.000238875,
+      "loss": 1.0101,
+      "step": 2623
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 0.0002388461538461538,
+      "loss": 1.0537,
+      "step": 2624
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 0.00023881730769230767,
+      "loss": 1.0375,
+      "step": 2625
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 0.00023878846153846153,
+      "loss": 1.0,
+      "step": 2626
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 0.00023875961538461534,
+      "loss": 0.9791,
+      "step": 2627
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 0.0002387307692307692,
+      "loss": 0.9529,
+      "step": 2628
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 0.00023870192307692305,
+      "loss": 0.9091,
+      "step": 2629
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 0.00023867307692307691,
+      "loss": 0.9877,
+      "step": 2630
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 0.00023864423076923075,
+      "loss": 0.8715,
+      "step": 2631
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 0.0002386153846153846,
+      "loss": 0.9729,
+      "step": 2632
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 0.00023858653846153846,
+      "loss": 1.077,
+      "step": 2633
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 0.00023855769230769227,
+      "loss": 1.0093,
+      "step": 2634
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 0.00023852884615384613,
+      "loss": 1.0488,
+      "step": 2635
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 0.0002385,
+      "loss": 0.972,
+      "step": 2636
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 0.0002384711538461538,
+      "loss": 1.0079,
+      "step": 2637
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 0.00023844230769230767,
+      "loss": 0.9188,
+      "step": 2638
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 0.0002384134615384615,
+      "loss": 1.0419,
+      "step": 2639
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 0.00023838461538461535,
+      "loss": 0.9396,
+      "step": 2640
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 0.00023835576923076921,
+      "loss": 1.0672,
+      "step": 2641
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 0.00023832692307692305,
+      "loss": 0.9391,
+      "step": 2642
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 0.00023829807692307692,
+      "loss": 1.0703,
+      "step": 2643
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 0.00023826923076923073,
+      "loss": 0.9194,
+      "step": 2644
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 0.0002382403846153846,
+      "loss": 1.0457,
+      "step": 2645
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 0.00023821153846153846,
+      "loss": 1.0835,
+      "step": 2646
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 0.00023818269230769227,
+      "loss": 1.1428,
+      "step": 2647
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 0.00023815384615384614,
+      "loss": 0.8919,
+      "step": 2648
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 0.00023812499999999997,
+      "loss": 0.9169,
+      "step": 2649
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 0.0002380961538461538,
+      "loss": 1.0673,
+      "step": 2650
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 0.00023806730769230768,
+      "loss": 1.1015,
+      "step": 2651
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 0.00023803846153846152,
+      "loss": 0.9986,
+      "step": 2652
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 0.00023800961538461538,
+      "loss": 0.9879,
+      "step": 2653
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 0.0002379807692307692,
+      "loss": 1.1196,
+      "step": 2654
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 0.00023795192307692306,
+      "loss": 0.9341,
+      "step": 2655
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 0.00023792307692307692,
+      "loss": 0.867,
+      "step": 2656
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 0.00023789423076923073,
+      "loss": 1.0139,
+      "step": 2657
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 0.0002378653846153846,
+      "loss": 1.0075,
+      "step": 2658
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 0.00023783653846153844,
+      "loss": 1.0133,
+      "step": 2659
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 0.00023780769230769228,
+      "loss": 1.1136,
+      "step": 2660
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 0.00023777884615384614,
+      "loss": 1.0049,
+      "step": 2661
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 0.00023774999999999998,
+      "loss": 0.8914,
+      "step": 2662
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 0.00023772115384615384,
+      "loss": 0.9494,
+      "step": 2663
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 0.00023769230769230765,
+      "loss": 1.016,
+      "step": 2664
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 0.00023766346153846152,
+      "loss": 1.0131,
+      "step": 2665
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 0.00023763461538461539,
+      "loss": 0.841,
+      "step": 2666
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 0.0002376057692307692,
+      "loss": 0.977,
+      "step": 2667
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 0.00023757692307692306,
+      "loss": 1.0514,
+      "step": 2668
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 0.0002375480769230769,
+      "loss": 0.9373,
+      "step": 2669
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 0.00023751923076923074,
+      "loss": 1.1145,
+      "step": 2670
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 0.0002374903846153846,
+      "loss": 1.0972,
+      "step": 2671
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 0.00023746153846153844,
+      "loss": 1.0414,
+      "step": 2672
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 0.0002374326923076923,
+      "loss": 1.0218,
+      "step": 2673
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 0.00023740384615384612,
+      "loss": 0.8954,
+      "step": 2674
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 0.00023737499999999998,
+      "loss": 0.8945,
+      "step": 2675
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 0.00023734615384615385,
+      "loss": 1.0756,
+      "step": 2676
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 0.00023731730769230766,
+      "loss": 1.1196,
+      "step": 2677
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 0.00023728846153846152,
+      "loss": 1.0709,
+      "step": 2678
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 0.00023725961538461536,
+      "loss": 0.9956,
+      "step": 2679
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 0.0002372307692307692,
+      "loss": 1.0179,
+      "step": 2680
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 0.00023720192307692307,
+      "loss": 0.9863,
+      "step": 2681
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 0.0002371730769230769,
+      "loss": 1.029,
+      "step": 2682
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 0.00023714423076923077,
+      "loss": 1.0582,
+      "step": 2683
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 0.00023711538461538458,
+      "loss": 0.9888,
+      "step": 2684
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 0.00023708653846153845,
+      "loss": 0.9607,
+      "step": 2685
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 0.00023705769230769228,
+      "loss": 0.8798,
+      "step": 2686
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 0.00023702884615384612,
+      "loss": 0.9399,
+      "step": 2687
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 0.000237,
+      "loss": 1.0365,
+      "step": 2688
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 0.00023697115384615383,
+      "loss": 0.9688,
+      "step": 2689
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 0.00023694230769230766,
+      "loss": 1.1015,
+      "step": 2690
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 0.0002369134615384615,
+      "loss": 0.9692,
+      "step": 2691
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 0.00023688461538461537,
+      "loss": 0.9623,
+      "step": 2692
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 0.00023685576923076923,
+      "loss": 0.9899,
+      "step": 2693
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 0.00023682692307692304,
+      "loss": 0.9615,
+      "step": 2694
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 0.0002367980769230769,
+      "loss": 0.9339,
+      "step": 2695
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 0.00023676923076923075,
+      "loss": 1.0056,
+      "step": 2696
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 0.00023674038461538458,
+      "loss": 1.0326,
+      "step": 2697
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 0.00023671153846153845,
+      "loss": 0.957,
+      "step": 2698
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 0.0002366826923076923,
+      "loss": 0.993,
+      "step": 2699
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 0.00023665384615384613,
+      "loss": 0.9926,
+      "step": 2700
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 0.00023662499999999996,
+      "loss": 1.0046,
+      "step": 2701
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 0.00023659615384615383,
+      "loss": 1.0829,
+      "step": 2702
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 0.0002365673076923077,
+      "loss": 1.1442,
+      "step": 2703
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 0.0002365384615384615,
+      "loss": 1.0408,
+      "step": 2704
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 0.00023650961538461537,
+      "loss": 0.9268,
+      "step": 2705
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 0.00023648076923076918,
+      "loss": 0.9436,
+      "step": 2706
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 0.00023645192307692305,
+      "loss": 0.9662,
+      "step": 2707
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 0.0002364230769230769,
+      "loss": 1.031,
+      "step": 2708
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 0.00023639423076923075,
+      "loss": 1.0035,
+      "step": 2709
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 0.0002363653846153846,
+      "loss": 1.0448,
+      "step": 2710
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 0.00023633653846153843,
+      "loss": 0.9809,
+      "step": 2711
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 0.0002363076923076923,
+      "loss": 0.9198,
+      "step": 2712
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 0.00023627884615384616,
+      "loss": 1.0456,
+      "step": 2713
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 0.00023624999999999997,
+      "loss": 0.9449,
+      "step": 2714
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 0.00023622115384615383,
+      "loss": 0.9812,
+      "step": 2715
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 0.00023619230769230765,
+      "loss": 1.032,
+      "step": 2716
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 0.0002361634615384615,
+      "loss": 0.9564,
+      "step": 2717
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 0.00023613461538461538,
+      "loss": 0.9548,
+      "step": 2718
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 0.00023610576923076921,
+      "loss": 0.8484,
+      "step": 2719
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 0.00023607692307692305,
+      "loss": 0.984,
+      "step": 2720
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 0.0002360480769230769,
+      "loss": 0.9453,
+      "step": 2721
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 0.00023601923076923076,
+      "loss": 1.0067,
+      "step": 2722
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 0.00023599038461538462,
+      "loss": 1.027,
+      "step": 2723
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 0.00023596153846153843,
+      "loss": 0.8671,
+      "step": 2724
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 0.0002359326923076923,
+      "loss": 0.8894,
+      "step": 2725
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 0.0002359038461538461,
+      "loss": 1.0032,
+      "step": 2726
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 0.00023587499999999997,
+      "loss": 0.9077,
+      "step": 2727
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 0.00023584615384615384,
+      "loss": 1.0701,
+      "step": 2728
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 0.00023581730769230768,
+      "loss": 0.8545,
+      "step": 2729
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 0.00023578846153846152,
+      "loss": 1.0503,
+      "step": 2730
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 0.00023575961538461535,
+      "loss": 1.071,
+      "step": 2731
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 0.00023573076923076922,
+      "loss": 0.8629,
+      "step": 2732
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 0.00023570192307692308,
+      "loss": 0.9499,
+      "step": 2733
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 0.0002356730769230769,
+      "loss": 0.9854,
+      "step": 2734
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 0.00023564423076923076,
+      "loss": 0.9517,
+      "step": 2735
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 0.00023561538461538457,
+      "loss": 1.004,
+      "step": 2736
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 0.00023558653846153844,
+      "loss": 0.9287,
+      "step": 2737
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 0.0002355576923076923,
+      "loss": 1.0507,
+      "step": 2738
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 0.00023552884615384614,
+      "loss": 0.8848,
+      "step": 2739
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 0.00023549999999999998,
+      "loss": 0.9722,
+      "step": 2740
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 0.00023547115384615382,
+      "loss": 1.0156,
+      "step": 2741
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 0.00023544230769230768,
+      "loss": 1.1699,
+      "step": 2742
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 0.00023541346153846155,
+      "loss": 1.1063,
+      "step": 2743
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 0.00023538461538461536,
+      "loss": 0.8953,
+      "step": 2744
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 0.00023535576923076922,
+      "loss": 1.0395,
+      "step": 2745
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 0.00023532692307692303,
+      "loss": 0.9972,
+      "step": 2746
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 0.0002352980769230769,
+      "loss": 1.0894,
+      "step": 2747
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 0.00023526923076923074,
+      "loss": 0.9414,
+      "step": 2748
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 0.0002352403846153846,
+      "loss": 0.8582,
+      "step": 2749
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 0.00023521153846153844,
+      "loss": 1.0297,
+      "step": 2750
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 0.00023518269230769228,
+      "loss": 0.9998,
+      "step": 2751
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 0.00023515384615384614,
+      "loss": 0.9315,
+      "step": 2752
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 0.00023512499999999996,
+      "loss": 0.8833,
+      "step": 2753
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 0.00023509615384615382,
+      "loss": 0.882,
+      "step": 2754
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 0.00023506730769230769,
+      "loss": 1.0875,
+      "step": 2755
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 0.0002350384615384615,
+      "loss": 1.1512,
+      "step": 2756
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 0.00023500961538461536,
+      "loss": 1.0212,
+      "step": 2757
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 0.0002349807692307692,
+      "loss": 1.0595,
+      "step": 2758
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 0.00023495192307692307,
+      "loss": 1.0457,
+      "step": 2759
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 0.0002349230769230769,
+      "loss": 1.0674,
+      "step": 2760
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 0.00023489423076923074,
+      "loss": 1.0412,
+      "step": 2761
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 0.0002348653846153846,
+      "loss": 1.1541,
+      "step": 2762
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 0.00023483653846153842,
+      "loss": 0.8444,
+      "step": 2763
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 0.00023480769230769228,
+      "loss": 0.9383,
+      "step": 2764
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 0.00023477884615384615,
+      "loss": 0.8923,
+      "step": 2765
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 0.00023474999999999996,
+      "loss": 1.0799,
+      "step": 2766
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 0.00023472115384615383,
+      "loss": 0.9781,
+      "step": 2767
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 0.00023469230769230766,
+      "loss": 1.0694,
+      "step": 2768
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 0.00023466346153846153,
+      "loss": 0.97,
+      "step": 2769
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 0.00023463461538461537,
+      "loss": 1.0157,
+      "step": 2770
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 0.0002346057692307692,
+      "loss": 0.9839,
+      "step": 2771
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 0.00023457692307692307,
+      "loss": 1.077,
+      "step": 2772
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 0.00023454807692307688,
+      "loss": 0.9978,
+      "step": 2773
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 0.00023451923076923075,
+      "loss": 1.0551,
+      "step": 2774
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 0.0002344903846153846,
+      "loss": 1.0444,
+      "step": 2775
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 0.00023446153846153842,
+      "loss": 1.0203,
+      "step": 2776
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 0.0002344326923076923,
+      "loss": 0.9964,
+      "step": 2777
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 0.00023440384615384613,
+      "loss": 0.9041,
+      "step": 2778
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 0.000234375,
+      "loss": 1.0092,
+      "step": 2779
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 0.00023434615384615383,
+      "loss": 0.9835,
+      "step": 2780
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 0.00023431730769230767,
+      "loss": 0.9713,
+      "step": 2781
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 0.00023428846153846153,
+      "loss": 0.9386,
+      "step": 2782
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 0.00023425961538461534,
+      "loss": 1.0201,
+      "step": 2783
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 0.0002342307692307692,
+      "loss": 0.9765,
+      "step": 2784
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 0.00023420192307692307,
+      "loss": 0.963,
+      "step": 2785
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 0.00023417307692307689,
+      "loss": 0.9594,
+      "step": 2786
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 0.00023414423076923075,
+      "loss": 1.0212,
+      "step": 2787
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 0.0002341153846153846,
+      "loss": 1.0073,
+      "step": 2788
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 0.00023408653846153845,
+      "loss": 0.9011,
+      "step": 2789
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 0.0002340576923076923,
+      "loss": 0.8452,
+      "step": 2790
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 0.00023402884615384613,
+      "loss": 0.9157,
+      "step": 2791
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 0.000234,
+      "loss": 1.0201,
+      "step": 2792
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 0.0002339711538461538,
+      "loss": 0.9164,
+      "step": 2793
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 0.00023394230769230767,
+      "loss": 1.0183,
+      "step": 2794
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 0.00023391346153846154,
+      "loss": 0.9536,
+      "step": 2795
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 0.00023388461538461535,
+      "loss": 0.8979,
+      "step": 2796
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 0.00023385576923076921,
+      "loss": 0.9786,
+      "step": 2797
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 0.00023382692307692305,
+      "loss": 0.9942,
+      "step": 2798
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 0.00023379807692307692,
+      "loss": 0.912,
+      "step": 2799
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 0.00023376923076923076,
+      "loss": 1.1286,
+      "step": 2800
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 0.0002337403846153846,
+      "loss": 0.981,
+      "step": 2801
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 0.00023371153846153846,
+      "loss": 0.9875,
+      "step": 2802
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 0.00023368269230769227,
+      "loss": 0.9702,
+      "step": 2803
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 0.00023365384615384613,
+      "loss": 0.9914,
+      "step": 2804
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 0.000233625,
+      "loss": 0.9561,
+      "step": 2805
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 0.0002335961538461538,
+      "loss": 0.945,
+      "step": 2806
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 0.00023356730769230768,
+      "loss": 0.9639,
+      "step": 2807
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 0.00023353846153846151,
+      "loss": 0.8438,
+      "step": 2808
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 0.00023350961538461538,
+      "loss": 0.8396,
+      "step": 2809
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 0.00023348076923076922,
+      "loss": 0.9756,
+      "step": 2810
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 0.00023345192307692306,
+      "loss": 1.1054,
+      "step": 2811
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 0.00023342307692307692,
+      "loss": 1.0288,
+      "step": 2812
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 0.00023339423076923073,
+      "loss": 1.0908,
+      "step": 2813
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 0.0002333653846153846,
+      "loss": 0.9218,
+      "step": 2814
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 0.00023333653846153844,
+      "loss": 0.8513,
+      "step": 2815
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 0.00023330769230769227,
+      "loss": 0.8694,
+      "step": 2816
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 0.00023327884615384614,
+      "loss": 1.0066,
+      "step": 2817
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 0.00023324999999999998,
+      "loss": 0.9183,
+      "step": 2818
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 0.00023322115384615384,
+      "loss": 0.9953,
+      "step": 2819
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 0.00023319230769230765,
+      "loss": 0.9374,
+      "step": 2820
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 0.00023316346153846152,
+      "loss": 0.9689,
+      "step": 2821
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 0.00023313461538461538,
+      "loss": 0.8937,
+      "step": 2822
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 0.0002331057692307692,
+      "loss": 1.0183,
+      "step": 2823
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 0.00023307692307692306,
+      "loss": 1.0309,
+      "step": 2824
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 0.0002330480769230769,
+      "loss": 0.9639,
+      "step": 2825
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 0.00023301923076923074,
+      "loss": 0.9701,
+      "step": 2826
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 0.0002329903846153846,
+      "loss": 0.8341,
+      "step": 2827
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 0.00023296153846153844,
+      "loss": 0.859,
+      "step": 2828
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 0.0002329326923076923,
+      "loss": 0.9383,
+      "step": 2829
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 0.00023290384615384612,
+      "loss": 0.9877,
+      "step": 2830
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 0.00023287499999999998,
+      "loss": 0.9729,
+      "step": 2831
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 0.00023284615384615385,
+      "loss": 0.8662,
+      "step": 2832
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 0.00023281730769230766,
+      "loss": 0.9447,
+      "step": 2833
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 0.00023278846153846152,
+      "loss": 0.9733,
+      "step": 2834
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 0.00023275961538461536,
+      "loss": 0.9663,
+      "step": 2835
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 0.0002327307692307692,
+      "loss": 1.0509,
+      "step": 2836
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 0.00023270192307692307,
+      "loss": 0.9961,
+      "step": 2837
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 0.0002326730769230769,
+      "loss": 0.9086,
+      "step": 2838
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 0.00023264423076923077,
+      "loss": 0.9391,
+      "step": 2839
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 0.00023261538461538458,
+      "loss": 1.1448,
+      "step": 2840
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 0.00023258653846153844,
+      "loss": 0.8751,
+      "step": 2841
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 0.0002325576923076923,
+      "loss": 0.996,
+      "step": 2842
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 0.00023252884615384612,
+      "loss": 0.9759,
+      "step": 2843
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 0.00023249999999999999,
+      "loss": 1.0193,
+      "step": 2844
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 0.00023247115384615382,
+      "loss": 0.9936,
+      "step": 2845
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 0.00023244230769230766,
+      "loss": 0.9702,
+      "step": 2846
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 0.00023241346153846153,
+      "loss": 0.9848,
+      "step": 2847
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 0.00023238461538461537,
+      "loss": 0.9974,
+      "step": 2848
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 0.0002323557692307692,
+      "loss": 0.9961,
+      "step": 2849
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 0.00023232692307692304,
+      "loss": 0.9197,
+      "step": 2850
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 0.0002322980769230769,
+      "loss": 1.1451,
+      "step": 2851
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 0.00023226923076923077,
+      "loss": 1.1711,
+      "step": 2852
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 0.00023224038461538458,
+      "loss": 1.0515,
+      "step": 2853
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 0.00023221153846153845,
+      "loss": 1.0113,
+      "step": 2854
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 0.0002321826923076923,
+      "loss": 0.8901,
+      "step": 2855
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 0.00023215384615384613,
+      "loss": 0.8921,
+      "step": 2856
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 0.000232125,
+      "loss": 1.1401,
+      "step": 2857
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 0.00023209615384615383,
+      "loss": 0.8854,
+      "step": 2858
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 0.00023206730769230767,
+      "loss": 1.0948,
+      "step": 2859
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 0.0002320384615384615,
+      "loss": 1.0171,
+      "step": 2860
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 0.00023200961538461537,
+      "loss": 1.0538,
+      "step": 2861
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 0.00023198076923076924,
+      "loss": 0.9039,
+      "step": 2862
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 0.00023195192307692305,
+      "loss": 0.9909,
+      "step": 2863
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 0.0002319230769230769,
+      "loss": 0.9434,
+      "step": 2864
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 0.00023189423076923075,
+      "loss": 0.996,
+      "step": 2865
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 0.0002318653846153846,
+      "loss": 1.0295,
+      "step": 2866
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 0.00023183653846153845,
+      "loss": 0.9508,
+      "step": 2867
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 0.0002318076923076923,
+      "loss": 0.9032,
+      "step": 2868
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 0.00023177884615384613,
+      "loss": 1.1038,
+      "step": 2869
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 0.00023174999999999997,
+      "loss": 0.8602,
+      "step": 2870
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 0.00023172115384615383,
+      "loss": 1.0498,
+      "step": 2871
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 0.0002316923076923077,
+      "loss": 0.9317,
+      "step": 2872
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 0.0002316634615384615,
+      "loss": 0.97,
+      "step": 2873
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 0.00023163461538461538,
+      "loss": 1.042,
+      "step": 2874
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 0.0002316057692307692,
+      "loss": 0.8044,
+      "step": 2875
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 0.00023157692307692305,
+      "loss": 0.989,
+      "step": 2876
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 0.0002315480769230769,
+      "loss": 0.9897,
+      "step": 2877
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 0.00023151923076923075,
+      "loss": 0.9931,
+      "step": 2878
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 0.0002314903846153846,
+      "loss": 0.8748,
+      "step": 2879
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 0.00023146153846153843,
+      "loss": 0.9517,
+      "step": 2880
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 0.0002314326923076923,
+      "loss": 0.9103,
+      "step": 2881
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 0.0002314038461538461,
+      "loss": 1.1293,
+      "step": 2882
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 0.00023137499999999997,
+      "loss": 0.8936,
+      "step": 2883
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 0.00023134615384615384,
+      "loss": 1.0006,
+      "step": 2884
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 0.00023131730769230768,
+      "loss": 0.9962,
+      "step": 2885
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 0.00023128846153846151,
+      "loss": 0.9515,
+      "step": 2886
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 0.00023125961538461535,
+      "loss": 1.0119,
+      "step": 2887
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 0.00023123076923076922,
+      "loss": 0.9287,
+      "step": 2888
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 0.00023120192307692306,
+      "loss": 1.0289,
+      "step": 2889
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 0.0002311730769230769,
+      "loss": 0.9163,
+      "step": 2890
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 0.00023114423076923076,
+      "loss": 0.9641,
+      "step": 2891
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 0.00023111538461538457,
+      "loss": 0.9098,
+      "step": 2892
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 0.00023108653846153844,
+      "loss": 0.8743,
+      "step": 2893
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 0.0002310576923076923,
+      "loss": 0.9347,
+      "step": 2894
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 0.00023102884615384614,
+      "loss": 0.9011,
+      "step": 2895
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 0.9992,
+      "step": 2896
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 0.00023097115384615382,
+      "loss": 0.9659,
+      "step": 2897
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 0.00023094230769230768,
+      "loss": 0.9647,
+      "step": 2898
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 0.00023091346153846152,
+      "loss": 0.8933,
+      "step": 2899
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 0.00023088461538461536,
+      "loss": 0.9605,
+      "step": 2900
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 0.00023085576923076922,
+      "loss": 0.9287,
+      "step": 2901
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 0.00023082692307692303,
+      "loss": 1.0517,
+      "step": 2902
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 0.0002307980769230769,
+      "loss": 1.0275,
+      "step": 2903
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 0.00023076923076923076,
+      "loss": 0.9705,
+      "step": 2904
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 0.00023074038461538457,
+      "loss": 0.9357,
+      "step": 2905
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 0.00023071153846153844,
+      "loss": 1.0383,
+      "step": 2906
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 0.00023068269230769228,
+      "loss": 0.9326,
+      "step": 2907
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 0.00023065384615384614,
+      "loss": 1.0818,
+      "step": 2908
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 0.00023062499999999998,
+      "loss": 0.8033,
+      "step": 2909
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 0.00023059615384615382,
+      "loss": 1.0605,
+      "step": 2910
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 0.00023056730769230768,
+      "loss": 0.8993,
+      "step": 2911
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 0.0002305384615384615,
+      "loss": 0.7971,
+      "step": 2912
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 0.00023050961538461536,
+      "loss": 1.0363,
+      "step": 2913
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 0.00023048076923076923,
+      "loss": 1.0726,
+      "step": 2914
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 0.00023045192307692304,
+      "loss": 1.0054,
+      "step": 2915
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 0.0002304230769230769,
+      "loss": 0.9765,
+      "step": 2916
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 0.00023039423076923074,
+      "loss": 0.9337,
+      "step": 2917
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 0.0002303653846153846,
+      "loss": 1.0114,
+      "step": 2918
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 0.00023033653846153844,
+      "loss": 0.9495,
+      "step": 2919
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 0.00023030769230769228,
+      "loss": 1.0202,
+      "step": 2920
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 0.00023027884615384615,
+      "loss": 1.0068,
+      "step": 2921
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 0.00023024999999999996,
+      "loss": 0.9452,
+      "step": 2922
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 0.00023022115384615382,
+      "loss": 0.8533,
+      "step": 2923
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 0.0002301923076923077,
+      "loss": 0.8575,
+      "step": 2924
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 0.0002301634615384615,
+      "loss": 0.9527,
+      "step": 2925
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 0.00023013461538461537,
+      "loss": 0.9422,
+      "step": 2926
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 0.0002301057692307692,
+      "loss": 0.8829,
+      "step": 2927
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 0.00023007692307692307,
+      "loss": 0.9465,
+      "step": 2928
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 0.0002300480769230769,
+      "loss": 0.9599,
+      "step": 2929
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 0.00023001923076923075,
+      "loss": 0.9606,
+      "step": 2930
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 0.0002299903846153846,
+      "loss": 0.935,
+      "step": 2931
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 0.00022996153846153842,
+      "loss": 0.8594,
+      "step": 2932
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 0.0002299326923076923,
+      "loss": 0.8879,
+      "step": 2933
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 0.00022990384615384615,
+      "loss": 0.9657,
+      "step": 2934
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 0.00022987499999999996,
+      "loss": 1.0739,
+      "step": 2935
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 0.00022984615384615383,
+      "loss": 1.0341,
+      "step": 2936
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 0.00022981730769230767,
+      "loss": 0.9731,
+      "step": 2937
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 0.00022978846153846153,
+      "loss": 1.037,
+      "step": 2938
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 0.00022975961538461537,
+      "loss": 1.0224,
+      "step": 2939
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 0.0002297307692307692,
+      "loss": 1.0722,
+      "step": 2940
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 0.00022970192307692307,
+      "loss": 0.8558,
+      "step": 2941
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 0.00022967307692307688,
+      "loss": 0.9403,
+      "step": 2942
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 0.00022964423076923075,
+      "loss": 0.924,
+      "step": 2943
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 0.0002296153846153846,
+      "loss": 1.0115,
+      "step": 2944
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 0.00022958653846153843,
+      "loss": 0.9588,
+      "step": 2945
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 0.0002295576923076923,
+      "loss": 0.8319,
+      "step": 2946
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 0.00022952884615384613,
+      "loss": 0.8855,
+      "step": 2947
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 0.0002295,
+      "loss": 0.9733,
+      "step": 2948
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 0.0002294711538461538,
+      "loss": 0.9015,
+      "step": 2949
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 0.00022944230769230767,
+      "loss": 0.9919,
+      "step": 2950
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 0.00022941346153846154,
+      "loss": 0.9091,
+      "step": 2951
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 0.00022938461538461535,
+      "loss": 0.8429,
+      "step": 2952
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 0.0002293557692307692,
+      "loss": 0.9202,
+      "step": 2953
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 0.00022932692307692305,
+      "loss": 0.9777,
+      "step": 2954
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 0.0002292980769230769,
+      "loss": 0.8539,
+      "step": 2955
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 0.00022926923076923075,
+      "loss": 0.8057,
+      "step": 2956
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 0.0002292403846153846,
+      "loss": 1.0102,
+      "step": 2957
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 0.00022921153846153846,
+      "loss": 0.9091,
+      "step": 2958
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 0.00022918269230769227,
+      "loss": 0.9548,
+      "step": 2959
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 0.00022915384615384613,
+      "loss": 0.9048,
+      "step": 2960
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 0.000229125,
+      "loss": 0.9605,
+      "step": 2961
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 0.0002290961538461538,
+      "loss": 1.1223,
+      "step": 2962
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 0.00022906730769230768,
+      "loss": 0.9301,
+      "step": 2963
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 0.00022903846153846151,
+      "loss": 0.8287,
+      "step": 2964
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 0.00022900961538461535,
+      "loss": 1.0443,
+      "step": 2965
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 0.00022898076923076922,
+      "loss": 0.935,
+      "step": 2966
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 0.00022895192307692306,
+      "loss": 0.9898,
+      "step": 2967
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 0.00022892307692307692,
+      "loss": 0.9651,
+      "step": 2968
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 0.00022889423076923073,
+      "loss": 1.0671,
+      "step": 2969
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 0.0002288653846153846,
+      "loss": 0.8691,
+      "step": 2970
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 0.00022883653846153846,
+      "loss": 1.0132,
+      "step": 2971
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 0.00022880769230769227,
+      "loss": 0.8134,
+      "step": 2972
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 0.00022877884615384614,
+      "loss": 0.8147,
+      "step": 2973
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 0.00022874999999999998,
+      "loss": 1.0835,
+      "step": 2974
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 0.00022872115384615381,
+      "loss": 1.0073,
+      "step": 2975
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 0.00022869230769230768,
+      "loss": 0.9416,
+      "step": 2976
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 0.00022866346153846152,
+      "loss": 0.9659,
+      "step": 2977
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 0.00022863461538461538,
+      "loss": 0.9628,
+      "step": 2978
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 0.0002286057692307692,
+      "loss": 1.0587,
+      "step": 2979
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 0.00022857692307692306,
+      "loss": 1.079,
+      "step": 2980
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 0.00022854807692307692,
+      "loss": 0.9909,
+      "step": 2981
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 0.00022851923076923074,
+      "loss": 0.9755,
+      "step": 2982
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 0.0002284903846153846,
+      "loss": 0.8568,
+      "step": 2983
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 0.00022846153846153844,
+      "loss": 0.9668,
+      "step": 2984
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 0.00022843269230769228,
+      "loss": 0.9782,
+      "step": 2985
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 0.00022840384615384614,
+      "loss": 1.0285,
+      "step": 2986
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 0.00022837499999999998,
+      "loss": 1.0176,
+      "step": 2987
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 0.00022834615384615385,
+      "loss": 0.9582,
+      "step": 2988
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 0.00022831730769230766,
+      "loss": 0.8385,
+      "step": 2989
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 0.00022828846153846152,
+      "loss": 0.935,
+      "step": 2990
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 0.0002282596153846154,
+      "loss": 0.9632,
+      "step": 2991
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 0.0002282307692307692,
+      "loss": 1.1575,
+      "step": 2992
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 0.00022820192307692306,
+      "loss": 0.944,
+      "step": 2993
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 0.0002281730769230769,
+      "loss": 0.8061,
+      "step": 2994
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 0.00022814423076923074,
+      "loss": 0.9082,
+      "step": 2995
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 0.0002281153846153846,
+      "loss": 0.9103,
+      "step": 2996
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 0.00022808653846153844,
+      "loss": 0.9155,
+      "step": 2997
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 0.0002280576923076923,
+      "loss": 0.9781,
+      "step": 2998
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 0.00022802884615384612,
+      "loss": 1.0322,
+      "step": 2999
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 0.8303,
+      "step": 3000
+    },
+    {
+      "epoch": 27.52,
+      "eval_cer": 0.1291773489857419,
+      "eval_loss": 0.3826153576374054,
+      "eval_runtime": 13.9905,
+      "eval_samples_per_second": 117.722,
+      "eval_steps_per_second": 1.858,
+      "eval_wer": 0.511285874782964,
+      "step": 3000
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 0.00022797115384615385,
+      "loss": 0.9371,
+      "step": 3001
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 0.00022794230769230766,
+      "loss": 1.0306,
+      "step": 3002
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 0.00022791346153846153,
+      "loss": 0.8098,
+      "step": 3003
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 0.00022788461538461537,
+      "loss": 1.0815,
+      "step": 3004
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 0.0002278557692307692,
+      "loss": 0.8722,
+      "step": 3005
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 0.00022782692307692307,
+      "loss": 0.8735,
+      "step": 3006
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 0.0002277980769230769,
+      "loss": 0.8609,
+      "step": 3007
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 0.00022776923076923077,
+      "loss": 0.9097,
+      "step": 3008
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 0.00022774038461538458,
+      "loss": 0.9431,
+      "step": 3009
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 0.00022771153846153845,
+      "loss": 1.0031,
+      "step": 3010
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 0.00022768269230769226,
+      "loss": 1.1546,
+      "step": 3011
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 0.00022765384615384612,
+      "loss": 1.0253,
+      "step": 3012
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 0.000227625,
+      "loss": 0.9538,
+      "step": 3013
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 0.00022759615384615383,
+      "loss": 0.9732,
+      "step": 3014
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 0.00022756730769230767,
+      "loss": 0.845,
+      "step": 3015
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 0.0002275384615384615,
+      "loss": 0.8982,
+      "step": 3016
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 0.00022750961538461537,
+      "loss": 0.805,
+      "step": 3017
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 0.00022748076923076923,
+      "loss": 1.0698,
+      "step": 3018
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 0.00022745192307692305,
+      "loss": 1.0087,
+      "step": 3019
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 0.0002274230769230769,
+      "loss": 0.9586,
+      "step": 3020
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 0.00022739423076923072,
+      "loss": 0.9449,
+      "step": 3021
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 0.0002273653846153846,
+      "loss": 0.9317,
+      "step": 3022
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 0.00022733653846153845,
+      "loss": 0.9799,
+      "step": 3023
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 0.0002273076923076923,
+      "loss": 0.9691,
+      "step": 3024
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 0.00022727884615384613,
+      "loss": 0.9614,
+      "step": 3025
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 0.00022724999999999997,
+      "loss": 0.8883,
+      "step": 3026
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 0.00022722115384615383,
+      "loss": 0.9432,
+      "step": 3027
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 0.0002271923076923077,
+      "loss": 0.9993,
+      "step": 3028
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 0.0002271634615384615,
+      "loss": 1.0641,
+      "step": 3029
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 0.00022713461538461537,
+      "loss": 0.9059,
+      "step": 3030
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 0.00022710576923076919,
+      "loss": 0.9575,
+      "step": 3031
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 0.00022707692307692305,
+      "loss": 0.9937,
+      "step": 3032
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 0.00022704807692307692,
+      "loss": 1.0052,
+      "step": 3033
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 0.00022701923076923075,
+      "loss": 0.9299,
+      "step": 3034
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 0.0002269903846153846,
+      "loss": 0.8955,
+      "step": 3035
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 0.00022696153846153843,
+      "loss": 1.0806,
+      "step": 3036
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 0.0002269326923076923,
+      "loss": 1.0231,
+      "step": 3037
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 0.00022690384615384616,
+      "loss": 0.8701,
+      "step": 3038
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 0.00022687499999999997,
+      "loss": 1.0138,
+      "step": 3039
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 0.00022684615384615384,
+      "loss": 1.0629,
+      "step": 3040
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 0.00022681730769230765,
+      "loss": 0.9817,
+      "step": 3041
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 0.0002267884615384615,
+      "loss": 1.0438,
+      "step": 3042
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 0.00022675961538461538,
+      "loss": 0.9901,
+      "step": 3043
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 0.00022673076923076922,
+      "loss": 1.0345,
+      "step": 3044
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 0.00022670192307692305,
+      "loss": 0.948,
+      "step": 3045
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 0.0002266730769230769,
+      "loss": 0.9169,
+      "step": 3046
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 0.00022664423076923076,
+      "loss": 0.9273,
+      "step": 3047
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 0.0002266153846153846,
+      "loss": 0.9166,
+      "step": 3048
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 0.00022658653846153843,
+      "loss": 0.9921,
+      "step": 3049
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 0.0002265576923076923,
+      "loss": 1.0211,
+      "step": 3050
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 0.0002265288461538461,
+      "loss": 0.8871,
+      "step": 3051
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 0.00022649999999999998,
+      "loss": 0.917,
+      "step": 3052
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 0.00022647115384615384,
+      "loss": 0.9078,
+      "step": 3053
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 0.00022644230769230768,
+      "loss": 0.948,
+      "step": 3054
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 0.00022641346153846152,
+      "loss": 0.9405,
+      "step": 3055
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 0.00022638461538461536,
+      "loss": 0.9323,
+      "step": 3056
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 0.00022635576923076922,
+      "loss": 0.9968,
+      "step": 3057
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 0.00022632692307692306,
+      "loss": 0.7508,
+      "step": 3058
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 0.0002262980769230769,
+      "loss": 0.9427,
+      "step": 3059
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 0.00022626923076923076,
+      "loss": 1.0357,
+      "step": 3060
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 0.00022624038461538457,
+      "loss": 0.9963,
+      "step": 3061
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 0.00022621153846153844,
+      "loss": 0.9919,
+      "step": 3062
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 0.0002261826923076923,
+      "loss": 1.0005,
+      "step": 3063
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 0.00022615384615384614,
+      "loss": 0.841,
+      "step": 3064
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 0.00022612499999999998,
+      "loss": 1.104,
+      "step": 3065
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 0.00022609615384615382,
+      "loss": 1.024,
+      "step": 3066
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 0.00022606730769230768,
+      "loss": 1.1057,
+      "step": 3067
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 0.00022603846153846152,
+      "loss": 0.9633,
+      "step": 3068
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 0.00022600961538461536,
+      "loss": 0.9147,
+      "step": 3069
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 0.00022598076923076923,
+      "loss": 1.1779,
+      "step": 3070
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 0.00022595192307692304,
+      "loss": 0.8819,
+      "step": 3071
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 0.0002259230769230769,
+      "loss": 0.8559,
+      "step": 3072
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 0.00022589423076923074,
+      "loss": 0.8864,
+      "step": 3073
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 0.0002258653846153846,
+      "loss": 0.8502,
+      "step": 3074
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 0.00022583653846153844,
+      "loss": 0.8928,
+      "step": 3075
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 0.00022580769230769228,
+      "loss": 1.113,
+      "step": 3076
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 0.00022577884615384615,
+      "loss": 1.119,
+      "step": 3077
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 0.00022574999999999996,
+      "loss": 0.9957,
+      "step": 3078
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 0.00022572115384615382,
+      "loss": 0.9449,
+      "step": 3079
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 0.0002256923076923077,
+      "loss": 0.9083,
+      "step": 3080
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 0.0002256634615384615,
+      "loss": 0.8866,
+      "step": 3081
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 0.00022563461538461536,
+      "loss": 1.0047,
+      "step": 3082
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 0.0002256057692307692,
+      "loss": 1.0268,
+      "step": 3083
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 0.00022557692307692307,
+      "loss": 1.1004,
+      "step": 3084
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 0.0002255480769230769,
+      "loss": 1.1434,
+      "step": 3085
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 0.00022551923076923074,
+      "loss": 0.9057,
+      "step": 3086
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 0.0002254903846153846,
+      "loss": 0.98,
+      "step": 3087
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 0.00022546153846153842,
+      "loss": 1.0577,
+      "step": 3088
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 0.00022543269230769229,
+      "loss": 0.9055,
+      "step": 3089
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 0.00022540384615384615,
+      "loss": 1.0036,
+      "step": 3090
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 0.00022537499999999996,
+      "loss": 0.9119,
+      "step": 3091
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 0.00022534615384615383,
+      "loss": 0.9817,
+      "step": 3092
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 0.00022531730769230767,
+      "loss": 1.0482,
+      "step": 3093
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 0.00022528846153846153,
+      "loss": 0.9497,
+      "step": 3094
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 0.00022525961538461537,
+      "loss": 1.0835,
+      "step": 3095
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 0.0002252307692307692,
+      "loss": 1.0393,
+      "step": 3096
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 0.00022520192307692307,
+      "loss": 1.047,
+      "step": 3097
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 0.00022517307692307688,
+      "loss": 0.9291,
+      "step": 3098
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 0.00022514423076923075,
+      "loss": 0.9218,
+      "step": 3099
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 0.00022511538461538461,
+      "loss": 0.9685,
+      "step": 3100
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 0.00022508653846153843,
+      "loss": 0.968,
+      "step": 3101
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 0.0002250576923076923,
+      "loss": 0.8528,
+      "step": 3102
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 0.00022502884615384613,
+      "loss": 0.9814,
+      "step": 3103
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 0.000225,
+      "loss": 0.8852,
+      "step": 3104
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 0.00022497115384615383,
+      "loss": 0.9165,
+      "step": 3105
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 0.00022494230769230767,
+      "loss": 0.9667,
+      "step": 3106
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 0.00022491346153846154,
+      "loss": 0.9455,
+      "step": 3107
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 0.00022488461538461535,
+      "loss": 1.0139,
+      "step": 3108
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 0.0002248557692307692,
+      "loss": 0.9738,
+      "step": 3109
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 0.00022482692307692308,
+      "loss": 0.9373,
+      "step": 3110
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 0.0002247980769230769,
+      "loss": 0.8958,
+      "step": 3111
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 0.00022476923076923075,
+      "loss": 0.9053,
+      "step": 3112
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 0.0002247403846153846,
+      "loss": 0.991,
+      "step": 3113
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 0.00022471153846153843,
+      "loss": 1.0854,
+      "step": 3114
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 0.0002246826923076923,
+      "loss": 0.9438,
+      "step": 3115
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 0.00022465384615384613,
+      "loss": 1.0953,
+      "step": 3116
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 0.000224625,
+      "loss": 0.8477,
+      "step": 3117
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 0.0002245961538461538,
+      "loss": 0.9203,
+      "step": 3118
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 0.00022456730769230767,
+      "loss": 1.0566,
+      "step": 3119
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 0.00022453846153846154,
+      "loss": 1.1361,
+      "step": 3120
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 0.00022450961538461535,
+      "loss": 0.899,
+      "step": 3121
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 0.00022448076923076922,
+      "loss": 0.9306,
+      "step": 3122
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 0.00022445192307692305,
+      "loss": 0.9024,
+      "step": 3123
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 0.0002244230769230769,
+      "loss": 0.8481,
+      "step": 3124
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 0.00022439423076923076,
+      "loss": 1.0204,
+      "step": 3125
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 0.0002243653846153846,
+      "loss": 0.9457,
+      "step": 3126
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 0.00022433653846153846,
+      "loss": 0.8723,
+      "step": 3127
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 0.00022430769230769227,
+      "loss": 0.871,
+      "step": 3128
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 0.00022427884615384614,
+      "loss": 0.8401,
+      "step": 3129
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 0.00022425,
+      "loss": 0.9749,
+      "step": 3130
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 0.00022422115384615381,
+      "loss": 1.0262,
+      "step": 3131
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 0.00022419230769230768,
+      "loss": 0.901,
+      "step": 3132
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 0.00022416346153846152,
+      "loss": 0.8949,
+      "step": 3133
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 0.00022413461538461536,
+      "loss": 1.0227,
+      "step": 3134
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 0.00022410576923076922,
+      "loss": 0.8931,
+      "step": 3135
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 0.00022407692307692306,
+      "loss": 0.7696,
+      "step": 3136
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 0.00022404807692307692,
+      "loss": 0.9435,
+      "step": 3137
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 0.00022401923076923074,
+      "loss": 0.9795,
+      "step": 3138
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 0.0002239903846153846,
+      "loss": 0.9354,
+      "step": 3139
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 0.00022396153846153844,
+      "loss": 0.9624,
+      "step": 3140
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 0.00022393269230769228,
+      "loss": 0.8631,
+      "step": 3141
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 0.00022390384615384614,
+      "loss": 0.9194,
+      "step": 3142
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 0.00022387499999999998,
+      "loss": 1.0486,
+      "step": 3143
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 0.00022384615384615382,
+      "loss": 1.0201,
+      "step": 3144
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 0.00022381730769230766,
+      "loss": 1.0055,
+      "step": 3145
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 0.00022378846153846152,
+      "loss": 0.9703,
+      "step": 3146
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 0.0002237596153846154,
+      "loss": 0.9907,
+      "step": 3147
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 0.0002237307692307692,
+      "loss": 0.894,
+      "step": 3148
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 0.00022370192307692306,
+      "loss": 0.9643,
+      "step": 3149
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 0.0002236730769230769,
+      "loss": 0.9714,
+      "step": 3150
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 0.00022364423076923074,
+      "loss": 0.9227,
+      "step": 3151
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 0.0002236153846153846,
+      "loss": 1.0777,
+      "step": 3152
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 0.00022358653846153844,
+      "loss": 0.9636,
+      "step": 3153
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 0.00022355769230769228,
+      "loss": 0.9241,
+      "step": 3154
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 0.00022352884615384612,
+      "loss": 0.8716,
+      "step": 3155
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 0.00022349999999999998,
+      "loss": 0.9346,
+      "step": 3156
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 0.00022347115384615385,
+      "loss": 1.0162,
+      "step": 3157
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 0.00022344230769230766,
+      "loss": 0.943,
+      "step": 3158
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 0.00022341346153846153,
+      "loss": 0.9814,
+      "step": 3159
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 0.00022338461538461536,
+      "loss": 0.9964,
+      "step": 3160
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 0.0002233557692307692,
+      "loss": 0.8529,
+      "step": 3161
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 0.00022332692307692307,
+      "loss": 0.8645,
+      "step": 3162
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 0.0002232980769230769,
+      "loss": 0.8783,
+      "step": 3163
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 0.00022326923076923074,
+      "loss": 0.9953,
+      "step": 3164
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 0.00022324038461538458,
+      "loss": 0.8863,
+      "step": 3165
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 0.00022321153846153845,
+      "loss": 1.0506,
+      "step": 3166
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 0.0002231826923076923,
+      "loss": 0.9968,
+      "step": 3167
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 0.00022315384615384612,
+      "loss": 1.028,
+      "step": 3168
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 0.000223125,
+      "loss": 0.9033,
+      "step": 3169
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 0.0002230961538461538,
+      "loss": 0.8539,
+      "step": 3170
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 0.00022306730769230767,
+      "loss": 1.004,
+      "step": 3171
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 0.00022303846153846153,
+      "loss": 0.8845,
+      "step": 3172
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 0.00022300961538461537,
+      "loss": 1.0703,
+      "step": 3173
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 0.0002229807692307692,
+      "loss": 0.8963,
+      "step": 3174
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 0.00022295192307692304,
+      "loss": 0.8734,
+      "step": 3175
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 0.0002229230769230769,
+      "loss": 0.8374,
+      "step": 3176
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 0.00022289423076923078,
+      "loss": 1.0067,
+      "step": 3177
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 0.0002228653846153846,
+      "loss": 0.9271,
+      "step": 3178
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 0.00022283653846153845,
+      "loss": 0.8847,
+      "step": 3179
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 0.00022280769230769226,
+      "loss": 1.0037,
+      "step": 3180
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 0.00022277884615384613,
+      "loss": 0.7878,
+      "step": 3181
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 0.00022275,
+      "loss": 0.907,
+      "step": 3182
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 0.00022272115384615383,
+      "loss": 0.9871,
+      "step": 3183
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 0.00022269230769230767,
+      "loss": 0.9952,
+      "step": 3184
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 0.0002226634615384615,
+      "loss": 0.975,
+      "step": 3185
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 0.00022263461538461537,
+      "loss": 0.9654,
+      "step": 3186
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 0.00022260576923076924,
+      "loss": 0.975,
+      "step": 3187
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 0.00022257692307692305,
+      "loss": 0.9374,
+      "step": 3188
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 0.00022254807692307691,
+      "loss": 0.9479,
+      "step": 3189
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 0.00022251923076923073,
+      "loss": 0.9324,
+      "step": 3190
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 0.0002224903846153846,
+      "loss": 0.9731,
+      "step": 3191
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 0.00022246153846153846,
+      "loss": 0.9555,
+      "step": 3192
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 0.0002224326923076923,
+      "loss": 0.934,
+      "step": 3193
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 0.00022240384615384613,
+      "loss": 0.9087,
+      "step": 3194
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 0.00022237499999999997,
+      "loss": 0.8588,
+      "step": 3195
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 0.00022234615384615384,
+      "loss": 0.9476,
+      "step": 3196
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 0.0002223173076923077,
+      "loss": 1.0057,
+      "step": 3197
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 0.0002222884615384615,
+      "loss": 0.9438,
+      "step": 3198
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 0.00022225961538461538,
+      "loss": 1.1128,
+      "step": 3199
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 0.0002222307692307692,
+      "loss": 1.003,
+      "step": 3200
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 0.00022220192307692305,
+      "loss": 0.9601,
+      "step": 3201
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 0.0002221730769230769,
+      "loss": 0.778,
+      "step": 3202
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 0.00022214423076923076,
+      "loss": 0.9301,
+      "step": 3203
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 0.0002221153846153846,
+      "loss": 1.0209,
+      "step": 3204
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 0.00022208653846153843,
+      "loss": 1.0406,
+      "step": 3205
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 0.0002220576923076923,
+      "loss": 0.8925,
+      "step": 3206
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 0.0002220288461538461,
+      "loss": 0.9487,
+      "step": 3207
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 0.9887,
+      "step": 3208
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 0.00022197115384615384,
+      "loss": 0.9249,
+      "step": 3209
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 0.00022194230769230765,
+      "loss": 1.029,
+      "step": 3210
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 0.00022191346153846152,
+      "loss": 0.8838,
+      "step": 3211
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 0.00022188461538461535,
+      "loss": 0.917,
+      "step": 3212
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 0.00022185576923076922,
+      "loss": 0.9227,
+      "step": 3213
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 0.00022182692307692306,
+      "loss": 0.8965,
+      "step": 3214
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 0.0002217980769230769,
+      "loss": 0.9454,
+      "step": 3215
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 0.00022176923076923076,
+      "loss": 0.9264,
+      "step": 3216
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 0.00022174038461538457,
+      "loss": 1.0501,
+      "step": 3217
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 0.00022171153846153844,
+      "loss": 1.0465,
+      "step": 3218
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 0.0002216826923076923,
+      "loss": 0.9777,
+      "step": 3219
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 0.00022165384615384611,
+      "loss": 0.8989,
+      "step": 3220
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 0.00022162499999999998,
+      "loss": 0.8916,
+      "step": 3221
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 0.00022159615384615382,
+      "loss": 0.8909,
+      "step": 3222
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 0.00022156730769230768,
+      "loss": 1.109,
+      "step": 3223
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 0.00022153846153846152,
+      "loss": 0.9439,
+      "step": 3224
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 0.00022150961538461536,
+      "loss": 0.9819,
+      "step": 3225
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 0.00022148076923076922,
+      "loss": 0.9402,
+      "step": 3226
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 0.00022145192307692304,
+      "loss": 0.9643,
+      "step": 3227
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 0.0002214230769230769,
+      "loss": 0.9864,
+      "step": 3228
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 0.00022139423076923077,
+      "loss": 0.9247,
+      "step": 3229
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 0.00022136538461538458,
+      "loss": 0.8681,
+      "step": 3230
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 0.00022133653846153844,
+      "loss": 0.9308,
+      "step": 3231
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 0.00022130769230769228,
+      "loss": 0.9415,
+      "step": 3232
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 0.00022127884615384615,
+      "loss": 1.1058,
+      "step": 3233
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 0.00022124999999999998,
+      "loss": 1.1587,
+      "step": 3234
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 0.00022122115384615382,
+      "loss": 0.8801,
+      "step": 3235
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 0.0002211923076923077,
+      "loss": 1.0978,
+      "step": 3236
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 0.0002211634615384615,
+      "loss": 0.9129,
+      "step": 3237
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 0.00022113461538461536,
+      "loss": 0.9063,
+      "step": 3238
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 0.00022110576923076923,
+      "loss": 0.9602,
+      "step": 3239
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 0.00022107692307692304,
+      "loss": 0.9003,
+      "step": 3240
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 0.0002210480769230769,
+      "loss": 0.9113,
+      "step": 3241
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 0.00022101923076923074,
+      "loss": 0.9871,
+      "step": 3242
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 0.0002209903846153846,
+      "loss": 0.9173,
+      "step": 3243
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 0.00022096153846153845,
+      "loss": 0.9554,
+      "step": 3244
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 0.00022093269230769229,
+      "loss": 0.9655,
+      "step": 3245
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 0.00022090384615384615,
+      "loss": 1.0128,
+      "step": 3246
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 0.00022087499999999996,
+      "loss": 0.9916,
+      "step": 3247
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 0.00022084615384615383,
+      "loss": 1.0183,
+      "step": 3248
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 0.0002208173076923077,
+      "loss": 0.9293,
+      "step": 3249
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 0.0002207884615384615,
+      "loss": 1.0386,
+      "step": 3250
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 0.00022075961538461537,
+      "loss": 0.9236,
+      "step": 3251
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 0.0002207307692307692,
+      "loss": 0.9829,
+      "step": 3252
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 0.00022070192307692307,
+      "loss": 0.8578,
+      "step": 3253
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 0.0002206730769230769,
+      "loss": 0.9862,
+      "step": 3254
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 0.00022064423076923075,
+      "loss": 1.1676,
+      "step": 3255
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 0.0002206153846153846,
+      "loss": 0.8895,
+      "step": 3256
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 0.00022058653846153842,
+      "loss": 0.9057,
+      "step": 3257
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 0.0002205576923076923,
+      "loss": 0.8799,
+      "step": 3258
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 0.00022052884615384615,
+      "loss": 0.8892,
+      "step": 3259
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 0.00022049999999999997,
+      "loss": 1.0499,
+      "step": 3260
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 0.00022047115384615383,
+      "loss": 0.9461,
+      "step": 3261
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 0.00022044230769230767,
+      "loss": 0.9317,
+      "step": 3262
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 0.00022041346153846153,
+      "loss": 0.9001,
+      "step": 3263
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 0.00022038461538461537,
+      "loss": 0.9394,
+      "step": 3264
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 0.0002203557692307692,
+      "loss": 1.0294,
+      "step": 3265
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 0.00022032692307692308,
+      "loss": 0.9188,
+      "step": 3266
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 0.0002202980769230769,
+      "loss": 0.9253,
+      "step": 3267
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 0.00022026923076923075,
+      "loss": 0.8548,
+      "step": 3268
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 0.0002202403846153846,
+      "loss": 0.9814,
+      "step": 3269
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 0.00022021153846153843,
+      "loss": 0.8516,
+      "step": 3270
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 0.0002201826923076923,
+      "loss": 0.9711,
+      "step": 3271
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 0.00022015384615384613,
+      "loss": 0.8726,
+      "step": 3272
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 0.000220125,
+      "loss": 0.9063,
+      "step": 3273
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 0.0002200961538461538,
+      "loss": 0.9909,
+      "step": 3274
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 0.00022006730769230767,
+      "loss": 1.0041,
+      "step": 3275
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 0.00022003846153846154,
+      "loss": 0.9771,
+      "step": 3276
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 0.00022000961538461535,
+      "loss": 1.0173,
+      "step": 3277
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 0.00021998076923076922,
+      "loss": 0.9226,
+      "step": 3278
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 0.00021995192307692305,
+      "loss": 0.9286,
+      "step": 3279
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 0.0002199230769230769,
+      "loss": 0.9513,
+      "step": 3280
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 0.00021989423076923076,
+      "loss": 1.0628,
+      "step": 3281
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 0.0002198653846153846,
+      "loss": 1.0159,
+      "step": 3282
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 0.00021983653846153846,
+      "loss": 0.9596,
+      "step": 3283
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 0.00021980769230769227,
+      "loss": 0.8686,
+      "step": 3284
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 0.00021977884615384614,
+      "loss": 0.9232,
+      "step": 3285
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 0.00021975,
+      "loss": 0.8245,
+      "step": 3286
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 0.0002197211538461538,
+      "loss": 1.0071,
+      "step": 3287
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 0.00021969230769230768,
+      "loss": 0.8954,
+      "step": 3288
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 0.00021966346153846152,
+      "loss": 1.0508,
+      "step": 3289
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 0.00021963461538461535,
+      "loss": 0.9011,
+      "step": 3290
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 0.00021960576923076922,
+      "loss": 0.8618,
+      "step": 3291
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 0.00021957692307692306,
+      "loss": 0.9146,
+      "step": 3292
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 0.00021954807692307692,
+      "loss": 0.8078,
+      "step": 3293
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 0.00021951923076923073,
+      "loss": 0.929,
+      "step": 3294
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 0.0002194903846153846,
+      "loss": 1.016,
+      "step": 3295
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 0.00021946153846153846,
+      "loss": 0.9713,
+      "step": 3296
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 0.00021943269230769228,
+      "loss": 0.9012,
+      "step": 3297
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 0.00021940384615384614,
+      "loss": 0.8463,
+      "step": 3298
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 0.00021937499999999998,
+      "loss": 0.9435,
+      "step": 3299
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 0.00021934615384615382,
+      "loss": 0.9525,
+      "step": 3300
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 0.00021931730769230768,
+      "loss": 0.9866,
+      "step": 3301
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 0.00021928846153846152,
+      "loss": 0.9363,
+      "step": 3302
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 0.00021925961538461539,
+      "loss": 0.8369,
+      "step": 3303
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 0.0002192307692307692,
+      "loss": 1.0053,
+      "step": 3304
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 0.00021920192307692306,
+      "loss": 0.8238,
+      "step": 3305
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 0.00021917307692307693,
+      "loss": 0.8688,
+      "step": 3306
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 0.00021914423076923074,
+      "loss": 0.9183,
+      "step": 3307
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 0.0002191153846153846,
+      "loss": 0.9115,
+      "step": 3308
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 0.00021908653846153844,
+      "loss": 0.8552,
+      "step": 3309
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 0.00021905769230769228,
+      "loss": 0.9659,
+      "step": 3310
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 0.00021902884615384615,
+      "loss": 0.8646,
+      "step": 3311
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 0.9038,
+      "step": 3312
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 0.00021897115384615382,
+      "loss": 0.9283,
+      "step": 3313
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 0.00021894230769230766,
+      "loss": 0.8992,
+      "step": 3314
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 0.00021891346153846153,
+      "loss": 0.8547,
+      "step": 3315
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 0.0002188846153846154,
+      "loss": 0.8672,
+      "step": 3316
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 0.0002188557692307692,
+      "loss": 0.8524,
+      "step": 3317
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 0.00021882692307692307,
+      "loss": 0.9291,
+      "step": 3318
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 0.0002187980769230769,
+      "loss": 1.0914,
+      "step": 3319
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 0.00021876923076923074,
+      "loss": 0.782,
+      "step": 3320
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 0.0002187403846153846,
+      "loss": 0.8971,
+      "step": 3321
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 0.00021871153846153845,
+      "loss": 0.8976,
+      "step": 3322
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 0.00021868269230769228,
+      "loss": 0.9429,
+      "step": 3323
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 0.00021865384615384612,
+      "loss": 0.9436,
+      "step": 3324
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 0.000218625,
+      "loss": 0.9535,
+      "step": 3325
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 0.00021859615384615385,
+      "loss": 0.9444,
+      "step": 3326
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 0.00021856730769230766,
+      "loss": 0.7903,
+      "step": 3327
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 0.00021853846153846153,
+      "loss": 0.8579,
+      "step": 3328
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 0.00021850961538461537,
+      "loss": 0.9128,
+      "step": 3329
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 0.0002184807692307692,
+      "loss": 0.9136,
+      "step": 3330
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 0.00021845192307692307,
+      "loss": 1.008,
+      "step": 3331
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 0.0002184230769230769,
+      "loss": 1.0564,
+      "step": 3332
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 0.00021839423076923075,
+      "loss": 0.8982,
+      "step": 3333
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 0.00021836538461538459,
+      "loss": 0.8815,
+      "step": 3334
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 0.00021833653846153845,
+      "loss": 0.9659,
+      "step": 3335
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 0.00021830769230769226,
+      "loss": 0.8271,
+      "step": 3336
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 0.00021827884615384613,
+      "loss": 0.7842,
+      "step": 3337
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 0.00021825,
+      "loss": 1.0113,
+      "step": 3338
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 0.00021822115384615383,
+      "loss": 1.0384,
+      "step": 3339
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 0.00021819230769230767,
+      "loss": 0.8973,
+      "step": 3340
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 0.0002181634615384615,
+      "loss": 0.9277,
+      "step": 3341
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 0.00021813461538461537,
+      "loss": 0.9369,
+      "step": 3342
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 0.0002181057692307692,
+      "loss": 0.9693,
+      "step": 3343
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 0.00021807692307692305,
+      "loss": 1.0343,
+      "step": 3344
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 0.00021804807692307691,
+      "loss": 0.9635,
+      "step": 3345
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 0.00021801923076923072,
+      "loss": 0.8427,
+      "step": 3346
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 0.0002179903846153846,
+      "loss": 0.8949,
+      "step": 3347
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 0.00021796153846153846,
+      "loss": 0.8671,
+      "step": 3348
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 0.0002179326923076923,
+      "loss": 0.9462,
+      "step": 3349
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 0.00021790384615384613,
+      "loss": 0.9502,
+      "step": 3350
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 0.00021787499999999997,
+      "loss": 0.9486,
+      "step": 3351
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 0.00021784615384615383,
+      "loss": 1.0203,
+      "step": 3352
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 0.00021781730769230767,
+      "loss": 0.9484,
+      "step": 3353
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 0.0002177884615384615,
+      "loss": 0.9509,
+      "step": 3354
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 0.00021775961538461538,
+      "loss": 0.9556,
+      "step": 3355
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 0.0002177307692307692,
+      "loss": 0.96,
+      "step": 3356
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 0.00021770192307692305,
+      "loss": 1.0007,
+      "step": 3357
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 0.00021767307692307692,
+      "loss": 0.9066,
+      "step": 3358
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 0.00021764423076923076,
+      "loss": 0.9388,
+      "step": 3359
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 0.0002176153846153846,
+      "loss": 0.9135,
+      "step": 3360
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 0.00021758653846153843,
+      "loss": 0.8616,
+      "step": 3361
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 0.0002175576923076923,
+      "loss": 0.9232,
+      "step": 3362
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 0.00021752884615384614,
+      "loss": 0.9167,
+      "step": 3363
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 0.00021749999999999997,
+      "loss": 1.031,
+      "step": 3364
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 0.00021747115384615384,
+      "loss": 1.0269,
+      "step": 3365
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 0.00021744230769230765,
+      "loss": 1.0596,
+      "step": 3366
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 0.00021741346153846152,
+      "loss": 0.944,
+      "step": 3367
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 0.00021738461538461538,
+      "loss": 0.9757,
+      "step": 3368
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 0.0002173557692307692,
+      "loss": 0.9817,
+      "step": 3369
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 0.00021732692307692306,
+      "loss": 0.9663,
+      "step": 3370
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 0.0002172980769230769,
+      "loss": 0.8544,
+      "step": 3371
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 0.00021726923076923076,
+      "loss": 0.8684,
+      "step": 3372
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 0.0002172403846153846,
+      "loss": 0.9988,
+      "step": 3373
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 0.00021721153846153844,
+      "loss": 1.05,
+      "step": 3374
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 0.0002171826923076923,
+      "loss": 1.0442,
+      "step": 3375
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 0.0002171538461538461,
+      "loss": 1.0469,
+      "step": 3376
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 0.00021712499999999998,
+      "loss": 0.9785,
+      "step": 3377
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 0.00021709615384615384,
+      "loss": 0.9265,
+      "step": 3378
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 0.00021706730769230765,
+      "loss": 1.0739,
+      "step": 3379
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 0.00021703846153846152,
+      "loss": 0.8358,
+      "step": 3380
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 0.00021700961538461536,
+      "loss": 0.8594,
+      "step": 3381
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 0.00021698076923076922,
+      "loss": 0.9178,
+      "step": 3382
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 0.00021695192307692306,
+      "loss": 0.8037,
+      "step": 3383
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 0.0002169230769230769,
+      "loss": 0.8282,
+      "step": 3384
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 0.00021689423076923077,
+      "loss": 0.8014,
+      "step": 3385
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 0.00021686538461538458,
+      "loss": 0.9911,
+      "step": 3386
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 0.00021683653846153844,
+      "loss": 0.9523,
+      "step": 3387
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 0.0002168076923076923,
+      "loss": 0.8778,
+      "step": 3388
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 0.00021677884615384612,
+      "loss": 0.8736,
+      "step": 3389
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 0.00021674999999999998,
+      "loss": 1.0523,
+      "step": 3390
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 0.00021672115384615382,
+      "loss": 0.8153,
+      "step": 3391
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 0.0002166923076923077,
+      "loss": 0.8465,
+      "step": 3392
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 0.00021666346153846152,
+      "loss": 0.8027,
+      "step": 3393
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 0.00021663461538461536,
+      "loss": 0.9501,
+      "step": 3394
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 0.00021660576923076923,
+      "loss": 0.9837,
+      "step": 3395
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 0.00021657692307692304,
+      "loss": 0.9287,
+      "step": 3396
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 0.0002165480769230769,
+      "loss": 0.9199,
+      "step": 3397
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 0.00021651923076923074,
+      "loss": 0.9575,
+      "step": 3398
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 0.00021649038461538458,
+      "loss": 0.8729,
+      "step": 3399
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 0.00021646153846153845,
+      "loss": 0.9884,
+      "step": 3400
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 0.00021643269230769228,
+      "loss": 0.8354,
+      "step": 3401
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 0.00021640384615384615,
+      "loss": 0.9697,
+      "step": 3402
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 0.00021637499999999996,
+      "loss": 0.9673,
+      "step": 3403
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 0.00021634615384615383,
+      "loss": 1.0582,
+      "step": 3404
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 0.0002163173076923077,
+      "loss": 0.9375,
+      "step": 3405
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 0.0002162884615384615,
+      "loss": 0.8416,
+      "step": 3406
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 0.00021625961538461537,
+      "loss": 0.8373,
+      "step": 3407
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 0.0002162307692307692,
+      "loss": 0.9094,
+      "step": 3408
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 0.00021620192307692304,
+      "loss": 0.9301,
+      "step": 3409
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 0.0002161730769230769,
+      "loss": 0.9503,
+      "step": 3410
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 0.00021614423076923075,
+      "loss": 0.7841,
+      "step": 3411
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 0.0002161153846153846,
+      "loss": 0.9377,
+      "step": 3412
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 0.00021608653846153842,
+      "loss": 0.9633,
+      "step": 3413
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 0.0002160576923076923,
+      "loss": 0.8814,
+      "step": 3414
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 0.00021602884615384615,
+      "loss": 0.9908,
+      "step": 3415
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 0.9547,
+      "step": 3416
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 0.00021597115384615383,
+      "loss": 0.9338,
+      "step": 3417
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 0.00021594230769230767,
+      "loss": 0.8919,
+      "step": 3418
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 0.0002159134615384615,
+      "loss": 0.8111,
+      "step": 3419
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 0.00021588461538461537,
+      "loss": 0.8826,
+      "step": 3420
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 0.0002158557692307692,
+      "loss": 1.0464,
+      "step": 3421
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 0.00021582692307692308,
+      "loss": 0.9507,
+      "step": 3422
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 0.00021579807692307689,
+      "loss": 0.9692,
+      "step": 3423
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 0.00021576923076923075,
+      "loss": 0.8613,
+      "step": 3424
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 0.00021574038461538462,
+      "loss": 0.9092,
+      "step": 3425
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 0.00021571153846153843,
+      "loss": 0.9423,
+      "step": 3426
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 0.0002156826923076923,
+      "loss": 0.9321,
+      "step": 3427
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 0.00021565384615384613,
+      "loss": 0.8623,
+      "step": 3428
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 0.00021562499999999997,
+      "loss": 1.0136,
+      "step": 3429
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 0.00021559615384615383,
+      "loss": 0.9251,
+      "step": 3430
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 0.00021556730769230767,
+      "loss": 0.9008,
+      "step": 3431
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 0.00021553846153846154,
+      "loss": 0.7691,
+      "step": 3432
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 0.00021550961538461535,
+      "loss": 1.0001,
+      "step": 3433
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 0.00021548076923076921,
+      "loss": 0.95,
+      "step": 3434
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 0.00021545192307692308,
+      "loss": 1.0535,
+      "step": 3435
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 0.0002154230769230769,
+      "loss": 0.9233,
+      "step": 3436
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 0.00021539423076923076,
+      "loss": 0.8563,
+      "step": 3437
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 0.0002153653846153846,
+      "loss": 0.8348,
+      "step": 3438
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 0.00021533653846153843,
+      "loss": 1.0025,
+      "step": 3439
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 0.0002153076923076923,
+      "loss": 0.8919,
+      "step": 3440
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 0.00021527884615384614,
+      "loss": 0.8366,
+      "step": 3441
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 0.00021525,
+      "loss": 0.9725,
+      "step": 3442
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 0.0002152211538461538,
+      "loss": 0.9703,
+      "step": 3443
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 0.00021519230769230768,
+      "loss": 0.86,
+      "step": 3444
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 0.00021516346153846154,
+      "loss": 0.7775,
+      "step": 3445
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 0.00021513461538461535,
+      "loss": 1.0972,
+      "step": 3446
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 0.00021510576923076922,
+      "loss": 0.969,
+      "step": 3447
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 0.00021507692307692306,
+      "loss": 0.9191,
+      "step": 3448
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 0.0002150480769230769,
+      "loss": 0.9328,
+      "step": 3449
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 0.00021501923076923076,
+      "loss": 0.8426,
+      "step": 3450
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 0.0002149903846153846,
+      "loss": 0.869,
+      "step": 3451
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 0.00021496153846153846,
+      "loss": 0.9653,
+      "step": 3452
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 0.00021493269230769227,
+      "loss": 0.8919,
+      "step": 3453
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 0.00021490384615384614,
+      "loss": 0.9931,
+      "step": 3454
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 0.000214875,
+      "loss": 0.9465,
+      "step": 3455
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 0.00021484615384615382,
+      "loss": 0.9266,
+      "step": 3456
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 0.00021481730769230768,
+      "loss": 0.9317,
+      "step": 3457
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 0.00021478846153846152,
+      "loss": 1.0327,
+      "step": 3458
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 0.00021475961538461536,
+      "loss": 0.9928,
+      "step": 3459
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 0.00021473076923076922,
+      "loss": 0.9201,
+      "step": 3460
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 0.00021470192307692306,
+      "loss": 0.9222,
+      "step": 3461
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 0.00021467307692307693,
+      "loss": 0.9159,
+      "step": 3462
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 0.00021464423076923074,
+      "loss": 0.9962,
+      "step": 3463
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 0.0002146153846153846,
+      "loss": 0.9503,
+      "step": 3464
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 0.00021458653846153841,
+      "loss": 0.7989,
+      "step": 3465
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 0.00021455769230769228,
+      "loss": 0.9373,
+      "step": 3466
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 0.00021452884615384614,
+      "loss": 0.9008,
+      "step": 3467
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 0.00021449999999999998,
+      "loss": 0.986,
+      "step": 3468
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 0.00021447115384615382,
+      "loss": 0.9381,
+      "step": 3469
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 0.00021444230769230766,
+      "loss": 0.8374,
+      "step": 3470
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 0.00021441346153846152,
+      "loss": 0.8484,
+      "step": 3471
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 0.0002143846153846154,
+      "loss": 1.0003,
+      "step": 3472
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 0.0002143557692307692,
+      "loss": 0.9891,
+      "step": 3473
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 0.00021432692307692307,
+      "loss": 0.7966,
+      "step": 3474
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 0.00021429807692307688,
+      "loss": 1.0088,
+      "step": 3475
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 0.00021426923076923074,
+      "loss": 0.8645,
+      "step": 3476
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 0.0002142403846153846,
+      "loss": 0.7974,
+      "step": 3477
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 0.00021421153846153845,
+      "loss": 0.909,
+      "step": 3478
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 0.00021418269230769228,
+      "loss": 1.048,
+      "step": 3479
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 0.00021415384615384612,
+      "loss": 0.7916,
+      "step": 3480
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 0.000214125,
+      "loss": 1.0347,
+      "step": 3481
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 0.00021409615384615385,
+      "loss": 0.7833,
+      "step": 3482
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 0.00021406730769230766,
+      "loss": 0.8942,
+      "step": 3483
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 0.00021403846153846153,
+      "loss": 1.0734,
+      "step": 3484
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 0.00021400961538461534,
+      "loss": 0.8904,
+      "step": 3485
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 0.0002139807692307692,
+      "loss": 0.9139,
+      "step": 3486
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 0.00021395192307692307,
+      "loss": 1.0524,
+      "step": 3487
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 0.0002139230769230769,
+      "loss": 1.0062,
+      "step": 3488
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 0.00021389423076923075,
+      "loss": 0.8252,
+      "step": 3489
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 0.00021386538461538458,
+      "loss": 0.9026,
+      "step": 3490
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 0.00021383653846153845,
+      "loss": 0.7774,
+      "step": 3491
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 0.00021380769230769232,
+      "loss": 1.0065,
+      "step": 3492
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 0.00021377884615384613,
+      "loss": 0.9947,
+      "step": 3493
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 0.00021375,
+      "loss": 0.9454,
+      "step": 3494
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 0.0002137211538461538,
+      "loss": 0.9411,
+      "step": 3495
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 0.00021369230769230767,
+      "loss": 0.935,
+      "step": 3496
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 0.00021366346153846153,
+      "loss": 0.8842,
+      "step": 3497
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 0.00021363461538461537,
+      "loss": 0.7986,
+      "step": 3498
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 0.0002136057692307692,
+      "loss": 0.9036,
+      "step": 3499
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 0.00021357692307692305,
+      "loss": 0.9115,
+      "step": 3500
+    },
+    {
+      "epoch": 32.11,
+      "eval_cer": 0.10704808168038502,
+      "eval_loss": 0.38187265396118164,
+      "eval_runtime": 13.9362,
+      "eval_samples_per_second": 118.181,
+      "eval_steps_per_second": 1.866,
+      "eval_wer": 0.43243795322234707,
+      "step": 3500
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 0.0002135480769230769,
+      "loss": 1.0335,
+      "step": 3501
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 0.00021351923076923078,
+      "loss": 1.0465,
+      "step": 3502
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 0.0002134903846153846,
+      "loss": 0.8806,
+      "step": 3503
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 0.00021346153846153845,
+      "loss": 0.8867,
+      "step": 3504
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 0.00021343269230769227,
+      "loss": 0.9201,
+      "step": 3505
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 0.00021340384615384613,
+      "loss": 0.8864,
+      "step": 3506
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 0.000213375,
+      "loss": 0.8085,
+      "step": 3507
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 0.00021334615384615383,
+      "loss": 0.9917,
+      "step": 3508
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 0.00021331730769230767,
+      "loss": 1.0053,
+      "step": 3509
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 0.0002132884615384615,
+      "loss": 0.8779,
+      "step": 3510
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 0.00021325961538461538,
+      "loss": 0.7739,
+      "step": 3511
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 0.00021323076923076921,
+      "loss": 1.0578,
+      "step": 3512
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 0.00021320192307692305,
+      "loss": 0.9338,
+      "step": 3513
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 0.00021317307692307692,
+      "loss": 0.8658,
+      "step": 3514
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 0.00021314423076923073,
+      "loss": 0.926,
+      "step": 3515
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 0.0002131153846153846,
+      "loss": 1.0066,
+      "step": 3516
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 0.00021308653846153846,
+      "loss": 0.8308,
+      "step": 3517
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 0.0002130576923076923,
+      "loss": 0.8934,
+      "step": 3518
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 0.00021302884615384614,
+      "loss": 0.7972,
+      "step": 3519
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 0.8837,
+      "step": 3520
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 0.00021297115384615384,
+      "loss": 0.9736,
+      "step": 3521
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 0.00021294230769230768,
+      "loss": 0.8167,
+      "step": 3522
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 0.00021291346153846151,
+      "loss": 0.8675,
+      "step": 3523
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 0.00021288461538461538,
+      "loss": 1.0407,
+      "step": 3524
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 0.0002128557692307692,
+      "loss": 0.997,
+      "step": 3525
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 0.00021282692307692306,
+      "loss": 0.8551,
+      "step": 3526
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 0.0002127980769230769,
+      "loss": 0.8276,
+      "step": 3527
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 0.00021276923076923076,
+      "loss": 0.9584,
+      "step": 3528
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 0.0002127403846153846,
+      "loss": 0.9984,
+      "step": 3529
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 0.00021271153846153844,
+      "loss": 0.8328,
+      "step": 3530
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 0.0002126826923076923,
+      "loss": 0.8707,
+      "step": 3531
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 0.0002126538461538461,
+      "loss": 0.9787,
+      "step": 3532
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 0.00021262499999999998,
+      "loss": 0.9165,
+      "step": 3533
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 0.00021259615384615384,
+      "loss": 0.9208,
+      "step": 3534
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 0.00021256730769230765,
+      "loss": 0.9275,
+      "step": 3535
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 0.00021253846153846152,
+      "loss": 0.9072,
+      "step": 3536
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 0.00021250961538461536,
+      "loss": 0.9076,
+      "step": 3537
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 0.00021248076923076922,
+      "loss": 0.9955,
+      "step": 3538
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 0.00021245192307692306,
+      "loss": 0.8926,
+      "step": 3539
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 0.0002124230769230769,
+      "loss": 0.8759,
+      "step": 3540
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 0.00021239423076923076,
+      "loss": 0.909,
+      "step": 3541
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 0.00021236538461538458,
+      "loss": 0.9207,
+      "step": 3542
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 0.00021233653846153844,
+      "loss": 0.9935,
+      "step": 3543
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 0.0002123076923076923,
+      "loss": 0.9829,
+      "step": 3544
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 0.0002123076923076923,
+      "loss": 1.1515,
+      "step": 3545
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 0.00021227884615384612,
+      "loss": 0.905,
+      "step": 3546
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 0.00021224999999999998,
+      "loss": 0.966,
+      "step": 3547
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 0.00021222115384615382,
+      "loss": 0.9141,
+      "step": 3548
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 0.00021219230769230769,
+      "loss": 0.8009,
+      "step": 3549
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 0.00021216346153846152,
+      "loss": 0.8902,
+      "step": 3550
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 0.00021213461538461536,
+      "loss": 0.8514,
+      "step": 3551
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 0.00021210576923076923,
+      "loss": 0.7993,
+      "step": 3552
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 0.00021207692307692304,
+      "loss": 0.8225,
+      "step": 3553
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 0.0002120480769230769,
+      "loss": 0.9155,
+      "step": 3554
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 0.00021201923076923077,
+      "loss": 0.9838,
+      "step": 3555
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 0.00021199038461538458,
+      "loss": 0.888,
+      "step": 3556
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 0.00021196153846153844,
+      "loss": 1.0051,
+      "step": 3557
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 0.00021193269230769228,
+      "loss": 0.9226,
+      "step": 3558
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 0.00021190384615384615,
+      "loss": 0.9054,
+      "step": 3559
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 0.000211875,
+      "loss": 0.8373,
+      "step": 3560
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 0.00021184615384615382,
+      "loss": 0.8301,
+      "step": 3561
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 0.0002118173076923077,
+      "loss": 0.941,
+      "step": 3562
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 0.0002117884615384615,
+      "loss": 0.8393,
+      "step": 3563
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 0.00021175961538461537,
+      "loss": 0.9171,
+      "step": 3564
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 0.00021173076923076923,
+      "loss": 0.8859,
+      "step": 3565
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 0.00021170192307692304,
+      "loss": 0.717,
+      "step": 3566
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 0.0002116730769230769,
+      "loss": 0.9605,
+      "step": 3567
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 0.00021164423076923075,
+      "loss": 0.9046,
+      "step": 3568
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 0.0002116153846153846,
+      "loss": 1.0716,
+      "step": 3569
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 0.00021158653846153845,
+      "loss": 0.8966,
+      "step": 3570
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 0.0002115576923076923,
+      "loss": 0.9387,
+      "step": 3571
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 0.00021152884615384615,
+      "loss": 0.808,
+      "step": 3572
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 0.00021149999999999996,
+      "loss": 0.957,
+      "step": 3573
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 0.00021147115384615383,
+      "loss": 0.9781,
+      "step": 3574
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 0.0002114423076923077,
+      "loss": 0.8063,
+      "step": 3575
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 0.0002114134615384615,
+      "loss": 0.8828,
+      "step": 3576
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 0.00021138461538461537,
+      "loss": 0.9865,
+      "step": 3577
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 0.0002113557692307692,
+      "loss": 0.9625,
+      "step": 3578
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 0.00021132692307692305,
+      "loss": 0.8634,
+      "step": 3579
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 0.0002112980769230769,
+      "loss": 0.9544,
+      "step": 3580
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 0.00021126923076923075,
+      "loss": 0.901,
+      "step": 3581
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 0.00021124038461538462,
+      "loss": 0.9536,
+      "step": 3582
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 0.00021121153846153843,
+      "loss": 0.9314,
+      "step": 3583
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 0.0002111826923076923,
+      "loss": 0.9844,
+      "step": 3584
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 0.00021115384615384616,
+      "loss": 0.9951,
+      "step": 3585
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 0.00021112499999999997,
+      "loss": 0.8562,
+      "step": 3586
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 0.00021109615384615383,
+      "loss": 1.0167,
+      "step": 3587
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 0.00021106730769230767,
+      "loss": 0.9363,
+      "step": 3588
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 0.0002110384615384615,
+      "loss": 0.9374,
+      "step": 3589
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 0.00021100961538461538,
+      "loss": 0.8233,
+      "step": 3590
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 0.0002109807692307692,
+      "loss": 0.8536,
+      "step": 3591
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 0.00021095192307692308,
+      "loss": 0.9139,
+      "step": 3592
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 0.0002109230769230769,
+      "loss": 0.902,
+      "step": 3593
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 0.00021089423076923075,
+      "loss": 1.0207,
+      "step": 3594
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 0.0002108653846153846,
+      "loss": 0.8944,
+      "step": 3595
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 0.00021083653846153843,
+      "loss": 0.9962,
+      "step": 3596
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 0.0002108076923076923,
+      "loss": 0.8594,
+      "step": 3597
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 0.00021077884615384613,
+      "loss": 0.9574,
+      "step": 3598
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 0.00021074999999999997,
+      "loss": 0.8808,
+      "step": 3599
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 0.0002107211538461538,
+      "loss": 0.7837,
+      "step": 3600
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 0.00021069230769230768,
+      "loss": 0.8536,
+      "step": 3601
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 0.00021066346153846154,
+      "loss": 0.8034,
+      "step": 3602
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 0.00021063461538461535,
+      "loss": 0.977,
+      "step": 3603
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 0.00021060576923076922,
+      "loss": 0.8372,
+      "step": 3604
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 0.00021057692307692306,
+      "loss": 0.8674,
+      "step": 3605
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 0.0002105480769230769,
+      "loss": 0.8015,
+      "step": 3606
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 0.00021051923076923076,
+      "loss": 0.8669,
+      "step": 3607
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 0.0002104903846153846,
+      "loss": 0.9538,
+      "step": 3608
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 0.00021046153846153844,
+      "loss": 0.9435,
+      "step": 3609
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 0.00021043269230769227,
+      "loss": 0.9002,
+      "step": 3610
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 0.00021040384615384614,
+      "loss": 0.8315,
+      "step": 3611
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 0.000210375,
+      "loss": 0.8223,
+      "step": 3612
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 0.00021034615384615382,
+      "loss": 0.9016,
+      "step": 3613
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 0.00021031730769230768,
+      "loss": 0.9653,
+      "step": 3614
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 0.00021028846153846152,
+      "loss": 1.0171,
+      "step": 3615
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 0.00021025961538461536,
+      "loss": 0.8732,
+      "step": 3616
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 0.00021023076923076922,
+      "loss": 0.9245,
+      "step": 3617
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 0.00021020192307692306,
+      "loss": 0.8802,
+      "step": 3618
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 0.0002101730769230769,
+      "loss": 0.9193,
+      "step": 3619
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 0.00021014423076923074,
+      "loss": 1.0691,
+      "step": 3620
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 0.0002101153846153846,
+      "loss": 0.8515,
+      "step": 3621
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 0.00021008653846153847,
+      "loss": 0.9934,
+      "step": 3622
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 0.00021005769230769228,
+      "loss": 0.8991,
+      "step": 3623
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 0.00021002884615384614,
+      "loss": 0.8781,
+      "step": 3624
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 1.0386,
+      "step": 3625
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 0.00020997115384615382,
+      "loss": 0.8268,
+      "step": 3626
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 0.00020994230769230769,
+      "loss": 0.798,
+      "step": 3627
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 0.00020991346153846152,
+      "loss": 0.9274,
+      "step": 3628
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 0.00020988461538461536,
+      "loss": 0.803,
+      "step": 3629
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 0.0002098557692307692,
+      "loss": 0.8319,
+      "step": 3630
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 0.00020982692307692306,
+      "loss": 0.8246,
+      "step": 3631
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 0.00020979807692307693,
+      "loss": 0.9137,
+      "step": 3632
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 0.00020976923076923074,
+      "loss": 0.8902,
+      "step": 3633
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 0.0002097403846153846,
+      "loss": 0.9682,
+      "step": 3634
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 0.00020971153846153842,
+      "loss": 0.9725,
+      "step": 3635
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 0.00020968269230769228,
+      "loss": 0.8608,
+      "step": 3636
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 0.00020965384615384615,
+      "loss": 0.8457,
+      "step": 3637
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 0.00020962499999999999,
+      "loss": 0.9599,
+      "step": 3638
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 0.00020959615384615382,
+      "loss": 0.9287,
+      "step": 3639
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 0.00020956730769230766,
+      "loss": 0.9179,
+      "step": 3640
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 0.00020953846153846153,
+      "loss": 0.8441,
+      "step": 3641
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 0.0002095096153846154,
+      "loss": 0.8851,
+      "step": 3642
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 0.0002094807692307692,
+      "loss": 0.924,
+      "step": 3643
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 0.00020945192307692307,
+      "loss": 0.8285,
+      "step": 3644
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 0.00020942307692307688,
+      "loss": 1.0251,
+      "step": 3645
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 0.00020939423076923075,
+      "loss": 0.9377,
+      "step": 3646
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 0.0002093653846153846,
+      "loss": 0.9125,
+      "step": 3647
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 0.00020933653846153845,
+      "loss": 0.9303,
+      "step": 3648
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 0.0002093076923076923,
+      "loss": 0.9188,
+      "step": 3649
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 0.00020927884615384613,
+      "loss": 0.8474,
+      "step": 3650
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 0.00020925,
+      "loss": 0.7941,
+      "step": 3651
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 0.00020922115384615386,
+      "loss": 1.1177,
+      "step": 3652
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 0.00020919230769230767,
+      "loss": 0.8718,
+      "step": 3653
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 0.00020916346153846153,
+      "loss": 0.8553,
+      "step": 3654
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 0.00020913461538461534,
+      "loss": 0.757,
+      "step": 3655
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 0.0002091057692307692,
+      "loss": 0.7948,
+      "step": 3656
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 0.00020907692307692307,
+      "loss": 0.8606,
+      "step": 3657
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 0.0002090480769230769,
+      "loss": 0.8462,
+      "step": 3658
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 0.00020901923076923075,
+      "loss": 0.7962,
+      "step": 3659
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 0.0002089903846153846,
+      "loss": 0.7536,
+      "step": 3660
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 0.00020896153846153845,
+      "loss": 0.9706,
+      "step": 3661
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 0.00020893269230769226,
+      "loss": 0.8758,
+      "step": 3662
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 0.00020890384615384613,
+      "loss": 0.9335,
+      "step": 3663
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 0.000208875,
+      "loss": 0.9508,
+      "step": 3664
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 0.0002088461538461538,
+      "loss": 0.925,
+      "step": 3665
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 0.00020881730769230767,
+      "loss": 1.073,
+      "step": 3666
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 0.0002087884615384615,
+      "loss": 0.9105,
+      "step": 3667
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 0.00020875961538461537,
+      "loss": 0.9619,
+      "step": 3668
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 0.0002087307692307692,
+      "loss": 0.9176,
+      "step": 3669
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 0.00020870192307692305,
+      "loss": 0.9483,
+      "step": 3670
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 0.00020867307692307692,
+      "loss": 0.8511,
+      "step": 3671
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 0.00020864423076923073,
+      "loss": 0.7918,
+      "step": 3672
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 0.0002086153846153846,
+      "loss": 0.8099,
+      "step": 3673
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 0.00020858653846153846,
+      "loss": 0.8658,
+      "step": 3674
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 0.00020855769230769227,
+      "loss": 0.9393,
+      "step": 3675
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 0.00020852884615384613,
+      "loss": 0.968,
+      "step": 3676
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 0.00020849999999999997,
+      "loss": 0.8577,
+      "step": 3677
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 0.00020847115384615384,
+      "loss": 1.0062,
+      "step": 3678
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 0.00020844230769230768,
+      "loss": 0.9984,
+      "step": 3679
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 0.00020841346153846151,
+      "loss": 0.9485,
+      "step": 3680
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 0.00020838461538461538,
+      "loss": 0.8857,
+      "step": 3681
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 0.0002083557692307692,
+      "loss": 0.8141,
+      "step": 3682
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 0.00020832692307692306,
+      "loss": 1.0013,
+      "step": 3683
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 0.00020829807692307692,
+      "loss": 0.7629,
+      "step": 3684
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 0.00020826923076923073,
+      "loss": 0.8382,
+      "step": 3685
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 0.0002082403846153846,
+      "loss": 0.8224,
+      "step": 3686
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 0.00020821153846153844,
+      "loss": 0.7083,
+      "step": 3687
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 0.0002081826923076923,
+      "loss": 0.9673,
+      "step": 3688
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 0.00020815384615384614,
+      "loss": 0.8925,
+      "step": 3689
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 0.00020812499999999998,
+      "loss": 0.9967,
+      "step": 3690
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 0.00020809615384615384,
+      "loss": 1.0723,
+      "step": 3691
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 0.00020806730769230765,
+      "loss": 0.9149,
+      "step": 3692
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 0.00020803846153846152,
+      "loss": 0.7002,
+      "step": 3693
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 0.00020800961538461538,
+      "loss": 0.9753,
+      "step": 3694
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 0.0002079807692307692,
+      "loss": 0.8155,
+      "step": 3695
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 0.00020795192307692306,
+      "loss": 0.8443,
+      "step": 3696
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 0.0002079230769230769,
+      "loss": 0.9102,
+      "step": 3697
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 0.00020789423076923076,
+      "loss": 0.9355,
+      "step": 3698
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 0.0002078653846153846,
+      "loss": 0.7779,
+      "step": 3699
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 0.00020783653846153844,
+      "loss": 0.6781,
+      "step": 3700
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 0.0002078076923076923,
+      "loss": 0.8665,
+      "step": 3701
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 0.00020777884615384612,
+      "loss": 0.9624,
+      "step": 3702
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 0.00020774999999999998,
+      "loss": 0.9268,
+      "step": 3703
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 0.00020772115384615385,
+      "loss": 0.7279,
+      "step": 3704
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 0.00020769230769230766,
+      "loss": 0.8082,
+      "step": 3705
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 0.00020766346153846152,
+      "loss": 0.9563,
+      "step": 3706
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 0.00020763461538461536,
+      "loss": 0.9164,
+      "step": 3707
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 0.00020760576923076923,
+      "loss": 0.7751,
+      "step": 3708
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 0.00020757692307692306,
+      "loss": 0.8952,
+      "step": 3709
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 0.0002075480769230769,
+      "loss": 0.8479,
+      "step": 3710
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 0.00020751923076923077,
+      "loss": 0.8643,
+      "step": 3711
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 0.00020749038461538458,
+      "loss": 0.8954,
+      "step": 3712
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 0.00020746153846153844,
+      "loss": 0.8703,
+      "step": 3713
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 0.0002074326923076923,
+      "loss": 0.8526,
+      "step": 3714
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 0.00020740384615384612,
+      "loss": 0.8213,
+      "step": 3715
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 0.00020737499999999999,
+      "loss": 0.8857,
+      "step": 3716
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 0.00020734615384615382,
+      "loss": 0.8729,
+      "step": 3717
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 0.0002073173076923077,
+      "loss": 0.9186,
+      "step": 3718
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 0.00020728846153846153,
+      "loss": 0.8857,
+      "step": 3719
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 0.00020725961538461537,
+      "loss": 0.8756,
+      "step": 3720
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 0.00020723076923076923,
+      "loss": 0.8347,
+      "step": 3721
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 0.00020720192307692304,
+      "loss": 0.8928,
+      "step": 3722
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 0.0002071730769230769,
+      "loss": 0.88,
+      "step": 3723
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 0.00020714423076923075,
+      "loss": 0.9347,
+      "step": 3724
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 0.00020711538461538458,
+      "loss": 1.0125,
+      "step": 3725
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 0.00020708653846153845,
+      "loss": 0.9315,
+      "step": 3726
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 0.0002070576923076923,
+      "loss": 0.9137,
+      "step": 3727
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 0.00020702884615384615,
+      "loss": 0.8687,
+      "step": 3728
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 0.9614,
+      "step": 3729
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 0.00020697115384615383,
+      "loss": 0.909,
+      "step": 3730
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 0.0002069423076923077,
+      "loss": 0.9469,
+      "step": 3731
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 0.0002069134615384615,
+      "loss": 0.9233,
+      "step": 3732
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 0.00020688461538461537,
+      "loss": 1.0762,
+      "step": 3733
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 0.0002068557692307692,
+      "loss": 0.744,
+      "step": 3734
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 0.00020682692307692305,
+      "loss": 0.7674,
+      "step": 3735
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 0.0002067980769230769,
+      "loss": 0.8664,
+      "step": 3736
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 0.00020676923076923075,
+      "loss": 0.8523,
+      "step": 3737
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 0.00020674038461538461,
+      "loss": 0.9253,
+      "step": 3738
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 0.00020671153846153843,
+      "loss": 0.8354,
+      "step": 3739
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 0.0002066826923076923,
+      "loss": 0.7766,
+      "step": 3740
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 0.00020665384615384616,
+      "loss": 0.8706,
+      "step": 3741
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 0.00020662499999999997,
+      "loss": 0.9332,
+      "step": 3742
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 0.00020659615384615383,
+      "loss": 0.899,
+      "step": 3743
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 0.00020656730769230767,
+      "loss": 0.9273,
+      "step": 3744
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 0.0002065384615384615,
+      "loss": 0.7803,
+      "step": 3745
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 0.00020650961538461537,
+      "loss": 0.7552,
+      "step": 3746
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 0.0002064807692307692,
+      "loss": 0.9131,
+      "step": 3747
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 0.00020645192307692308,
+      "loss": 0.9366,
+      "step": 3748
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 0.0002064230769230769,
+      "loss": 0.8925,
+      "step": 3749
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 0.00020639423076923075,
+      "loss": 0.8608,
+      "step": 3750
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 0.00020636538461538462,
+      "loss": 0.895,
+      "step": 3751
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 0.00020633653846153843,
+      "loss": 0.8323,
+      "step": 3752
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 0.0002063076923076923,
+      "loss": 0.8715,
+      "step": 3753
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 0.00020627884615384613,
+      "loss": 0.8896,
+      "step": 3754
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 0.00020624999999999997,
+      "loss": 0.9205,
+      "step": 3755
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 0.00020622115384615384,
+      "loss": 0.8299,
+      "step": 3756
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 0.00020619230769230768,
+      "loss": 0.879,
+      "step": 3757
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 0.00020616346153846154,
+      "loss": 0.9888,
+      "step": 3758
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 0.00020613461538461535,
+      "loss": 0.8998,
+      "step": 3759
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 0.00020610576923076922,
+      "loss": 0.9125,
+      "step": 3760
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 0.00020607692307692308,
+      "loss": 0.97,
+      "step": 3761
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 0.0002060480769230769,
+      "loss": 0.9129,
+      "step": 3762
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 0.00020601923076923076,
+      "loss": 0.9332,
+      "step": 3763
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 0.0002059903846153846,
+      "loss": 0.8624,
+      "step": 3764
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 0.00020596153846153843,
+      "loss": 0.9804,
+      "step": 3765
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 0.0002059326923076923,
+      "loss": 0.8036,
+      "step": 3766
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 0.00020590384615384614,
+      "loss": 0.8908,
+      "step": 3767
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 0.000205875,
+      "loss": 0.9524,
+      "step": 3768
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 0.00020584615384615381,
+      "loss": 0.8301,
+      "step": 3769
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 0.00020581730769230768,
+      "loss": 0.8086,
+      "step": 3770
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 0.00020578846153846154,
+      "loss": 0.8093,
+      "step": 3771
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 0.00020575961538461536,
+      "loss": 0.8275,
+      "step": 3772
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 0.00020573076923076922,
+      "loss": 0.9284,
+      "step": 3773
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 0.00020570192307692306,
+      "loss": 0.8944,
+      "step": 3774
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 0.0002056730769230769,
+      "loss": 0.8524,
+      "step": 3775
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 0.00020564423076923076,
+      "loss": 0.8667,
+      "step": 3776
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 0.0002056153846153846,
+      "loss": 0.9619,
+      "step": 3777
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 0.00020558653846153844,
+      "loss": 0.8404,
+      "step": 3778
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 0.00020555769230769228,
+      "loss": 0.9521,
+      "step": 3779
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 0.00020552884615384614,
+      "loss": 0.9037,
+      "step": 3780
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 0.0002055,
+      "loss": 0.7594,
+      "step": 3781
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 0.00020547115384615382,
+      "loss": 0.8163,
+      "step": 3782
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 0.00020544230769230768,
+      "loss": 0.8054,
+      "step": 3783
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 0.00020541346153846152,
+      "loss": 0.8984,
+      "step": 3784
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 0.00020538461538461536,
+      "loss": 0.8004,
+      "step": 3785
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 0.00020535576923076923,
+      "loss": 0.9919,
+      "step": 3786
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 0.00020532692307692306,
+      "loss": 0.8079,
+      "step": 3787
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 0.0002052980769230769,
+      "loss": 0.9255,
+      "step": 3788
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 0.00020526923076923074,
+      "loss": 0.968,
+      "step": 3789
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 0.0002052403846153846,
+      "loss": 0.8787,
+      "step": 3790
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 0.00020521153846153842,
+      "loss": 0.8358,
+      "step": 3791
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 0.00020518269230769228,
+      "loss": 0.9388,
+      "step": 3792
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 0.00020515384615384615,
+      "loss": 0.9369,
+      "step": 3793
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 0.00020512499999999999,
+      "loss": 0.9166,
+      "step": 3794
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 0.00020509615384615382,
+      "loss": 0.8348,
+      "step": 3795
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 0.00020506730769230766,
+      "loss": 0.8068,
+      "step": 3796
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 0.00020503846153846153,
+      "loss": 0.9529,
+      "step": 3797
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 0.00020500961538461536,
+      "loss": 0.788,
+      "step": 3798
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 0.0002049807692307692,
+      "loss": 0.8546,
+      "step": 3799
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 0.00020495192307692307,
+      "loss": 1.1,
+      "step": 3800
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 0.00020492307692307688,
+      "loss": 0.8881,
+      "step": 3801
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 0.00020489423076923074,
+      "loss": 0.7621,
+      "step": 3802
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 0.0002048653846153846,
+      "loss": 0.839,
+      "step": 3803
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 0.00020483653846153845,
+      "loss": 0.851,
+      "step": 3804
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 0.00020480769230769229,
+      "loss": 0.9075,
+      "step": 3805
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 0.00020477884615384612,
+      "loss": 0.9067,
+      "step": 3806
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 0.00020475,
+      "loss": 0.8643,
+      "step": 3807
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 0.00020472115384615383,
+      "loss": 0.9278,
+      "step": 3808
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 0.00020469230769230767,
+      "loss": 0.9111,
+      "step": 3809
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 0.00020466346153846153,
+      "loss": 0.9851,
+      "step": 3810
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 0.00020463461538461534,
+      "loss": 0.8948,
+      "step": 3811
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 0.0002046057692307692,
+      "loss": 0.8843,
+      "step": 3812
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 0.00020457692307692307,
+      "loss": 0.7274,
+      "step": 3813
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 0.0002045480769230769,
+      "loss": 1.0277,
+      "step": 3814
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 0.00020451923076923075,
+      "loss": 0.7644,
+      "step": 3815
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 0.0002044903846153846,
+      "loss": 0.8985,
+      "step": 3816
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 0.00020446153846153845,
+      "loss": 0.7778,
+      "step": 3817
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 0.0002044326923076923,
+      "loss": 0.8887,
+      "step": 3818
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 0.00020440384615384613,
+      "loss": 0.7497,
+      "step": 3819
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 0.000204375,
+      "loss": 0.8571,
+      "step": 3820
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 0.0002043461538461538,
+      "loss": 0.9846,
+      "step": 3821
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 0.00020431730769230767,
+      "loss": 0.9719,
+      "step": 3822
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 0.00020428846153846154,
+      "loss": 0.86,
+      "step": 3823
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 0.00020425961538461537,
+      "loss": 0.9499,
+      "step": 3824
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 0.0002042307692307692,
+      "loss": 0.8884,
+      "step": 3825
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 0.00020420192307692305,
+      "loss": 0.8976,
+      "step": 3826
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 0.00020417307692307692,
+      "loss": 0.7967,
+      "step": 3827
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 0.00020414423076923075,
+      "loss": 0.8528,
+      "step": 3828
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 0.0002041153846153846,
+      "loss": 0.8637,
+      "step": 3829
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 0.00020408653846153846,
+      "loss": 0.8081,
+      "step": 3830
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 0.00020405769230769227,
+      "loss": 0.7699,
+      "step": 3831
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 0.00020402884615384613,
+      "loss": 0.9848,
+      "step": 3832
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 0.000204,
+      "loss": 0.8322,
+      "step": 3833
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 0.00020397115384615384,
+      "loss": 0.9297,
+      "step": 3834
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 0.00020394230769230767,
+      "loss": 0.8663,
+      "step": 3835
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 0.0002039134615384615,
+      "loss": 0.8263,
+      "step": 3836
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 0.00020388461538461538,
+      "loss": 0.9411,
+      "step": 3837
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 0.00020385576923076922,
+      "loss": 0.8546,
+      "step": 3838
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 0.00020382692307692305,
+      "loss": 0.9203,
+      "step": 3839
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 0.00020379807692307692,
+      "loss": 0.8662,
+      "step": 3840
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 0.00020376923076923073,
+      "loss": 0.8336,
+      "step": 3841
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 0.0002037403846153846,
+      "loss": 0.7953,
+      "step": 3842
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 0.00020371153846153846,
+      "loss": 0.8588,
+      "step": 3843
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 0.00020368269230769227,
+      "loss": 0.9358,
+      "step": 3844
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 0.00020365384615384614,
+      "loss": 0.8269,
+      "step": 3845
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 0.00020362499999999998,
+      "loss": 0.8368,
+      "step": 3846
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 0.00020362499999999998,
+      "loss": 0.8728,
+      "step": 3847
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 0.00020359615384615384,
+      "loss": 0.8955,
+      "step": 3848
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 0.00020356730769230768,
+      "loss": 0.9191,
+      "step": 3849
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 0.00020353846153846152,
+      "loss": 0.7879,
+      "step": 3850
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 0.00020350961538461538,
+      "loss": 0.9609,
+      "step": 3851
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 0.0002034807692307692,
+      "loss": 1.0855,
+      "step": 3852
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 0.00020345192307692306,
+      "loss": 0.7924,
+      "step": 3853
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 0.0002034230769230769,
+      "loss": 0.9272,
+      "step": 3854
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 0.00020339423076923074,
+      "loss": 0.8738,
+      "step": 3855
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 0.0002033653846153846,
+      "loss": 0.7501,
+      "step": 3856
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 0.00020333653846153844,
+      "loss": 0.9455,
+      "step": 3857
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 0.0002033076923076923,
+      "loss": 0.7964,
+      "step": 3858
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 0.00020327884615384611,
+      "loss": 0.9254,
+      "step": 3859
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 0.00020324999999999998,
+      "loss": 0.8247,
+      "step": 3860
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 0.00020322115384615385,
+      "loss": 0.8646,
+      "step": 3861
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 0.00020319230769230766,
+      "loss": 0.9851,
+      "step": 3862
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 0.00020316346153846152,
+      "loss": 0.877,
+      "step": 3863
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 0.00020313461538461536,
+      "loss": 0.9813,
+      "step": 3864
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 0.0002031057692307692,
+      "loss": 0.7621,
+      "step": 3865
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 0.00020307692307692306,
+      "loss": 0.9144,
+      "step": 3866
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 0.0002030480769230769,
+      "loss": 0.8007,
+      "step": 3867
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 0.00020301923076923077,
+      "loss": 0.8459,
+      "step": 3868
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 0.00020299038461538458,
+      "loss": 0.9571,
+      "step": 3869
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 0.00020296153846153844,
+      "loss": 0.8764,
+      "step": 3870
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 0.0002029326923076923,
+      "loss": 0.8914,
+      "step": 3871
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 0.00020290384615384612,
+      "loss": 1.0435,
+      "step": 3872
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 0.00020287499999999998,
+      "loss": 0.928,
+      "step": 3873
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 0.00020284615384615382,
+      "loss": 0.8818,
+      "step": 3874
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 0.00020281730769230766,
+      "loss": 0.909,
+      "step": 3875
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 0.00020278846153846153,
+      "loss": 0.9493,
+      "step": 3876
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 0.00020275961538461536,
+      "loss": 0.8821,
+      "step": 3877
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 0.00020273076923076923,
+      "loss": 0.8704,
+      "step": 3878
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 0.00020270192307692304,
+      "loss": 0.9601,
+      "step": 3879
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 0.0002026730769230769,
+      "loss": 0.7929,
+      "step": 3880
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 0.00020264423076923077,
+      "loss": 0.8394,
+      "step": 3881
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 0.00020261538461538458,
+      "loss": 0.8369,
+      "step": 3882
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 0.00020258653846153845,
+      "loss": 1.1072,
+      "step": 3883
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 0.00020255769230769229,
+      "loss": 0.8635,
+      "step": 3884
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 0.00020252884615384612,
+      "loss": 0.8926,
+      "step": 3885
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 0.0002025,
+      "loss": 0.8874,
+      "step": 3886
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 0.00020247115384615383,
+      "loss": 0.8868,
+      "step": 3887
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 0.0002024423076923077,
+      "loss": 0.9238,
+      "step": 3888
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 0.0002024134615384615,
+      "loss": 0.9817,
+      "step": 3889
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 0.00020238461538461537,
+      "loss": 0.8621,
+      "step": 3890
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 0.00020235576923076923,
+      "loss": 0.9238,
+      "step": 3891
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 0.00020232692307692305,
+      "loss": 0.8456,
+      "step": 3892
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 0.0002022980769230769,
+      "loss": 0.9176,
+      "step": 3893
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 0.00020226923076923075,
+      "loss": 0.9269,
+      "step": 3894
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 0.0002022403846153846,
+      "loss": 0.7718,
+      "step": 3895
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 0.00020221153846153845,
+      "loss": 0.8941,
+      "step": 3896
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 0.0002021826923076923,
+      "loss": 0.8151,
+      "step": 3897
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 0.00020215384615384616,
+      "loss": 0.8509,
+      "step": 3898
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 0.00020212499999999997,
+      "loss": 0.869,
+      "step": 3899
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 0.00020209615384615383,
+      "loss": 0.8123,
+      "step": 3900
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 0.0002020673076923077,
+      "loss": 0.9575,
+      "step": 3901
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 0.0002020384615384615,
+      "loss": 0.9782,
+      "step": 3902
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 0.00020200961538461537,
+      "loss": 0.7684,
+      "step": 3903
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 0.0002019807692307692,
+      "loss": 0.8965,
+      "step": 3904
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 0.00020195192307692305,
+      "loss": 0.873,
+      "step": 3905
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 0.00020192307692307691,
+      "loss": 0.9057,
+      "step": 3906
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 0.00020189423076923075,
+      "loss": 0.8915,
+      "step": 3907
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 0.00020186538461538462,
+      "loss": 0.9352,
+      "step": 3908
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 0.00020183653846153843,
+      "loss": 0.9836,
+      "step": 3909
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 0.0002018076923076923,
+      "loss": 0.9851,
+      "step": 3910
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 0.00020177884615384616,
+      "loss": 0.8727,
+      "step": 3911
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 0.00020174999999999997,
+      "loss": 0.9213,
+      "step": 3912
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 0.00020172115384615384,
+      "loss": 0.8113,
+      "step": 3913
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 0.00020169230769230767,
+      "loss": 0.8484,
+      "step": 3914
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 0.0002016634615384615,
+      "loss": 0.6815,
+      "step": 3915
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 0.00020163461538461538,
+      "loss": 0.8446,
+      "step": 3916
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 0.00020160576923076922,
+      "loss": 0.848,
+      "step": 3917
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 0.00020157692307692308,
+      "loss": 0.7723,
+      "step": 3918
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 0.0002015480769230769,
+      "loss": 0.9377,
+      "step": 3919
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 0.00020151923076923076,
+      "loss": 0.8461,
+      "step": 3920
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 0.00020149038461538457,
+      "loss": 0.8204,
+      "step": 3921
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 0.00020146153846153843,
+      "loss": 0.8735,
+      "step": 3922
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 0.0002014326923076923,
+      "loss": 0.8167,
+      "step": 3923
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 0.00020140384615384614,
+      "loss": 0.8394,
+      "step": 3924
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 0.00020137499999999998,
+      "loss": 0.9003,
+      "step": 3925
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 0.0002013461538461538,
+      "loss": 0.8127,
+      "step": 3926
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 0.00020131730769230768,
+      "loss": 0.937,
+      "step": 3927
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 0.00020128846153846154,
+      "loss": 0.8415,
+      "step": 3928
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 0.00020125961538461535,
+      "loss": 0.9667,
+      "step": 3929
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 0.00020123076923076922,
+      "loss": 0.8672,
+      "step": 3930
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 0.00020120192307692303,
+      "loss": 0.7742,
+      "step": 3931
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 0.0002011730769230769,
+      "loss": 0.8505,
+      "step": 3932
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 0.00020114423076923076,
+      "loss": 0.7654,
+      "step": 3933
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 0.0002011153846153846,
+      "loss": 0.9821,
+      "step": 3934
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 0.00020108653846153844,
+      "loss": 0.9095,
+      "step": 3935
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 0.00020105769230769228,
+      "loss": 0.8374,
+      "step": 3936
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 0.00020102884615384614,
+      "loss": 0.8893,
+      "step": 3937
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 0.000201,
+      "loss": 0.8729,
+      "step": 3938
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 0.00020097115384615382,
+      "loss": 0.9369,
+      "step": 3939
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 0.00020094230769230768,
+      "loss": 0.7737,
+      "step": 3940
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 0.0002009134615384615,
+      "loss": 0.9172,
+      "step": 3941
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 0.00020088461538461536,
+      "loss": 0.8956,
+      "step": 3942
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 0.00020085576923076922,
+      "loss": 0.9637,
+      "step": 3943
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 0.00020082692307692306,
+      "loss": 0.8173,
+      "step": 3944
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 0.0002007980769230769,
+      "loss": 0.7861,
+      "step": 3945
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 0.00020076923076923074,
+      "loss": 0.8699,
+      "step": 3946
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 0.0002007403846153846,
+      "loss": 0.8601,
+      "step": 3947
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 0.00020071153846153847,
+      "loss": 0.9298,
+      "step": 3948
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 0.00020068269230769228,
+      "loss": 0.8079,
+      "step": 3949
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 0.00020065384615384615,
+      "loss": 0.8377,
+      "step": 3950
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 0.00020062499999999996,
+      "loss": 0.8312,
+      "step": 3951
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 0.00020059615384615382,
+      "loss": 0.8214,
+      "step": 3952
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 0.0002005673076923077,
+      "loss": 1.0897,
+      "step": 3953
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 0.00020053846153846153,
+      "loss": 0.8114,
+      "step": 3954
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 0.00020050961538461536,
+      "loss": 1.0263,
+      "step": 3955
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 0.0002004807692307692,
+      "loss": 0.9331,
+      "step": 3956
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 0.00020045192307692307,
+      "loss": 0.9068,
+      "step": 3957
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 0.00020042307692307693,
+      "loss": 0.7226,
+      "step": 3958
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 0.00020039423076923074,
+      "loss": 0.8876,
+      "step": 3959
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 0.0002003653846153846,
+      "loss": 0.8183,
+      "step": 3960
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 0.00020033653846153842,
+      "loss": 0.8963,
+      "step": 3961
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 0.00020030769230769229,
+      "loss": 0.928,
+      "step": 3962
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 0.00020027884615384615,
+      "loss": 0.9497,
+      "step": 3963
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 0.00020025,
+      "loss": 0.8051,
+      "step": 3964
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 0.00020022115384615383,
+      "loss": 0.7953,
+      "step": 3965
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 0.00020019230769230766,
+      "loss": 0.8853,
+      "step": 3966
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 0.00020016346153846153,
+      "loss": 0.7695,
+      "step": 3967
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 0.0002001346153846154,
+      "loss": 0.9789,
+      "step": 3968
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 0.0002001057692307692,
+      "loss": 0.804,
+      "step": 3969
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 0.00020007692307692307,
+      "loss": 0.7808,
+      "step": 3970
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 0.00020004807692307688,
+      "loss": 0.9268,
+      "step": 3971
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 0.00020001923076923075,
+      "loss": 0.9455,
+      "step": 3972
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 0.0001999903846153846,
+      "loss": 0.7445,
+      "step": 3973
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 0.00019996153846153845,
+      "loss": 0.9473,
+      "step": 3974
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 0.0001999326923076923,
+      "loss": 0.823,
+      "step": 3975
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 0.00019990384615384613,
+      "loss": 0.8547,
+      "step": 3976
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 0.000199875,
+      "loss": 0.848,
+      "step": 3977
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 0.00019984615384615386,
+      "loss": 0.8269,
+      "step": 3978
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 0.00019981730769230767,
+      "loss": 0.9398,
+      "step": 3979
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 0.00019978846153846153,
+      "loss": 0.8175,
+      "step": 3980
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 0.00019975961538461535,
+      "loss": 0.8688,
+      "step": 3981
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 0.0001997307692307692,
+      "loss": 0.8406,
+      "step": 3982
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 0.00019970192307692308,
+      "loss": 0.8334,
+      "step": 3983
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 0.00019967307692307691,
+      "loss": 0.9133,
+      "step": 3984
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 0.00019964423076923075,
+      "loss": 0.8744,
+      "step": 3985
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 0.0001996153846153846,
+      "loss": 0.8869,
+      "step": 3986
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 0.00019958653846153846,
+      "loss": 0.8982,
+      "step": 3987
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 0.00019955769230769227,
+      "loss": 0.8739,
+      "step": 3988
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 0.00019952884615384613,
+      "loss": 0.7555,
+      "step": 3989
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 0.0001995,
+      "loss": 0.905,
+      "step": 3990
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 0.0001994711538461538,
+      "loss": 0.7935,
+      "step": 3991
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 0.00019944230769230767,
+      "loss": 1.0293,
+      "step": 3992
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 0.0001994134615384615,
+      "loss": 0.9678,
+      "step": 3993
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 0.00019938461538461538,
+      "loss": 0.8701,
+      "step": 3994
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 0.00019935576923076922,
+      "loss": 0.7955,
+      "step": 3995
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 0.00019932692307692305,
+      "loss": 0.9627,
+      "step": 3996
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 0.00019929807692307692,
+      "loss": 0.669,
+      "step": 3997
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 0.00019926923076923073,
+      "loss": 0.9616,
+      "step": 3998
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 0.0001992403846153846,
+      "loss": 0.8615,
+      "step": 3999
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 0.00019921153846153846,
+      "loss": 0.8193,
+      "step": 4000
+    },
+    {
+      "epoch": 36.7,
+      "eval_cer": 0.10359975411925217,
+      "eval_loss": 0.3694167733192444,
+      "eval_runtime": 13.8555,
+      "eval_samples_per_second": 118.869,
+      "eval_steps_per_second": 1.877,
+      "eval_wer": 0.4223266264937187,
+      "step": 4000
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 0.00019918269230769227,
+      "loss": 0.89,
+      "step": 4001
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 0.00019915384615384614,
+      "loss": 0.911,
+      "step": 4002
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 0.00019912499999999997,
+      "loss": 0.8788,
+      "step": 4003
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 0.00019909615384615384,
+      "loss": 0.7958,
+      "step": 4004
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 0.00019906730769230768,
+      "loss": 0.8986,
+      "step": 4005
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 0.00019903846153846152,
+      "loss": 0.9424,
+      "step": 4006
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 0.00019900961538461538,
+      "loss": 0.8277,
+      "step": 4007
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 0.0001989807692307692,
+      "loss": 0.8133,
+      "step": 4008
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 0.00019895192307692306,
+      "loss": 0.7445,
+      "step": 4009
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 0.00019892307692307692,
+      "loss": 0.9926,
+      "step": 4010
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 0.00019889423076923073,
+      "loss": 0.9315,
+      "step": 4011
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 0.0001988653846153846,
+      "loss": 0.8855,
+      "step": 4012
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 0.00019883653846153844,
+      "loss": 0.954,
+      "step": 4013
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 0.0001988076923076923,
+      "loss": 0.827,
+      "step": 4014
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 0.00019877884615384614,
+      "loss": 0.8971,
+      "step": 4015
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 0.00019874999999999998,
+      "loss": 1.0008,
+      "step": 4016
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 0.00019872115384615384,
+      "loss": 0.8573,
+      "step": 4017
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 0.00019869230769230766,
+      "loss": 0.9337,
+      "step": 4018
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 0.00019866346153846152,
+      "loss": 0.8482,
+      "step": 4019
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 0.00019863461538461539,
+      "loss": 0.7079,
+      "step": 4020
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 0.0001986057692307692,
+      "loss": 0.8583,
+      "step": 4021
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 0.00019857692307692306,
+      "loss": 0.8073,
+      "step": 4022
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 0.0001985480769230769,
+      "loss": 0.8738,
+      "step": 4023
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 0.00019851923076923077,
+      "loss": 0.8772,
+      "step": 4024
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 0.0001984903846153846,
+      "loss": 0.8323,
+      "step": 4025
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 0.00019846153846153844,
+      "loss": 0.8795,
+      "step": 4026
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 0.0001984326923076923,
+      "loss": 0.959,
+      "step": 4027
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 0.00019840384615384612,
+      "loss": 0.9062,
+      "step": 4028
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 0.00019837499999999998,
+      "loss": 0.8658,
+      "step": 4029
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 0.00019834615384615385,
+      "loss": 0.849,
+      "step": 4030
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 0.00019831730769230766,
+      "loss": 0.8838,
+      "step": 4031
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 0.00019828846153846153,
+      "loss": 0.896,
+      "step": 4032
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 0.00019825961538461536,
+      "loss": 0.7994,
+      "step": 4033
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 0.00019823076923076923,
+      "loss": 0.8271,
+      "step": 4034
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 0.00019820192307692307,
+      "loss": 0.8933,
+      "step": 4035
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 0.0001981730769230769,
+      "loss": 0.8735,
+      "step": 4036
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 0.00019814423076923077,
+      "loss": 0.8814,
+      "step": 4037
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 0.00019811538461538458,
+      "loss": 0.859,
+      "step": 4038
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 0.00019808653846153845,
+      "loss": 0.9138,
+      "step": 4039
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 0.0001980576923076923,
+      "loss": 0.8517,
+      "step": 4040
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 0.00019802884615384612,
+      "loss": 0.7521,
+      "step": 4041
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 0.000198,
+      "loss": 0.7688,
+      "step": 4042
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 0.00019797115384615383,
+      "loss": 1.0019,
+      "step": 4043
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 0.00019794230769230766,
+      "loss": 0.9702,
+      "step": 4044
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 0.00019791346153846153,
+      "loss": 0.7081,
+      "step": 4045
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 0.00019788461538461537,
+      "loss": 0.7524,
+      "step": 4046
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 0.00019785576923076923,
+      "loss": 0.875,
+      "step": 4047
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 0.00019782692307692304,
+      "loss": 0.7704,
+      "step": 4048
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 0.0001977980769230769,
+      "loss": 0.8036,
+      "step": 4049
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 0.00019776923076923075,
+      "loss": 0.981,
+      "step": 4050
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 0.00019774038461538459,
+      "loss": 0.8231,
+      "step": 4051
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 0.00019771153846153845,
+      "loss": 0.8184,
+      "step": 4052
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 0.0001976826923076923,
+      "loss": 0.9,
+      "step": 4053
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 0.00019765384615384613,
+      "loss": 0.9152,
+      "step": 4054
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 0.00019762499999999997,
+      "loss": 0.8597,
+      "step": 4055
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 0.00019759615384615383,
+      "loss": 0.9396,
+      "step": 4056
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 0.0001975673076923077,
+      "loss": 0.9219,
+      "step": 4057
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 0.0001975384615384615,
+      "loss": 0.8418,
+      "step": 4058
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 0.00019750961538461537,
+      "loss": 0.8462,
+      "step": 4059
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 0.0001974807692307692,
+      "loss": 0.7243,
+      "step": 4060
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 0.00019745192307692305,
+      "loss": 0.9801,
+      "step": 4061
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 0.00019742307692307691,
+      "loss": 0.8723,
+      "step": 4062
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 0.00019739423076923075,
+      "loss": 0.9216,
+      "step": 4063
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 0.0001973653846153846,
+      "loss": 0.8339,
+      "step": 4064
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 0.00019733653846153843,
+      "loss": 0.7834,
+      "step": 4065
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 0.0001973076923076923,
+      "loss": 0.8157,
+      "step": 4066
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 0.00019727884615384616,
+      "loss": 0.8052,
+      "step": 4067
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 0.00019724999999999997,
+      "loss": 0.8701,
+      "step": 4068
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 0.00019722115384615384,
+      "loss": 0.8506,
+      "step": 4069
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 0.00019719230769230767,
+      "loss": 0.8462,
+      "step": 4070
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 0.0001971634615384615,
+      "loss": 0.8906,
+      "step": 4071
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 0.00019713461538461538,
+      "loss": 0.915,
+      "step": 4072
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 0.00019710576923076921,
+      "loss": 0.7607,
+      "step": 4073
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 0.00019707692307692305,
+      "loss": 0.7509,
+      "step": 4074
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 0.0001970480769230769,
+      "loss": 0.8329,
+      "step": 4075
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 0.00019701923076923076,
+      "loss": 0.8861,
+      "step": 4076
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 0.00019699038461538462,
+      "loss": 0.728,
+      "step": 4077
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 0.00019696153846153843,
+      "loss": 0.8979,
+      "step": 4078
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 0.0001969326923076923,
+      "loss": 0.75,
+      "step": 4079
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 0.00019690384615384614,
+      "loss": 0.8771,
+      "step": 4080
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 0.00019687499999999997,
+      "loss": 0.8692,
+      "step": 4081
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 0.00019684615384615384,
+      "loss": 0.9286,
+      "step": 4082
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 0.00019681730769230768,
+      "loss": 0.8092,
+      "step": 4083
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 0.00019678846153846152,
+      "loss": 0.8786,
+      "step": 4084
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 0.00019675961538461535,
+      "loss": 0.8085,
+      "step": 4085
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 0.00019673076923076922,
+      "loss": 0.8606,
+      "step": 4086
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 0.00019670192307692308,
+      "loss": 0.7795,
+      "step": 4087
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 0.0001966730769230769,
+      "loss": 0.9224,
+      "step": 4088
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 0.00019664423076923076,
+      "loss": 0.7952,
+      "step": 4089
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 0.0001966153846153846,
+      "loss": 0.7356,
+      "step": 4090
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 0.00019658653846153844,
+      "loss": 0.9082,
+      "step": 4091
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 0.0001965576923076923,
+      "loss": 0.7883,
+      "step": 4092
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 0.00019652884615384614,
+      "loss": 0.841,
+      "step": 4093
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 0.00019649999999999998,
+      "loss": 0.8717,
+      "step": 4094
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 0.00019647115384615382,
+      "loss": 0.7857,
+      "step": 4095
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 0.00019644230769230768,
+      "loss": 0.7557,
+      "step": 4096
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 0.00019641346153846155,
+      "loss": 0.9144,
+      "step": 4097
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 0.00019638461538461536,
+      "loss": 0.7096,
+      "step": 4098
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 0.00019635576923076922,
+      "loss": 0.8648,
+      "step": 4099
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 0.00019632692307692306,
+      "loss": 0.7757,
+      "step": 4100
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 0.0001962980769230769,
+      "loss": 0.8968,
+      "step": 4101
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 0.00019626923076923077,
+      "loss": 0.9065,
+      "step": 4102
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 0.0001962403846153846,
+      "loss": 0.8856,
+      "step": 4103
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 0.00019621153846153844,
+      "loss": 0.7671,
+      "step": 4104
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 0.00019618269230769228,
+      "loss": 0.8353,
+      "step": 4105
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 0.00019615384615384615,
+      "loss": 0.8487,
+      "step": 4106
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 0.000196125,
+      "loss": 0.9914,
+      "step": 4107
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 0.00019609615384615382,
+      "loss": 0.8343,
+      "step": 4108
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 0.0001960673076923077,
+      "loss": 0.8452,
+      "step": 4109
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 0.0001960384615384615,
+      "loss": 0.8499,
+      "step": 4110
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 0.00019600961538461536,
+      "loss": 0.8013,
+      "step": 4111
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 0.00019598076923076923,
+      "loss": 0.7669,
+      "step": 4112
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 0.00019595192307692307,
+      "loss": 0.7291,
+      "step": 4113
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 0.0001959230769230769,
+      "loss": 0.8037,
+      "step": 4114
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 0.00019589423076923074,
+      "loss": 0.8567,
+      "step": 4115
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 0.0001958653846153846,
+      "loss": 0.7944,
+      "step": 4116
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 0.00019583653846153842,
+      "loss": 0.7427,
+      "step": 4117
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 0.00019580769230769228,
+      "loss": 1.0378,
+      "step": 4118
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 0.00019577884615384615,
+      "loss": 0.927,
+      "step": 4119
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 0.00019574999999999996,
+      "loss": 0.9413,
+      "step": 4120
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 0.00019572115384615383,
+      "loss": 0.7734,
+      "step": 4121
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 0.00019569230769230766,
+      "loss": 0.8527,
+      "step": 4122
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 0.00019566346153846153,
+      "loss": 0.8146,
+      "step": 4123
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 0.00019563461538461537,
+      "loss": 0.7667,
+      "step": 4124
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 0.0001956057692307692,
+      "loss": 0.9965,
+      "step": 4125
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 0.00019557692307692307,
+      "loss": 0.9427,
+      "step": 4126
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 0.00019554807692307688,
+      "loss": 0.9469,
+      "step": 4127
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 0.00019551923076923075,
+      "loss": 0.7711,
+      "step": 4128
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 0.0001954903846153846,
+      "loss": 0.8477,
+      "step": 4129
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 0.00019546153846153842,
+      "loss": 0.8523,
+      "step": 4130
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 0.0001954326923076923,
+      "loss": 0.7941,
+      "step": 4131
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 0.00019540384615384613,
+      "loss": 0.8309,
+      "step": 4132
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 0.000195375,
+      "loss": 1.03,
+      "step": 4133
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 0.00019534615384615383,
+      "loss": 0.8905,
+      "step": 4134
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 0.00019531730769230767,
+      "loss": 0.7621,
+      "step": 4135
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 0.00019528846153846153,
+      "loss": 0.9066,
+      "step": 4136
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 0.00019525961538461534,
+      "loss": 0.9265,
+      "step": 4137
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 0.0001952307692307692,
+      "loss": 0.7745,
+      "step": 4138
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 0.00019520192307692308,
+      "loss": 0.8942,
+      "step": 4139
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 0.00019517307692307689,
+      "loss": 0.9003,
+      "step": 4140
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 0.00019514423076923075,
+      "loss": 0.7688,
+      "step": 4141
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 0.0001951153846153846,
+      "loss": 0.8349,
+      "step": 4142
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 0.00019508653846153845,
+      "loss": 0.82,
+      "step": 4143
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 0.0001950576923076923,
+      "loss": 0.7985,
+      "step": 4144
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 0.00019502884615384613,
+      "loss": 0.9304,
+      "step": 4145
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 0.000195,
+      "loss": 0.9517,
+      "step": 4146
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 0.0001949711538461538,
+      "loss": 0.8806,
+      "step": 4147
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 0.00019494230769230767,
+      "loss": 0.7724,
+      "step": 4148
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 0.00019491346153846154,
+      "loss": 0.8021,
+      "step": 4149
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 0.00019488461538461535,
+      "loss": 0.7546,
+      "step": 4150
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 0.00019485576923076921,
+      "loss": 0.7503,
+      "step": 4151
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 0.00019482692307692305,
+      "loss": 0.7656,
+      "step": 4152
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 0.00019479807692307692,
+      "loss": 0.8682,
+      "step": 4153
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 0.00019476923076923076,
+      "loss": 0.8659,
+      "step": 4154
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 0.0001947403846153846,
+      "loss": 0.8722,
+      "step": 4155
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 0.00019471153846153846,
+      "loss": 0.8624,
+      "step": 4156
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 0.00019468269230769227,
+      "loss": 0.7819,
+      "step": 4157
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 0.00019465384615384614,
+      "loss": 0.8487,
+      "step": 4158
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 0.000194625,
+      "loss": 0.9169,
+      "step": 4159
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 0.0001945961538461538,
+      "loss": 0.8849,
+      "step": 4160
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 0.00019456730769230768,
+      "loss": 0.8205,
+      "step": 4161
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 0.00019453846153846152,
+      "loss": 0.8577,
+      "step": 4162
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 0.00019450961538461538,
+      "loss": 0.7789,
+      "step": 4163
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 0.00019448076923076922,
+      "loss": 0.9189,
+      "step": 4164
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 0.00019445192307692306,
+      "loss": 0.824,
+      "step": 4165
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 0.00019442307692307692,
+      "loss": 0.8029,
+      "step": 4166
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 0.00019439423076923073,
+      "loss": 0.9408,
+      "step": 4167
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 0.0001943653846153846,
+      "loss": 0.8798,
+      "step": 4168
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 0.00019433653846153846,
+      "loss": 0.7898,
+      "step": 4169
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 0.00019430769230769227,
+      "loss": 0.8992,
+      "step": 4170
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 0.00019427884615384614,
+      "loss": 0.7971,
+      "step": 4171
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 0.00019424999999999998,
+      "loss": 1.0344,
+      "step": 4172
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 0.00019422115384615384,
+      "loss": 0.9948,
+      "step": 4173
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 0.00019419230769230768,
+      "loss": 0.8349,
+      "step": 4174
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 0.00019416346153846152,
+      "loss": 0.8522,
+      "step": 4175
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 0.00019413461538461539,
+      "loss": 0.7951,
+      "step": 4176
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 0.0001941057692307692,
+      "loss": 0.805,
+      "step": 4177
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 0.00019407692307692306,
+      "loss": 0.7551,
+      "step": 4178
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 0.0001940480769230769,
+      "loss": 0.9504,
+      "step": 4179
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 0.00019401923076923074,
+      "loss": 0.9664,
+      "step": 4180
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 0.0001939903846153846,
+      "loss": 0.8063,
+      "step": 4181
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 0.00019396153846153844,
+      "loss": 0.7322,
+      "step": 4182
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 0.0001939326923076923,
+      "loss": 0.7388,
+      "step": 4183
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 0.00019390384615384612,
+      "loss": 0.8658,
+      "step": 4184
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 0.00019387499999999998,
+      "loss": 0.9165,
+      "step": 4185
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 0.00019384615384615385,
+      "loss": 0.8076,
+      "step": 4186
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 0.00019381730769230766,
+      "loss": 0.8318,
+      "step": 4187
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 0.00019378846153846152,
+      "loss": 0.9131,
+      "step": 4188
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 0.00019375961538461536,
+      "loss": 0.8074,
+      "step": 4189
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 0.0001937307692307692,
+      "loss": 0.8405,
+      "step": 4190
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 0.00019370192307692307,
+      "loss": 0.8811,
+      "step": 4191
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 0.0001936730769230769,
+      "loss": 0.9656,
+      "step": 4192
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 0.00019364423076923077,
+      "loss": 0.9017,
+      "step": 4193
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 0.00019361538461538458,
+      "loss": 0.8508,
+      "step": 4194
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 0.00019358653846153845,
+      "loss": 0.8689,
+      "step": 4195
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 0.0001935576923076923,
+      "loss": 0.8934,
+      "step": 4196
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 0.00019352884615384612,
+      "loss": 0.9164,
+      "step": 4197
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 0.0001935,
+      "loss": 0.9273,
+      "step": 4198
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 0.00019347115384615383,
+      "loss": 0.8674,
+      "step": 4199
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 0.00019344230769230766,
+      "loss": 0.8401,
+      "step": 4200
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 0.00019341346153846153,
+      "loss": 0.7384,
+      "step": 4201
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 0.00019338461538461537,
+      "loss": 0.7453,
+      "step": 4202
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 0.00019335576923076923,
+      "loss": 0.9407,
+      "step": 4203
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 0.00019332692307692304,
+      "loss": 0.8337,
+      "step": 4204
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 0.0001932980769230769,
+      "loss": 0.8274,
+      "step": 4205
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 0.00019326923076923077,
+      "loss": 0.8405,
+      "step": 4206
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 0.00019324038461538458,
+      "loss": 0.7806,
+      "step": 4207
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 0.00019321153846153845,
+      "loss": 0.8231,
+      "step": 4208
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 0.0001931826923076923,
+      "loss": 0.851,
+      "step": 4209
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 0.00019315384615384613,
+      "loss": 0.7125,
+      "step": 4210
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 0.000193125,
+      "loss": 0.782,
+      "step": 4211
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 0.00019309615384615383,
+      "loss": 0.914,
+      "step": 4212
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 0.0001930673076923077,
+      "loss": 0.9692,
+      "step": 4213
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 0.0001930384615384615,
+      "loss": 0.7576,
+      "step": 4214
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 0.00019300961538461537,
+      "loss": 0.9033,
+      "step": 4215
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 0.00019298076923076924,
+      "loss": 0.8629,
+      "step": 4216
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 0.00019295192307692305,
+      "loss": 0.8145,
+      "step": 4217
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 0.0001929230769230769,
+      "loss": 0.8521,
+      "step": 4218
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 0.00019289423076923075,
+      "loss": 0.8124,
+      "step": 4219
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 0.0001928653846153846,
+      "loss": 0.8232,
+      "step": 4220
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 0.00019283653846153845,
+      "loss": 0.8173,
+      "step": 4221
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 0.0001928076923076923,
+      "loss": 0.7966,
+      "step": 4222
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 0.00019277884615384616,
+      "loss": 0.8787,
+      "step": 4223
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 0.00019274999999999997,
+      "loss": 0.9537,
+      "step": 4224
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 0.00019272115384615383,
+      "loss": 0.7796,
+      "step": 4225
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 0.0001926923076923077,
+      "loss": 0.9692,
+      "step": 4226
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 0.0001926634615384615,
+      "loss": 0.9271,
+      "step": 4227
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 0.00019263461538461538,
+      "loss": 0.7584,
+      "step": 4228
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 0.00019260576923076921,
+      "loss": 0.8014,
+      "step": 4229
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 0.00019257692307692305,
+      "loss": 0.7879,
+      "step": 4230
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 0.00019254807692307692,
+      "loss": 0.922,
+      "step": 4231
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 0.00019251923076923076,
+      "loss": 0.795,
+      "step": 4232
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 0.00019249038461538462,
+      "loss": 0.8207,
+      "step": 4233
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 0.00019246153846153843,
+      "loss": 0.8538,
+      "step": 4234
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 0.0001924326923076923,
+      "loss": 0.7904,
+      "step": 4235
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 0.00019240384615384616,
+      "loss": 0.7892,
+      "step": 4236
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 0.00019237499999999997,
+      "loss": 0.8823,
+      "step": 4237
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 0.00019234615384615384,
+      "loss": 0.7648,
+      "step": 4238
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 0.00019231730769230768,
+      "loss": 0.7777,
+      "step": 4239
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 0.00019228846153846151,
+      "loss": 0.9034,
+      "step": 4240
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 0.00019225961538461538,
+      "loss": 0.8049,
+      "step": 4241
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 0.00019223076923076922,
+      "loss": 0.8321,
+      "step": 4242
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 0.00019220192307692308,
+      "loss": 0.824,
+      "step": 4243
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 0.0001921730769230769,
+      "loss": 0.7378,
+      "step": 4244
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 0.00019214423076923076,
+      "loss": 0.8168,
+      "step": 4245
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 0.00019211538461538457,
+      "loss": 0.8844,
+      "step": 4246
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 0.00019208653846153844,
+      "loss": 1.0311,
+      "step": 4247
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 0.0001920576923076923,
+      "loss": 0.875,
+      "step": 4248
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 0.00019202884615384614,
+      "loss": 0.8079,
+      "step": 4249
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 0.7708,
+      "step": 4250
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 0.00019197115384615382,
+      "loss": 0.8643,
+      "step": 4251
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 0.00019194230769230768,
+      "loss": 0.787,
+      "step": 4252
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 0.00019191346153846152,
+      "loss": 0.7804,
+      "step": 4253
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 0.00019188461538461536,
+      "loss": 0.8621,
+      "step": 4254
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 0.00019185576923076922,
+      "loss": 0.7634,
+      "step": 4255
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 0.00019182692307692303,
+      "loss": 0.7125,
+      "step": 4256
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 0.0001917980769230769,
+      "loss": 0.8905,
+      "step": 4257
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 0.00019176923076923076,
+      "loss": 0.8226,
+      "step": 4258
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 0.0001917403846153846,
+      "loss": 0.7479,
+      "step": 4259
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 0.00019171153846153844,
+      "loss": 0.8921,
+      "step": 4260
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 0.00019168269230769228,
+      "loss": 0.9309,
+      "step": 4261
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 0.00019165384615384614,
+      "loss": 0.8074,
+      "step": 4262
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 0.00019162499999999998,
+      "loss": 0.7594,
+      "step": 4263
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 0.00019159615384615382,
+      "loss": 0.9795,
+      "step": 4264
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 0.00019156730769230769,
+      "loss": 0.8906,
+      "step": 4265
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 0.0001915384615384615,
+      "loss": 0.7917,
+      "step": 4266
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 0.00019150961538461536,
+      "loss": 0.8758,
+      "step": 4267
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 0.00019148076923076923,
+      "loss": 0.8316,
+      "step": 4268
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 0.00019145192307692307,
+      "loss": 0.8702,
+      "step": 4269
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 0.0001914230769230769,
+      "loss": 0.9747,
+      "step": 4270
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 0.00019139423076923074,
+      "loss": 0.799,
+      "step": 4271
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 0.0001913653846153846,
+      "loss": 0.7483,
+      "step": 4272
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 0.00019133653846153845,
+      "loss": 0.9354,
+      "step": 4273
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 0.00019130769230769228,
+      "loss": 0.9161,
+      "step": 4274
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 0.00019127884615384615,
+      "loss": 0.8529,
+      "step": 4275
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 0.00019124999999999996,
+      "loss": 0.6842,
+      "step": 4276
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 0.00019122115384615382,
+      "loss": 0.9158,
+      "step": 4277
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 0.0001911923076923077,
+      "loss": 0.8652,
+      "step": 4278
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 0.00019116346153846153,
+      "loss": 0.886,
+      "step": 4279
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 0.00019113461538461537,
+      "loss": 0.8522,
+      "step": 4280
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 0.0001911057692307692,
+      "loss": 0.8539,
+      "step": 4281
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 0.00019107692307692307,
+      "loss": 0.8989,
+      "step": 4282
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 0.0001910480769230769,
+      "loss": 0.7969,
+      "step": 4283
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 0.00019101923076923075,
+      "loss": 0.7794,
+      "step": 4284
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 0.0001909903846153846,
+      "loss": 0.8171,
+      "step": 4285
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 0.00019096153846153842,
+      "loss": 0.8622,
+      "step": 4286
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 0.0001909326923076923,
+      "loss": 0.9009,
+      "step": 4287
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 0.00019090384615384615,
+      "loss": 0.7837,
+      "step": 4288
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 0.000190875,
+      "loss": 0.7349,
+      "step": 4289
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 0.00019084615384615383,
+      "loss": 0.7589,
+      "step": 4290
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 0.00019081730769230767,
+      "loss": 0.707,
+      "step": 4291
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 0.00019078846153846153,
+      "loss": 0.9258,
+      "step": 4292
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 0.00019075961538461537,
+      "loss": 0.9761,
+      "step": 4293
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 0.0001907307692307692,
+      "loss": 0.8467,
+      "step": 4294
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 0.00019070192307692307,
+      "loss": 0.7836,
+      "step": 4295
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 0.00019067307692307689,
+      "loss": 0.8232,
+      "step": 4296
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 0.00019064423076923075,
+      "loss": 0.9013,
+      "step": 4297
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 0.00019061538461538462,
+      "loss": 0.8726,
+      "step": 4298
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 0.00019058653846153845,
+      "loss": 0.6896,
+      "step": 4299
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 0.0001905576923076923,
+      "loss": 0.8644,
+      "step": 4300
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 0.00019052884615384613,
+      "loss": 0.755,
+      "step": 4301
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 0.0001905,
+      "loss": 0.7961,
+      "step": 4302
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 0.00019047115384615383,
+      "loss": 0.8841,
+      "step": 4303
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 0.00019044230769230767,
+      "loss": 0.801,
+      "step": 4304
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 0.00019041346153846154,
+      "loss": 0.8227,
+      "step": 4305
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 0.00019038461538461535,
+      "loss": 0.8083,
+      "step": 4306
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 0.0001903557692307692,
+      "loss": 0.9444,
+      "step": 4307
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 0.00019032692307692308,
+      "loss": 0.8687,
+      "step": 4308
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 0.0001902980769230769,
+      "loss": 0.8009,
+      "step": 4309
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 0.00019026923076923075,
+      "loss": 0.7954,
+      "step": 4310
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 0.0001902403846153846,
+      "loss": 0.7881,
+      "step": 4311
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 0.00019021153846153846,
+      "loss": 0.804,
+      "step": 4312
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 0.00019018269230769227,
+      "loss": 0.8508,
+      "step": 4313
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 0.00019015384615384613,
+      "loss": 0.855,
+      "step": 4314
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 0.000190125,
+      "loss": 0.7965,
+      "step": 4315
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 0.0001900961538461538,
+      "loss": 0.8193,
+      "step": 4316
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 0.00019006730769230768,
+      "loss": 0.7496,
+      "step": 4317
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 0.00019003846153846151,
+      "loss": 0.7795,
+      "step": 4318
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 0.00019000961538461535,
+      "loss": 0.845,
+      "step": 4319
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 0.00018998076923076922,
+      "loss": 0.8373,
+      "step": 4320
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 0.00018995192307692306,
+      "loss": 0.8287,
+      "step": 4321
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 0.00018992307692307692,
+      "loss": 0.7976,
+      "step": 4322
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 0.00018989423076923073,
+      "loss": 0.856,
+      "step": 4323
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 0.0001898653846153846,
+      "loss": 0.8516,
+      "step": 4324
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 0.00018983653846153846,
+      "loss": 0.9312,
+      "step": 4325
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 0.00018980769230769227,
+      "loss": 0.8188,
+      "step": 4326
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 0.00018977884615384614,
+      "loss": 0.9284,
+      "step": 4327
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 0.00018974999999999998,
+      "loss": 0.8454,
+      "step": 4328
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 0.00018972115384615382,
+      "loss": 0.6833,
+      "step": 4329
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 0.00018969230769230768,
+      "loss": 1.0084,
+      "step": 4330
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 0.00018966346153846152,
+      "loss": 0.7065,
+      "step": 4331
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 0.00018963461538461538,
+      "loss": 0.865,
+      "step": 4332
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 0.0001896057692307692,
+      "loss": 0.8892,
+      "step": 4333
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 0.00018957692307692306,
+      "loss": 0.8017,
+      "step": 4334
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 0.00018954807692307693,
+      "loss": 0.8272,
+      "step": 4335
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 0.00018951923076923074,
+      "loss": 0.824,
+      "step": 4336
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 0.0001894903846153846,
+      "loss": 0.835,
+      "step": 4337
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 0.00018946153846153844,
+      "loss": 0.7358,
+      "step": 4338
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 0.00018943269230769228,
+      "loss": 0.9655,
+      "step": 4339
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 0.00018940384615384614,
+      "loss": 0.8567,
+      "step": 4340
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 0.00018937499999999998,
+      "loss": 0.8742,
+      "step": 4341
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 0.00018934615384615385,
+      "loss": 0.8806,
+      "step": 4342
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 0.00018931730769230766,
+      "loss": 0.8662,
+      "step": 4343
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 0.00018928846153846152,
+      "loss": 0.9298,
+      "step": 4344
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 0.0001892596153846154,
+      "loss": 0.8047,
+      "step": 4345
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 0.0001892307692307692,
+      "loss": 0.7671,
+      "step": 4346
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 0.00018920192307692306,
+      "loss": 0.7841,
+      "step": 4347
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 0.0001891730769230769,
+      "loss": 0.6909,
+      "step": 4348
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 0.00018914423076923074,
+      "loss": 0.8252,
+      "step": 4349
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 0.0001891153846153846,
+      "loss": 0.8596,
+      "step": 4350
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 0.00018908653846153844,
+      "loss": 0.9463,
+      "step": 4351
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 0.0001890576923076923,
+      "loss": 0.8947,
+      "step": 4352
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 0.00018902884615384612,
+      "loss": 0.8477,
+      "step": 4353
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 0.6889,
+      "step": 4354
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 0.00018897115384615385,
+      "loss": 0.886,
+      "step": 4355
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 0.00018894230769230766,
+      "loss": 0.8151,
+      "step": 4356
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 0.00018891346153846153,
+      "loss": 0.7846,
+      "step": 4357
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 0.00018888461538461537,
+      "loss": 0.9215,
+      "step": 4358
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 0.0001888557692307692,
+      "loss": 0.7921,
+      "step": 4359
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 0.00018882692307692307,
+      "loss": 0.8044,
+      "step": 4360
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 0.0001887980769230769,
+      "loss": 0.8149,
+      "step": 4361
+    },
+    {
+      "epoch": 40.02,
+      "learning_rate": 0.00018876923076923077,
+      "loss": 0.7315,
+      "step": 4362
+    },
+    {
+      "epoch": 40.03,
+      "learning_rate": 0.00018874038461538458,
+      "loss": 0.8343,
+      "step": 4363
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 0.00018871153846153845,
+      "loss": 0.8424,
+      "step": 4364
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 0.00018868269230769231,
+      "loss": 0.8314,
+      "step": 4365
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 0.00018865384615384613,
+      "loss": 0.8295,
+      "step": 4366
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 0.000188625,
+      "loss": 1.0012,
+      "step": 4367
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 0.00018859615384615383,
+      "loss": 0.7465,
+      "step": 4368
+    },
+    {
+      "epoch": 40.08,
+      "learning_rate": 0.00018856730769230767,
+      "loss": 0.5933,
+      "step": 4369
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 0.00018853846153846153,
+      "loss": 0.8773,
+      "step": 4370
+    },
+    {
+      "epoch": 40.1,
+      "learning_rate": 0.00018850961538461537,
+      "loss": 0.906,
+      "step": 4371
+    },
+    {
+      "epoch": 40.11,
+      "learning_rate": 0.00018848076923076924,
+      "loss": 0.8616,
+      "step": 4372
+    },
+    {
+      "epoch": 40.12,
+      "learning_rate": 0.00018845192307692305,
+      "loss": 1.0085,
+      "step": 4373
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 0.0001884230769230769,
+      "loss": 0.902,
+      "step": 4374
+    },
+    {
+      "epoch": 40.14,
+      "learning_rate": 0.00018839423076923072,
+      "loss": 0.8388,
+      "step": 4375
+    },
+    {
+      "epoch": 40.15,
+      "learning_rate": 0.0001883653846153846,
+      "loss": 0.8891,
+      "step": 4376
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 0.00018833653846153845,
+      "loss": 0.8011,
+      "step": 4377
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 0.0001883076923076923,
+      "loss": 0.8151,
+      "step": 4378
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 0.00018827884615384613,
+      "loss": 0.7754,
+      "step": 4379
+    },
+    {
+      "epoch": 40.18,
+      "learning_rate": 0.00018824999999999997,
+      "loss": 0.8657,
+      "step": 4380
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 0.00018822115384615383,
+      "loss": 0.8011,
+      "step": 4381
+    },
+    {
+      "epoch": 40.2,
+      "learning_rate": 0.0001881923076923077,
+      "loss": 0.8019,
+      "step": 4382
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 0.0001881634615384615,
+      "loss": 0.8483,
+      "step": 4383
+    },
+    {
+      "epoch": 40.22,
+      "learning_rate": 0.00018813461538461537,
+      "loss": 0.8813,
+      "step": 4384
+    },
+    {
+      "epoch": 40.23,
+      "learning_rate": 0.00018810576923076919,
+      "loss": 0.9546,
+      "step": 4385
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 0.00018807692307692305,
+      "loss": 0.9045,
+      "step": 4386
+    },
+    {
+      "epoch": 40.25,
+      "learning_rate": 0.00018804807692307692,
+      "loss": 0.8365,
+      "step": 4387
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 0.00018801923076923075,
+      "loss": 0.8044,
+      "step": 4388
+    },
+    {
+      "epoch": 40.27,
+      "learning_rate": 0.0001879903846153846,
+      "loss": 0.8024,
+      "step": 4389
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 0.00018796153846153843,
+      "loss": 0.7806,
+      "step": 4390
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 0.0001879326923076923,
+      "loss": 0.8172,
+      "step": 4391
+    },
+    {
+      "epoch": 40.29,
+      "learning_rate": 0.00018790384615384616,
+      "loss": 0.6542,
+      "step": 4392
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 0.00018787499999999997,
+      "loss": 0.6941,
+      "step": 4393
+    },
+    {
+      "epoch": 40.31,
+      "learning_rate": 0.00018784615384615384,
+      "loss": 0.8064,
+      "step": 4394
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 0.00018781730769230765,
+      "loss": 0.704,
+      "step": 4395
+    },
+    {
+      "epoch": 40.33,
+      "learning_rate": 0.00018778846153846151,
+      "loss": 0.8606,
+      "step": 4396
+    },
+    {
+      "epoch": 40.34,
+      "learning_rate": 0.00018775961538461538,
+      "loss": 0.9368,
+      "step": 4397
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 0.00018773076923076922,
+      "loss": 0.7649,
+      "step": 4398
+    },
+    {
+      "epoch": 40.36,
+      "learning_rate": 0.00018770192307692306,
+      "loss": 0.8815,
+      "step": 4399
+    },
+    {
+      "epoch": 40.37,
+      "learning_rate": 0.0001876730769230769,
+      "loss": 0.7668,
+      "step": 4400
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 0.00018764423076923076,
+      "loss": 0.8785,
+      "step": 4401
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 0.00018761538461538462,
+      "loss": 0.8348,
+      "step": 4402
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 0.00018758653846153844,
+      "loss": 0.7896,
+      "step": 4403
+    },
+    {
+      "epoch": 40.4,
+      "learning_rate": 0.0001875576923076923,
+      "loss": 0.9604,
+      "step": 4404
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 0.0001875288461538461,
+      "loss": 0.8164,
+      "step": 4405
+    },
+    {
+      "epoch": 40.42,
+      "learning_rate": 0.00018749999999999998,
+      "loss": 0.7329,
+      "step": 4406
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 0.00018747115384615384,
+      "loss": 0.8868,
+      "step": 4407
+    },
+    {
+      "epoch": 40.44,
+      "learning_rate": 0.00018744230769230768,
+      "loss": 0.7833,
+      "step": 4408
+    },
+    {
+      "epoch": 40.45,
+      "learning_rate": 0.00018741346153846152,
+      "loss": 0.9236,
+      "step": 4409
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 0.00018738461538461536,
+      "loss": 0.903,
+      "step": 4410
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 0.00018735576923076922,
+      "loss": 0.7335,
+      "step": 4411
+    },
+    {
+      "epoch": 40.48,
+      "learning_rate": 0.0001873269230769231,
+      "loss": 0.7893,
+      "step": 4412
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 0.0001872980769230769,
+      "loss": 0.7362,
+      "step": 4413
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 0.00018726923076923076,
+      "loss": 0.8008,
+      "step": 4414
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 0.00018724038461538457,
+      "loss": 0.918,
+      "step": 4415
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 0.00018721153846153844,
+      "loss": 0.9587,
+      "step": 4416
+    },
+    {
+      "epoch": 40.52,
+      "learning_rate": 0.0001871826923076923,
+      "loss": 0.8271,
+      "step": 4417
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 0.00018715384615384614,
+      "loss": 0.9438,
+      "step": 4418
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 0.00018712499999999998,
+      "loss": 0.8702,
+      "step": 4419
+    },
+    {
+      "epoch": 40.55,
+      "learning_rate": 0.00018709615384615382,
+      "loss": 0.8373,
+      "step": 4420
+    },
+    {
+      "epoch": 40.56,
+      "learning_rate": 0.00018706730769230768,
+      "loss": 0.9904,
+      "step": 4421
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 0.00018703846153846155,
+      "loss": 0.8361,
+      "step": 4422
+    },
+    {
+      "epoch": 40.58,
+      "learning_rate": 0.00018700961538461536,
+      "loss": 0.8749,
+      "step": 4423
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 0.00018698076923076923,
+      "loss": 0.8781,
+      "step": 4424
+    },
+    {
+      "epoch": 40.6,
+      "learning_rate": 0.00018695192307692304,
+      "loss": 0.769,
+      "step": 4425
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 0.0001869230769230769,
+      "loss": 0.8195,
+      "step": 4426
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 0.00018689423076923077,
+      "loss": 0.7899,
+      "step": 4427
+    },
+    {
+      "epoch": 40.62,
+      "learning_rate": 0.0001868653846153846,
+      "loss": 0.8968,
+      "step": 4428
+    },
+    {
+      "epoch": 40.63,
+      "learning_rate": 0.00018683653846153844,
+      "loss": 0.8218,
+      "step": 4429
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 0.00018680769230769228,
+      "loss": 0.9187,
+      "step": 4430
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 0.00018677884615384615,
+      "loss": 0.8182,
+      "step": 4431
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 0.00018675,
+      "loss": 0.7926,
+      "step": 4432
+    },
+    {
+      "epoch": 40.67,
+      "learning_rate": 0.00018672115384615382,
+      "loss": 0.7235,
+      "step": 4433
+    },
+    {
+      "epoch": 40.68,
+      "learning_rate": 0.0001866923076923077,
+      "loss": 0.8637,
+      "step": 4434
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 0.0001866634615384615,
+      "loss": 0.8035,
+      "step": 4435
+    },
+    {
+      "epoch": 40.7,
+      "learning_rate": 0.00018663461538461537,
+      "loss": 0.9892,
+      "step": 4436
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 0.00018660576923076923,
+      "loss": 0.835,
+      "step": 4437
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 0.00018657692307692307,
+      "loss": 0.7764,
+      "step": 4438
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 0.0001865480769230769,
+      "loss": 0.7526,
+      "step": 4439
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 0.00018651923076923075,
+      "loss": 0.783,
+      "step": 4440
+    },
+    {
+      "epoch": 40.74,
+      "learning_rate": 0.0001864903846153846,
+      "loss": 0.8963,
+      "step": 4441
+    },
+    {
+      "epoch": 40.75,
+      "learning_rate": 0.00018646153846153842,
+      "loss": 0.7791,
+      "step": 4442
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 0.0001864326923076923,
+      "loss": 0.7449,
+      "step": 4443
+    },
+    {
+      "epoch": 40.77,
+      "learning_rate": 0.00018640384615384615,
+      "loss": 0.7852,
+      "step": 4444
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 0.00018637499999999996,
+      "loss": 0.8811,
+      "step": 4445
+    },
+    {
+      "epoch": 40.79,
+      "learning_rate": 0.00018634615384615383,
+      "loss": 0.6417,
+      "step": 4446
+    },
+    {
+      "epoch": 40.8,
+      "learning_rate": 0.00018631730769230767,
+      "loss": 0.835,
+      "step": 4447
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 0.00018628846153846153,
+      "loss": 0.7482,
+      "step": 4448
+    },
+    {
+      "epoch": 40.82,
+      "learning_rate": 0.00018625961538461537,
+      "loss": 0.7351,
+      "step": 4449
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 0.0001862307692307692,
+      "loss": 0.8151,
+      "step": 4450
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 0.00018620192307692307,
+      "loss": 0.758,
+      "step": 4451
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 0.00018617307692307688,
+      "loss": 0.8577,
+      "step": 4452
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 0.00018614423076923075,
+      "loss": 0.8038,
+      "step": 4453
+    },
+    {
+      "epoch": 40.86,
+      "learning_rate": 0.00018611538461538461,
+      "loss": 0.7583,
+      "step": 4454
+    },
+    {
+      "epoch": 40.87,
+      "learning_rate": 0.00018608653846153843,
+      "loss": 0.8414,
+      "step": 4455
+    },
+    {
+      "epoch": 40.88,
+      "learning_rate": 0.0001860576923076923,
+      "loss": 0.7599,
+      "step": 4456
+    },
+    {
+      "epoch": 40.89,
+      "learning_rate": 0.00018602884615384613,
+      "loss": 0.9211,
+      "step": 4457
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 0.000186,
+      "loss": 0.7822,
+      "step": 4458
+    },
+    {
+      "epoch": 40.91,
+      "learning_rate": 0.00018597115384615383,
+      "loss": 0.8426,
+      "step": 4459
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 0.00018594230769230767,
+      "loss": 0.8044,
+      "step": 4460
+    },
+    {
+      "epoch": 40.93,
+      "learning_rate": 0.00018591346153846154,
+      "loss": 0.8558,
+      "step": 4461
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 0.00018588461538461535,
+      "loss": 0.8912,
+      "step": 4462
+    },
+    {
+      "epoch": 40.94,
+      "learning_rate": 0.0001858557692307692,
+      "loss": 0.9353,
+      "step": 4463
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 0.00018582692307692308,
+      "loss": 0.9187,
+      "step": 4464
+    },
+    {
+      "epoch": 40.96,
+      "learning_rate": 0.0001857980769230769,
+      "loss": 0.8249,
+      "step": 4465
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 0.00018576923076923075,
+      "loss": 0.892,
+      "step": 4466
+    },
+    {
+      "epoch": 40.98,
+      "learning_rate": 0.0001857403846153846,
+      "loss": 0.828,
+      "step": 4467
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 0.00018571153846153846,
+      "loss": 0.8345,
+      "step": 4468
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 0.0001856826923076923,
+      "loss": 0.8322,
+      "step": 4469
+    },
+    {
+      "epoch": 41.01,
+      "learning_rate": 0.00018565384615384613,
+      "loss": 0.9028,
+      "step": 4470
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 0.000185625,
+      "loss": 0.9005,
+      "step": 4471
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 0.0001855961538461538,
+      "loss": 0.8048,
+      "step": 4472
+    },
+    {
+      "epoch": 41.04,
+      "learning_rate": 0.00018556730769230768,
+      "loss": 0.7604,
+      "step": 4473
+    },
+    {
+      "epoch": 41.05,
+      "learning_rate": 0.00018553846153846154,
+      "loss": 0.9223,
+      "step": 4474
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 0.00018550961538461535,
+      "loss": 0.8549,
+      "step": 4475
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 0.00018548076923076922,
+      "loss": 0.7758,
+      "step": 4476
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 0.00018545192307692306,
+      "loss": 0.7783,
+      "step": 4477
+    },
+    {
+      "epoch": 41.08,
+      "learning_rate": 0.00018542307692307692,
+      "loss": 0.8385,
+      "step": 4478
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 0.00018539423076923076,
+      "loss": 0.8834,
+      "step": 4479
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 0.0001853653846153846,
+      "loss": 0.8508,
+      "step": 4480
+    },
+    {
+      "epoch": 41.11,
+      "learning_rate": 0.00018533653846153846,
+      "loss": 0.8268,
+      "step": 4481
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 0.00018530769230769227,
+      "loss": 0.8861,
+      "step": 4482
+    },
+    {
+      "epoch": 41.13,
+      "learning_rate": 0.00018527884615384614,
+      "loss": 0.8852,
+      "step": 4483
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 0.00018525,
+      "loss": 0.8356,
+      "step": 4484
+    },
+    {
+      "epoch": 41.15,
+      "learning_rate": 0.00018522115384615381,
+      "loss": 0.8189,
+      "step": 4485
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 0.00018519230769230768,
+      "loss": 0.8119,
+      "step": 4486
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 0.00018516346153846152,
+      "loss": 0.8097,
+      "step": 4487
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 0.00018513461538461538,
+      "loss": 0.8274,
+      "step": 4488
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 0.00018510576923076922,
+      "loss": 0.8357,
+      "step": 4489
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 0.00018507692307692306,
+      "loss": 0.9129,
+      "step": 4490
+    },
+    {
+      "epoch": 41.2,
+      "learning_rate": 0.00018504807692307692,
+      "loss": 0.7671,
+      "step": 4491
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 0.00018501923076923074,
+      "loss": 0.8492,
+      "step": 4492
+    },
+    {
+      "epoch": 41.22,
+      "learning_rate": 0.0001849903846153846,
+      "loss": 0.9262,
+      "step": 4493
+    },
+    {
+      "epoch": 41.23,
+      "learning_rate": 0.00018496153846153847,
+      "loss": 0.7761,
+      "step": 4494
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 0.00018493269230769228,
+      "loss": 0.8751,
+      "step": 4495
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 0.00018490384615384614,
+      "loss": 0.9248,
+      "step": 4496
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 0.00018487499999999998,
+      "loss": 0.782,
+      "step": 4497
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 0.00018484615384615385,
+      "loss": 0.9537,
+      "step": 4498
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 0.00018481730769230768,
+      "loss": 0.9254,
+      "step": 4499
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 0.00018478846153846152,
+      "loss": 0.8948,
+      "step": 4500
+    },
+    {
+      "epoch": 41.28,
+      "eval_cer": 0.10054123749981259,
+      "eval_loss": 0.3714151084423065,
+      "eval_runtime": 14.1186,
+      "eval_samples_per_second": 116.655,
+      "eval_steps_per_second": 1.842,
+      "eval_wer": 0.40996833826983964,
+      "step": 4500
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 0.0001847596153846154,
+      "loss": 0.8462,
+      "step": 4501
+    },
+    {
+      "epoch": 41.3,
+      "learning_rate": 0.0001847307692307692,
+      "loss": 0.8771,
+      "step": 4502
+    },
+    {
+      "epoch": 41.31,
+      "learning_rate": 0.00018470192307692306,
+      "loss": 0.7588,
+      "step": 4503
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 0.0001846730769230769,
+      "loss": 0.7639,
+      "step": 4504
+    },
+    {
+      "epoch": 41.33,
+      "learning_rate": 0.00018464423076923074,
+      "loss": 0.9728,
+      "step": 4505
+    },
+    {
+      "epoch": 41.34,
+      "learning_rate": 0.0001846153846153846,
+      "loss": 0.9085,
+      "step": 4506
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 0.00018458653846153844,
+      "loss": 0.7594,
+      "step": 4507
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 0.00018455769230769228,
+      "loss": 0.8631,
+      "step": 4508
+    },
+    {
+      "epoch": 41.37,
+      "learning_rate": 0.00018452884615384612,
+      "loss": 0.9258,
+      "step": 4509
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 0.00018449999999999999,
+      "loss": 0.814,
+      "step": 4510
+    },
+    {
+      "epoch": 41.39,
+      "learning_rate": 0.00018447115384615385,
+      "loss": 0.8139,
+      "step": 4511
+    },
+    {
+      "epoch": 41.39,
+      "learning_rate": 0.00018444230769230766,
+      "loss": 0.7875,
+      "step": 4512
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 0.00018441346153846153,
+      "loss": 0.7858,
+      "step": 4513
+    },
+    {
+      "epoch": 41.41,
+      "learning_rate": 0.00018438461538461536,
+      "loss": 0.7635,
+      "step": 4514
+    },
+    {
+      "epoch": 41.42,
+      "learning_rate": 0.0001843557692307692,
+      "loss": 0.8186,
+      "step": 4515
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 0.00018432692307692307,
+      "loss": 0.744,
+      "step": 4516
+    },
+    {
+      "epoch": 41.44,
+      "learning_rate": 0.0001842980769230769,
+      "loss": 0.9035,
+      "step": 4517
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 0.00018426923076923074,
+      "loss": 0.7772,
+      "step": 4518
+    },
+    {
+      "epoch": 41.46,
+      "learning_rate": 0.00018424038461538458,
+      "loss": 0.7777,
+      "step": 4519
+    },
+    {
+      "epoch": 41.47,
+      "learning_rate": 0.00018421153846153845,
+      "loss": 0.8138,
+      "step": 4520
+    },
+    {
+      "epoch": 41.48,
+      "learning_rate": 0.0001841826923076923,
+      "loss": 0.8098,
+      "step": 4521
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 0.00018415384615384612,
+      "loss": 0.9035,
+      "step": 4522
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 0.000184125,
+      "loss": 0.7502,
+      "step": 4523
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 0.00018409615384615383,
+      "loss": 0.8714,
+      "step": 4524
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 0.00018406730769230767,
+      "loss": 0.8501,
+      "step": 4525
+    },
+    {
+      "epoch": 41.52,
+      "learning_rate": 0.00018403846153846153,
+      "loss": 0.9593,
+      "step": 4526
+    },
+    {
+      "epoch": 41.53,
+      "learning_rate": 0.00018400961538461537,
+      "loss": 0.8857,
+      "step": 4527
+    },
+    {
+      "epoch": 41.54,
+      "learning_rate": 0.0001839807692307692,
+      "loss": 0.7886,
+      "step": 4528
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 0.00018395192307692305,
+      "loss": 0.8189,
+      "step": 4529
+    },
+    {
+      "epoch": 41.56,
+      "learning_rate": 0.0001839230769230769,
+      "loss": 1.0303,
+      "step": 4530
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 0.00018389423076923078,
+      "loss": 0.7672,
+      "step": 4531
+    },
+    {
+      "epoch": 41.58,
+      "learning_rate": 0.0001838653846153846,
+      "loss": 0.9014,
+      "step": 4532
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 0.00018383653846153845,
+      "loss": 0.8232,
+      "step": 4533
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 0.0001838076923076923,
+      "loss": 0.7972,
+      "step": 4534
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 0.00018377884615384613,
+      "loss": 0.8299,
+      "step": 4535
+    },
+    {
+      "epoch": 41.61,
+      "learning_rate": 0.00018375,
+      "loss": 0.688,
+      "step": 4536
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 0.00018372115384615383,
+      "loss": 0.9025,
+      "step": 4537
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 0.00018369230769230767,
+      "loss": 0.8071,
+      "step": 4538
+    },
+    {
+      "epoch": 41.64,
+      "learning_rate": 0.0001836634615384615,
+      "loss": 0.8864,
+      "step": 4539
+    },
+    {
+      "epoch": 41.65,
+      "learning_rate": 0.00018363461538461537,
+      "loss": 0.7722,
+      "step": 4540
+    },
+    {
+      "epoch": 41.66,
+      "learning_rate": 0.00018360576923076924,
+      "loss": 0.796,
+      "step": 4541
+    },
+    {
+      "epoch": 41.67,
+      "learning_rate": 0.00018357692307692305,
+      "loss": 0.9036,
+      "step": 4542
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 0.00018354807692307692,
+      "loss": 0.8245,
+      "step": 4543
+    },
+    {
+      "epoch": 41.69,
+      "learning_rate": 0.00018351923076923075,
+      "loss": 0.8955,
+      "step": 4544
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 0.0001834903846153846,
+      "loss": 0.9439,
+      "step": 4545
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 0.00018346153846153846,
+      "loss": 0.7831,
+      "step": 4546
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 0.0001834326923076923,
+      "loss": 0.7325,
+      "step": 4547
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 0.00018340384615384613,
+      "loss": 0.8636,
+      "step": 4548
+    },
+    {
+      "epoch": 41.73,
+      "learning_rate": 0.00018337499999999997,
+      "loss": 0.8776,
+      "step": 4549
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 0.00018334615384615384,
+      "loss": 0.8831,
+      "step": 4550
+    },
+    {
+      "epoch": 41.75,
+      "learning_rate": 0.0001833173076923077,
+      "loss": 0.7164,
+      "step": 4551
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 0.0001832884615384615,
+      "loss": 0.8552,
+      "step": 4552
+    },
+    {
+      "epoch": 41.77,
+      "learning_rate": 0.00018325961538461538,
+      "loss": 0.7892,
+      "step": 4553
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 0.00018323076923076922,
+      "loss": 0.9359,
+      "step": 4554
+    },
+    {
+      "epoch": 41.79,
+      "learning_rate": 0.00018320192307692305,
+      "loss": 0.8373,
+      "step": 4555
+    },
+    {
+      "epoch": 41.8,
+      "learning_rate": 0.00018317307692307692,
+      "loss": 0.6541,
+      "step": 4556
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 0.00018314423076923076,
+      "loss": 0.9103,
+      "step": 4557
+    },
+    {
+      "epoch": 41.82,
+      "learning_rate": 0.0001831153846153846,
+      "loss": 0.7049,
+      "step": 4558
+    },
+    {
+      "epoch": 41.83,
+      "learning_rate": 0.00018308653846153843,
+      "loss": 0.7905,
+      "step": 4559
+    },
+    {
+      "epoch": 41.83,
+      "learning_rate": 0.0001830576923076923,
+      "loss": 0.8173,
+      "step": 4560
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 0.00018302884615384616,
+      "loss": 0.9057,
+      "step": 4561
+    },
+    {
+      "epoch": 41.85,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 0.8691,
+      "step": 4562
+    },
+    {
+      "epoch": 41.86,
+      "learning_rate": 0.00018297115384615384,
+      "loss": 0.9936,
+      "step": 4563
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 0.00018294230769230768,
+      "loss": 0.9255,
+      "step": 4564
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 0.00018291346153846152,
+      "loss": 0.7524,
+      "step": 4565
+    },
+    {
+      "epoch": 41.89,
+      "learning_rate": 0.00018288461538461538,
+      "loss": 0.8586,
+      "step": 4566
+    },
+    {
+      "epoch": 41.9,
+      "learning_rate": 0.00018285576923076922,
+      "loss": 0.8722,
+      "step": 4567
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 0.00018282692307692306,
+      "loss": 0.919,
+      "step": 4568
+    },
+    {
+      "epoch": 41.92,
+      "learning_rate": 0.0001827980769230769,
+      "loss": 0.8922,
+      "step": 4569
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 0.00018276923076923076,
+      "loss": 0.8286,
+      "step": 4570
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 0.00018274038461538457,
+      "loss": 0.8662,
+      "step": 4571
+    },
+    {
+      "epoch": 41.94,
+      "learning_rate": 0.00018271153846153844,
+      "loss": 0.7157,
+      "step": 4572
+    },
+    {
+      "epoch": 41.95,
+      "learning_rate": 0.0001826826923076923,
+      "loss": 0.8069,
+      "step": 4573
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 0.00018265384615384612,
+      "loss": 0.8377,
+      "step": 4574
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 0.00018262499999999998,
+      "loss": 0.9644,
+      "step": 4575
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 0.00018259615384615382,
+      "loss": 0.8853,
+      "step": 4576
+    },
+    {
+      "epoch": 41.99,
+      "learning_rate": 0.00018256730769230768,
+      "loss": 0.8275,
+      "step": 4577
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 0.00018253846153846152,
+      "loss": 0.7861,
+      "step": 4578
+    },
+    {
+      "epoch": 42.01,
+      "learning_rate": 0.00018250961538461536,
+      "loss": 0.7224,
+      "step": 4579
+    },
+    {
+      "epoch": 42.02,
+      "learning_rate": 0.00018248076923076923,
+      "loss": 0.7237,
+      "step": 4580
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 0.00018245192307692304,
+      "loss": 0.8011,
+      "step": 4581
+    },
+    {
+      "epoch": 42.04,
+      "learning_rate": 0.0001824230769230769,
+      "loss": 0.7373,
+      "step": 4582
+    },
+    {
+      "epoch": 42.05,
+      "learning_rate": 0.00018239423076923077,
+      "loss": 0.8563,
+      "step": 4583
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 0.00018236538461538458,
+      "loss": 0.8381,
+      "step": 4584
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 0.00018233653846153844,
+      "loss": 0.6981,
+      "step": 4585
+    },
+    {
+      "epoch": 42.07,
+      "learning_rate": 0.00018230769230769228,
+      "loss": 0.7535,
+      "step": 4586
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 0.00018227884615384615,
+      "loss": 0.7331,
+      "step": 4587
+    },
+    {
+      "epoch": 42.09,
+      "learning_rate": 0.00018224999999999998,
+      "loss": 0.8502,
+      "step": 4588
+    },
+    {
+      "epoch": 42.1,
+      "learning_rate": 0.00018222115384615382,
+      "loss": 0.8222,
+      "step": 4589
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 0.0001821923076923077,
+      "loss": 0.8337,
+      "step": 4590
+    },
+    {
+      "epoch": 42.12,
+      "learning_rate": 0.0001821634615384615,
+      "loss": 0.7525,
+      "step": 4591
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 0.00018213461538461536,
+      "loss": 0.9714,
+      "step": 4592
+    },
+    {
+      "epoch": 42.14,
+      "learning_rate": 0.00018210576923076923,
+      "loss": 0.8065,
+      "step": 4593
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 0.00018207692307692304,
+      "loss": 0.7553,
+      "step": 4594
+    },
+    {
+      "epoch": 42.16,
+      "learning_rate": 0.0001820480769230769,
+      "loss": 0.7892,
+      "step": 4595
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 0.00018201923076923074,
+      "loss": 0.8764,
+      "step": 4596
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 0.0001819903846153846,
+      "loss": 0.9039,
+      "step": 4597
+    },
+    {
+      "epoch": 42.18,
+      "learning_rate": 0.00018196153846153845,
+      "loss": 0.7819,
+      "step": 4598
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 0.00018193269230769229,
+      "loss": 0.6985,
+      "step": 4599
+    },
+    {
+      "epoch": 42.2,
+      "learning_rate": 0.00018190384615384615,
+      "loss": 0.7712,
+      "step": 4600
+    },
+    {
+      "epoch": 42.21,
+      "learning_rate": 0.00018187499999999996,
+      "loss": 0.8229,
+      "step": 4601
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 0.00018184615384615383,
+      "loss": 0.7643,
+      "step": 4602
+    },
+    {
+      "epoch": 42.23,
+      "learning_rate": 0.0001818173076923077,
+      "loss": 0.8431,
+      "step": 4603
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 0.0001817884615384615,
+      "loss": 0.8915,
+      "step": 4604
+    },
+    {
+      "epoch": 42.25,
+      "learning_rate": 0.00018175961538461537,
+      "loss": 0.934,
+      "step": 4605
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 0.0001817307692307692,
+      "loss": 0.8522,
+      "step": 4606
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 0.00018170192307692307,
+      "loss": 0.8705,
+      "step": 4607
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 0.0001816730769230769,
+      "loss": 0.8464,
+      "step": 4608
+    },
+    {
+      "epoch": 42.28,
+      "learning_rate": 0.00018164423076923075,
+      "loss": 0.7064,
+      "step": 4609
+    },
+    {
+      "epoch": 42.29,
+      "learning_rate": 0.00018161538461538461,
+      "loss": 0.9327,
+      "step": 4610
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 0.00018158653846153842,
+      "loss": 0.8839,
+      "step": 4611
+    },
+    {
+      "epoch": 42.31,
+      "learning_rate": 0.0001815576923076923,
+      "loss": 0.8012,
+      "step": 4612
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 0.00018152884615384616,
+      "loss": 0.8348,
+      "step": 4613
+    },
+    {
+      "epoch": 42.33,
+      "learning_rate": 0.00018149999999999997,
+      "loss": 0.8164,
+      "step": 4614
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 0.00018147115384615383,
+      "loss": 0.9141,
+      "step": 4615
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 0.00018144230769230767,
+      "loss": 0.8649,
+      "step": 4616
+    },
+    {
+      "epoch": 42.36,
+      "learning_rate": 0.00018141346153846154,
+      "loss": 0.7373,
+      "step": 4617
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 0.00018138461538461537,
+      "loss": 0.776,
+      "step": 4618
+    },
+    {
+      "epoch": 42.38,
+      "learning_rate": 0.0001813557692307692,
+      "loss": 0.7695,
+      "step": 4619
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 0.00018132692307692308,
+      "loss": 0.7783,
+      "step": 4620
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 0.0001812980769230769,
+      "loss": 0.7247,
+      "step": 4621
+    },
+    {
+      "epoch": 42.4,
+      "learning_rate": 0.00018126923076923075,
+      "loss": 0.9918,
+      "step": 4622
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 0.00018124038461538462,
+      "loss": 0.8179,
+      "step": 4623
+    },
+    {
+      "epoch": 42.42,
+      "learning_rate": 0.00018121153846153843,
+      "loss": 0.9547,
+      "step": 4624
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 0.0001811826923076923,
+      "loss": 0.9191,
+      "step": 4625
+    },
+    {
+      "epoch": 42.44,
+      "learning_rate": 0.00018115384615384613,
+      "loss": 0.8746,
+      "step": 4626
+    },
+    {
+      "epoch": 42.45,
+      "learning_rate": 0.000181125,
+      "loss": 0.7969,
+      "step": 4627
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 0.00018109615384615384,
+      "loss": 0.763,
+      "step": 4628
+    },
+    {
+      "epoch": 42.47,
+      "learning_rate": 0.00018106730769230767,
+      "loss": 0.8129,
+      "step": 4629
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 0.00018103846153846154,
+      "loss": 0.7443,
+      "step": 4630
+    },
+    {
+      "epoch": 42.49,
+      "learning_rate": 0.00018100961538461535,
+      "loss": 0.8187,
+      "step": 4631
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 0.00018098076923076922,
+      "loss": 0.725,
+      "step": 4632
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 0.00018095192307692308,
+      "loss": 0.8903,
+      "step": 4633
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 0.0001809230769230769,
+      "loss": 0.7094,
+      "step": 4634
+    },
+    {
+      "epoch": 42.52,
+      "learning_rate": 0.00018089423076923076,
+      "loss": 0.7761,
+      "step": 4635
+    },
+    {
+      "epoch": 42.53,
+      "learning_rate": 0.0001808653846153846,
+      "loss": 0.8038,
+      "step": 4636
+    },
+    {
+      "epoch": 42.54,
+      "learning_rate": 0.00018083653846153846,
+      "loss": 0.8496,
+      "step": 4637
+    },
+    {
+      "epoch": 42.55,
+      "learning_rate": 0.00018080769230769227,
+      "loss": 0.9032,
+      "step": 4638
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 0.00018077884615384614,
+      "loss": 0.7176,
+      "step": 4639
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 0.00018075,
+      "loss": 0.9401,
+      "step": 4640
+    },
+    {
+      "epoch": 42.58,
+      "learning_rate": 0.0001807211538461538,
+      "loss": 0.9093,
+      "step": 4641
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 0.00018069230769230768,
+      "loss": 0.9284,
+      "step": 4642
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 0.00018066346153846152,
+      "loss": 0.856,
+      "step": 4643
+    },
+    {
+      "epoch": 42.61,
+      "learning_rate": 0.00018063461538461536,
+      "loss": 0.8019,
+      "step": 4644
+    },
+    {
+      "epoch": 42.61,
+      "learning_rate": 0.00018060576923076922,
+      "loss": 0.8052,
+      "step": 4645
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 0.00018057692307692306,
+      "loss": 0.8813,
+      "step": 4646
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 0.00018054807692307692,
+      "loss": 0.7857,
+      "step": 4647
+    },
+    {
+      "epoch": 42.64,
+      "learning_rate": 0.00018051923076923073,
+      "loss": 0.866,
+      "step": 4648
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 0.0001804903846153846,
+      "loss": 0.781,
+      "step": 4649
+    },
+    {
+      "epoch": 42.66,
+      "learning_rate": 0.00018046153846153847,
+      "loss": 0.8379,
+      "step": 4650
+    },
+    {
+      "epoch": 42.67,
+      "learning_rate": 0.00018043269230769228,
+      "loss": 0.9162,
+      "step": 4651
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 0.00018040384615384614,
+      "loss": 0.7784,
+      "step": 4652
+    },
+    {
+      "epoch": 42.69,
+      "learning_rate": 0.00018037499999999998,
+      "loss": 0.7722,
+      "step": 4653
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 0.00018034615384615382,
+      "loss": 0.8053,
+      "step": 4654
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 0.00018031730769230768,
+      "loss": 0.8317,
+      "step": 4655
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 0.00018028846153846152,
+      "loss": 0.9101,
+      "step": 4656
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 0.0001802596153846154,
+      "loss": 0.7411,
+      "step": 4657
+    },
+    {
+      "epoch": 42.73,
+      "learning_rate": 0.0001802307692307692,
+      "loss": 0.8304,
+      "step": 4658
+    },
+    {
+      "epoch": 42.74,
+      "learning_rate": 0.00018020192307692306,
+      "loss": 0.8188,
+      "step": 4659
+    },
+    {
+      "epoch": 42.75,
+      "learning_rate": 0.00018017307692307693,
+      "loss": 0.7645,
+      "step": 4660
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 0.00018014423076923074,
+      "loss": 0.7805,
+      "step": 4661
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 0.0001801153846153846,
+      "loss": 0.8291,
+      "step": 4662
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 0.00018008653846153844,
+      "loss": 0.9076,
+      "step": 4663
+    },
+    {
+      "epoch": 42.79,
+      "learning_rate": 0.00018005769230769228,
+      "loss": 0.8454,
+      "step": 4664
+    },
+    {
+      "epoch": 42.8,
+      "learning_rate": 0.00018002884615384615,
+      "loss": 0.8487,
+      "step": 4665
+    },
+    {
+      "epoch": 42.81,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.7729,
+      "step": 4666
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 0.00017997115384615385,
+      "loss": 0.8391,
+      "step": 4667
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 0.00017994230769230766,
+      "loss": 0.7933,
+      "step": 4668
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 0.00017991346153846153,
+      "loss": 0.8418,
+      "step": 4669
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 0.0001798846153846154,
+      "loss": 0.8062,
+      "step": 4670
+    },
+    {
+      "epoch": 42.85,
+      "learning_rate": 0.0001798557692307692,
+      "loss": 0.8633,
+      "step": 4671
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 0.00017982692307692307,
+      "loss": 0.8506,
+      "step": 4672
+    },
+    {
+      "epoch": 42.87,
+      "learning_rate": 0.0001797980769230769,
+      "loss": 0.7055,
+      "step": 4673
+    },
+    {
+      "epoch": 42.88,
+      "learning_rate": 0.00017976923076923074,
+      "loss": 0.8121,
+      "step": 4674
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 0.0001797403846153846,
+      "loss": 0.8033,
+      "step": 4675
+    },
+    {
+      "epoch": 42.9,
+      "learning_rate": 0.00017971153846153845,
+      "loss": 0.7677,
+      "step": 4676
+    },
+    {
+      "epoch": 42.91,
+      "learning_rate": 0.0001796826923076923,
+      "loss": 0.6929,
+      "step": 4677
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 0.00017965384615384612,
+      "loss": 0.8833,
+      "step": 4678
+    },
+    {
+      "epoch": 42.93,
+      "learning_rate": 0.000179625,
+      "loss": 0.7844,
+      "step": 4679
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 0.00017959615384615385,
+      "loss": 0.879,
+      "step": 4680
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 0.00017956730769230766,
+      "loss": 0.8542,
+      "step": 4681
+    },
+    {
+      "epoch": 42.95,
+      "learning_rate": 0.00017953846153846153,
+      "loss": 0.905,
+      "step": 4682
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 0.00017950961538461537,
+      "loss": 0.7594,
+      "step": 4683
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 0.0001794807692307692,
+      "loss": 0.986,
+      "step": 4684
+    },
+    {
+      "epoch": 42.98,
+      "learning_rate": 0.00017945192307692307,
+      "loss": 0.806,
+      "step": 4685
+    },
+    {
+      "epoch": 42.99,
+      "learning_rate": 0.0001794230769230769,
+      "loss": 0.9354,
+      "step": 4686
+    },
+    {
+      "epoch": 43.0,
+      "learning_rate": 0.00017939423076923078,
+      "loss": 0.6766,
+      "step": 4687
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 0.00017936538461538459,
+      "loss": 0.8427,
+      "step": 4688
+    },
+    {
+      "epoch": 43.02,
+      "learning_rate": 0.00017933653846153845,
+      "loss": 0.8807,
+      "step": 4689
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 0.00017930769230769232,
+      "loss": 0.8423,
+      "step": 4690
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 0.00017927884615384613,
+      "loss": 0.7409,
+      "step": 4691
+    },
+    {
+      "epoch": 43.05,
+      "learning_rate": 0.00017925,
+      "loss": 0.8163,
+      "step": 4692
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 0.00017922115384615383,
+      "loss": 0.8497,
+      "step": 4693
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 0.00017919230769230767,
+      "loss": 0.9028,
+      "step": 4694
+    },
+    {
+      "epoch": 43.07,
+      "learning_rate": 0.00017916346153846153,
+      "loss": 0.7855,
+      "step": 4695
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 0.00017913461538461537,
+      "loss": 0.6573,
+      "step": 4696
+    },
+    {
+      "epoch": 43.09,
+      "learning_rate": 0.00017910576923076924,
+      "loss": 0.7488,
+      "step": 4697
+    },
+    {
+      "epoch": 43.1,
+      "learning_rate": 0.00017907692307692305,
+      "loss": 0.8182,
+      "step": 4698
+    },
+    {
+      "epoch": 43.11,
+      "learning_rate": 0.00017904807692307691,
+      "loss": 0.7115,
+      "step": 4699
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 0.00017901923076923073,
+      "loss": 0.8028,
+      "step": 4700
+    },
+    {
+      "epoch": 43.13,
+      "learning_rate": 0.0001789903846153846,
+      "loss": 0.7261,
+      "step": 4701
+    },
+    {
+      "epoch": 43.14,
+      "learning_rate": 0.00017896153846153846,
+      "loss": 0.735,
+      "step": 4702
+    },
+    {
+      "epoch": 43.15,
+      "learning_rate": 0.0001789326923076923,
+      "loss": 0.8271,
+      "step": 4703
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 0.00017890384615384613,
+      "loss": 0.8585,
+      "step": 4704
+    },
+    {
+      "epoch": 43.17,
+      "learning_rate": 0.00017887499999999997,
+      "loss": 0.8685,
+      "step": 4705
+    },
+    {
+      "epoch": 43.17,
+      "learning_rate": 0.00017884615384615384,
+      "loss": 0.6432,
+      "step": 4706
+    },
+    {
+      "epoch": 43.18,
+      "learning_rate": 0.0001788173076923077,
+      "loss": 0.7613,
+      "step": 4707
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 0.0001787884615384615,
+      "loss": 0.7892,
+      "step": 4708
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 0.00017875961538461538,
+      "loss": 0.9276,
+      "step": 4709
+    },
+    {
+      "epoch": 43.21,
+      "learning_rate": 0.0001787307692307692,
+      "loss": 0.9813,
+      "step": 4710
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 0.00017870192307692305,
+      "loss": 0.7829,
+      "step": 4711
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 0.00017867307692307692,
+      "loss": 0.8082,
+      "step": 4712
+    },
+    {
+      "epoch": 43.24,
+      "learning_rate": 0.00017864423076923076,
+      "loss": 0.888,
+      "step": 4713
+    },
+    {
+      "epoch": 43.25,
+      "learning_rate": 0.0001786153846153846,
+      "loss": 0.7916,
+      "step": 4714
+    },
+    {
+      "epoch": 43.26,
+      "learning_rate": 0.00017858653846153843,
+      "loss": 0.7499,
+      "step": 4715
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 0.0001785576923076923,
+      "loss": 0.8305,
+      "step": 4716
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 0.00017852884615384614,
+      "loss": 0.7635,
+      "step": 4717
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 0.00017849999999999997,
+      "loss": 0.6897,
+      "step": 4718
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 0.00017847115384615384,
+      "loss": 0.8111,
+      "step": 4719
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 0.00017844230769230765,
+      "loss": 0.8185,
+      "step": 4720
+    },
+    {
+      "epoch": 43.31,
+      "learning_rate": 0.00017841346153846152,
+      "loss": 0.8731,
+      "step": 4721
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 0.00017838461538461538,
+      "loss": 0.6431,
+      "step": 4722
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 0.00017835576923076922,
+      "loss": 0.7763,
+      "step": 4723
+    },
+    {
+      "epoch": 43.34,
+      "learning_rate": 0.00017832692307692306,
+      "loss": 0.8852,
+      "step": 4724
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 0.0001782980769230769,
+      "loss": 0.983,
+      "step": 4725
+    },
+    {
+      "epoch": 43.36,
+      "learning_rate": 0.00017826923076923076,
+      "loss": 0.7742,
+      "step": 4726
+    },
+    {
+      "epoch": 43.37,
+      "learning_rate": 0.0001782403846153846,
+      "loss": 0.7902,
+      "step": 4727
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 0.00017821153846153844,
+      "loss": 0.9161,
+      "step": 4728
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 0.0001781826923076923,
+      "loss": 0.7697,
+      "step": 4729
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 0.00017815384615384611,
+      "loss": 0.8807,
+      "step": 4730
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 0.00017812499999999998,
+      "loss": 0.9045,
+      "step": 4731
+    },
+    {
+      "epoch": 43.41,
+      "learning_rate": 0.00017809615384615384,
+      "loss": 0.8012,
+      "step": 4732
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 0.00017806730769230768,
+      "loss": 0.7483,
+      "step": 4733
+    },
+    {
+      "epoch": 43.43,
+      "learning_rate": 0.00017803846153846152,
+      "loss": 0.7351,
+      "step": 4734
+    },
+    {
+      "epoch": 43.44,
+      "learning_rate": 0.00017800961538461536,
+      "loss": 0.7864,
+      "step": 4735
+    },
+    {
+      "epoch": 43.45,
+      "learning_rate": 0.00017798076923076922,
+      "loss": 0.841,
+      "step": 4736
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 0.00017795192307692306,
+      "loss": 0.707,
+      "step": 4737
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 0.0001779230769230769,
+      "loss": 0.8411,
+      "step": 4738
+    },
+    {
+      "epoch": 43.48,
+      "learning_rate": 0.00017789423076923077,
+      "loss": 0.8754,
+      "step": 4739
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 0.00017786538461538458,
+      "loss": 0.6725,
+      "step": 4740
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 0.00017783653846153844,
+      "loss": 0.595,
+      "step": 4741
+    },
+    {
+      "epoch": 43.5,
+      "learning_rate": 0.0001778076923076923,
+      "loss": 0.9755,
+      "step": 4742
+    },
+    {
+      "epoch": 43.51,
+      "learning_rate": 0.00017777884615384615,
+      "loss": 0.9068,
+      "step": 4743
+    },
+    {
+      "epoch": 43.52,
+      "learning_rate": 0.00017774999999999998,
+      "loss": 0.7003,
+      "step": 4744
+    },
+    {
+      "epoch": 43.53,
+      "learning_rate": 0.00017772115384615382,
+      "loss": 0.7294,
+      "step": 4745
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 0.0001776923076923077,
+      "loss": 0.9443,
+      "step": 4746
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 0.00017766346153846153,
+      "loss": 0.8259,
+      "step": 4747
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 0.00017763461538461536,
+      "loss": 0.7832,
+      "step": 4748
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 0.00017760576923076923,
+      "loss": 0.868,
+      "step": 4749
+    },
+    {
+      "epoch": 43.58,
+      "learning_rate": 0.00017757692307692304,
+      "loss": 0.7367,
+      "step": 4750
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 0.0001775480769230769,
+      "loss": 0.6537,
+      "step": 4751
+    },
+    {
+      "epoch": 43.6,
+      "learning_rate": 0.00017751923076923077,
+      "loss": 0.6894,
+      "step": 4752
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 0.0001774903846153846,
+      "loss": 0.8075,
+      "step": 4753
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 0.00017746153846153845,
+      "loss": 0.8759,
+      "step": 4754
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 0.00017743269230769228,
+      "loss": 0.9058,
+      "step": 4755
+    },
+    {
+      "epoch": 43.63,
+      "learning_rate": 0.00017740384615384615,
+      "loss": 0.8136,
+      "step": 4756
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 0.000177375,
+      "loss": 0.7608,
+      "step": 4757
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 0.00017734615384615383,
+      "loss": 0.7496,
+      "step": 4758
+    },
+    {
+      "epoch": 43.66,
+      "learning_rate": 0.0001773173076923077,
+      "loss": 0.7801,
+      "step": 4759
+    },
+    {
+      "epoch": 43.67,
+      "learning_rate": 0.0001772884615384615,
+      "loss": 0.7323,
+      "step": 4760
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 0.00017725961538461537,
+      "loss": 0.8426,
+      "step": 4761
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 0.00017723076923076923,
+      "loss": 0.7239,
+      "step": 4762
+    },
+    {
+      "epoch": 43.7,
+      "learning_rate": 0.00017720192307692307,
+      "loss": 0.8008,
+      "step": 4763
+    },
+    {
+      "epoch": 43.71,
+      "learning_rate": 0.0001771730769230769,
+      "loss": 0.816,
+      "step": 4764
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 0.00017714423076923075,
+      "loss": 0.7657,
+      "step": 4765
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 0.0001771153846153846,
+      "loss": 0.7239,
+      "step": 4766
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 0.00017708653846153842,
+      "loss": 0.9287,
+      "step": 4767
+    },
+    {
+      "epoch": 43.74,
+      "learning_rate": 0.0001770576923076923,
+      "loss": 0.7943,
+      "step": 4768
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 0.00017702884615384615,
+      "loss": 0.7847,
+      "step": 4769
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 0.8961,
+      "step": 4770
+    },
+    {
+      "epoch": 43.77,
+      "learning_rate": 0.00017697115384615383,
+      "loss": 0.7147,
+      "step": 4771
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 0.00017694230769230767,
+      "loss": 0.771,
+      "step": 4772
+    },
+    {
+      "epoch": 43.79,
+      "learning_rate": 0.0001769134615384615,
+      "loss": 0.8054,
+      "step": 4773
+    },
+    {
+      "epoch": 43.8,
+      "learning_rate": 0.00017688461538461537,
+      "loss": 0.833,
+      "step": 4774
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 0.0001768557692307692,
+      "loss": 0.7193,
+      "step": 4775
+    },
+    {
+      "epoch": 43.82,
+      "learning_rate": 0.00017682692307692308,
+      "loss": 0.7028,
+      "step": 4776
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 0.0001767980769230769,
+      "loss": 0.7601,
+      "step": 4777
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 0.00017676923076923075,
+      "loss": 0.9021,
+      "step": 4778
+    },
+    {
+      "epoch": 43.84,
+      "learning_rate": 0.00017674038461538462,
+      "loss": 0.8928,
+      "step": 4779
+    },
+    {
+      "epoch": 43.85,
+      "learning_rate": 0.00017671153846153843,
+      "loss": 0.9538,
+      "step": 4780
+    },
+    {
+      "epoch": 43.86,
+      "learning_rate": 0.0001766826923076923,
+      "loss": 0.8579,
+      "step": 4781
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 0.00017665384615384613,
+      "loss": 0.8326,
+      "step": 4782
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 0.00017662499999999997,
+      "loss": 0.7949,
+      "step": 4783
+    },
+    {
+      "epoch": 43.89,
+      "learning_rate": 0.00017659615384615384,
+      "loss": 0.7513,
+      "step": 4784
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 0.00017656730769230767,
+      "loss": 0.8251,
+      "step": 4785
+    },
+    {
+      "epoch": 43.91,
+      "learning_rate": 0.00017653846153846154,
+      "loss": 0.7916,
+      "step": 4786
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 0.00017650961538461535,
+      "loss": 0.6471,
+      "step": 4787
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 0.00017648076923076921,
+      "loss": 0.8904,
+      "step": 4788
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 0.00017645192307692308,
+      "loss": 0.719,
+      "step": 4789
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 0.0001764230769230769,
+      "loss": 0.8358,
+      "step": 4790
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 0.00017639423076923076,
+      "loss": 0.9273,
+      "step": 4791
+    },
+    {
+      "epoch": 43.96,
+      "learning_rate": 0.0001763653846153846,
+      "loss": 0.8671,
+      "step": 4792
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 0.00017633653846153843,
+      "loss": 1.0598,
+      "step": 4793
+    },
+    {
+      "epoch": 43.98,
+      "learning_rate": 0.0001763076923076923,
+      "loss": 0.8626,
+      "step": 4794
+    },
+    {
+      "epoch": 43.99,
+      "learning_rate": 0.00017627884615384614,
+      "loss": 0.93,
+      "step": 4795
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 0.00017625,
+      "loss": 0.7073,
+      "step": 4796
+    },
+    {
+      "epoch": 44.01,
+      "learning_rate": 0.0001762211538461538,
+      "loss": 0.781,
+      "step": 4797
+    },
+    {
+      "epoch": 44.02,
+      "learning_rate": 0.00017619230769230768,
+      "loss": 0.832,
+      "step": 4798
+    },
+    {
+      "epoch": 44.03,
+      "learning_rate": 0.00017616346153846154,
+      "loss": 0.7284,
+      "step": 4799
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 0.00017613461538461535,
+      "loss": 0.7659,
+      "step": 4800
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 0.00017610576923076922,
+      "loss": 0.8274,
+      "step": 4801
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 0.00017607692307692306,
+      "loss": 0.6419,
+      "step": 4802
+    },
+    {
+      "epoch": 44.06,
+      "learning_rate": 0.0001760480769230769,
+      "loss": 0.8087,
+      "step": 4803
+    },
+    {
+      "epoch": 44.07,
+      "learning_rate": 0.00017601923076923076,
+      "loss": 0.8034,
+      "step": 4804
+    },
+    {
+      "epoch": 44.08,
+      "learning_rate": 0.0001759903846153846,
+      "loss": 0.8063,
+      "step": 4805
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 0.00017596153846153846,
+      "loss": 0.8505,
+      "step": 4806
+    },
+    {
+      "epoch": 44.1,
+      "learning_rate": 0.00017593269230769228,
+      "loss": 0.7837,
+      "step": 4807
+    },
+    {
+      "epoch": 44.11,
+      "learning_rate": 0.00017590384615384614,
+      "loss": 0.7478,
+      "step": 4808
+    },
+    {
+      "epoch": 44.12,
+      "learning_rate": 0.000175875,
+      "loss": 0.9508,
+      "step": 4809
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 0.00017584615384615382,
+      "loss": 0.7783,
+      "step": 4810
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 0.00017581730769230768,
+      "loss": 0.7277,
+      "step": 4811
+    },
+    {
+      "epoch": 44.15,
+      "learning_rate": 0.00017578846153846152,
+      "loss": 0.7561,
+      "step": 4812
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 0.00017575961538461536,
+      "loss": 0.7999,
+      "step": 4813
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 0.00017573076923076922,
+      "loss": 0.7703,
+      "step": 4814
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 0.00017570192307692306,
+      "loss": 0.8501,
+      "step": 4815
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 0.00017567307692307693,
+      "loss": 0.8412,
+      "step": 4816
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 0.00017564423076923074,
+      "loss": 0.8016,
+      "step": 4817
+    },
+    {
+      "epoch": 44.2,
+      "learning_rate": 0.0001756153846153846,
+      "loss": 0.8994,
+      "step": 4818
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 0.00017558653846153847,
+      "loss": 0.8324,
+      "step": 4819
+    },
+    {
+      "epoch": 44.22,
+      "learning_rate": 0.00017555769230769228,
+      "loss": 0.764,
+      "step": 4820
+    },
+    {
+      "epoch": 44.23,
+      "learning_rate": 0.00017552884615384615,
+      "loss": 0.6828,
+      "step": 4821
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 0.00017549999999999998,
+      "loss": 0.7965,
+      "step": 4822
+    },
+    {
+      "epoch": 44.25,
+      "learning_rate": 0.00017547115384615382,
+      "loss": 0.7505,
+      "step": 4823
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 0.0001754423076923077,
+      "loss": 0.8103,
+      "step": 4824
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 0.00017541346153846152,
+      "loss": 0.8114,
+      "step": 4825
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 0.0001753846153846154,
+      "loss": 0.7987,
+      "step": 4826
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 0.0001753557692307692,
+      "loss": 0.6793,
+      "step": 4827
+    },
+    {
+      "epoch": 44.29,
+      "learning_rate": 0.00017532692307692307,
+      "loss": 0.7716,
+      "step": 4828
+    },
+    {
+      "epoch": 44.3,
+      "learning_rate": 0.0001752980769230769,
+      "loss": 0.7271,
+      "step": 4829
+    },
+    {
+      "epoch": 44.31,
+      "learning_rate": 0.00017526923076923074,
+      "loss": 0.8139,
+      "step": 4830
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 0.0001752403846153846,
+      "loss": 0.8265,
+      "step": 4831
+    },
+    {
+      "epoch": 44.33,
+      "learning_rate": 0.00017521153846153845,
+      "loss": 0.7879,
+      "step": 4832
+    },
+    {
+      "epoch": 44.34,
+      "learning_rate": 0.00017518269230769228,
+      "loss": 0.8469,
+      "step": 4833
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 0.00017515384615384612,
+      "loss": 0.7929,
+      "step": 4834
+    },
+    {
+      "epoch": 44.36,
+      "learning_rate": 0.000175125,
+      "loss": 0.7842,
+      "step": 4835
+    },
+    {
+      "epoch": 44.37,
+      "learning_rate": 0.00017509615384615385,
+      "loss": 0.8105,
+      "step": 4836
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 0.00017506730769230766,
+      "loss": 0.7904,
+      "step": 4837
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 0.00017503846153846153,
+      "loss": 0.8045,
+      "step": 4838
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 0.00017500961538461534,
+      "loss": 0.8628,
+      "step": 4839
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 0.0001749807692307692,
+      "loss": 0.9635,
+      "step": 4840
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 0.00017495192307692307,
+      "loss": 0.868,
+      "step": 4841
+    },
+    {
+      "epoch": 44.42,
+      "learning_rate": 0.0001749230769230769,
+      "loss": 0.6908,
+      "step": 4842
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 0.00017489423076923075,
+      "loss": 0.7525,
+      "step": 4843
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 0.00017486538461538459,
+      "loss": 0.7882,
+      "step": 4844
+    },
+    {
+      "epoch": 44.45,
+      "learning_rate": 0.00017483653846153845,
+      "loss": 0.7337,
+      "step": 4845
+    },
+    {
+      "epoch": 44.46,
+      "learning_rate": 0.00017480769230769232,
+      "loss": 0.7172,
+      "step": 4846
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 0.00017477884615384613,
+      "loss": 0.8419,
+      "step": 4847
+    },
+    {
+      "epoch": 44.48,
+      "learning_rate": 0.00017475,
+      "loss": 0.7554,
+      "step": 4848
+    },
+    {
+      "epoch": 44.49,
+      "learning_rate": 0.0001747211538461538,
+      "loss": 0.6914,
+      "step": 4849
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 0.00017469230769230767,
+      "loss": 0.8753,
+      "step": 4850
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 0.00017466346153846153,
+      "loss": 0.75,
+      "step": 4851
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 0.00017463461538461537,
+      "loss": 0.8596,
+      "step": 4852
+    },
+    {
+      "epoch": 44.52,
+      "learning_rate": 0.0001746057692307692,
+      "loss": 0.8732,
+      "step": 4853
+    },
+    {
+      "epoch": 44.53,
+      "learning_rate": 0.00017457692307692305,
+      "loss": 0.8751,
+      "step": 4854
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 0.0001745480769230769,
+      "loss": 0.9472,
+      "step": 4855
+    },
+    {
+      "epoch": 44.55,
+      "learning_rate": 0.00017451923076923078,
+      "loss": 0.8726,
+      "step": 4856
+    },
+    {
+      "epoch": 44.56,
+      "learning_rate": 0.0001744903846153846,
+      "loss": 0.9112,
+      "step": 4857
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 0.00017446153846153846,
+      "loss": 0.8017,
+      "step": 4858
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 0.00017443269230769227,
+      "loss": 0.9206,
+      "step": 4859
+    },
+    {
+      "epoch": 44.59,
+      "learning_rate": 0.00017440384615384613,
+      "loss": 0.8308,
+      "step": 4860
+    },
+    {
+      "epoch": 44.6,
+      "learning_rate": 0.000174375,
+      "loss": 0.8261,
+      "step": 4861
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 0.00017434615384615383,
+      "loss": 0.8042,
+      "step": 4862
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 0.00017431730769230767,
+      "loss": 0.7932,
+      "step": 4863
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 0.0001742884615384615,
+      "loss": 0.7211,
+      "step": 4864
+    },
+    {
+      "epoch": 44.63,
+      "learning_rate": 0.00017425961538461538,
+      "loss": 0.805,
+      "step": 4865
+    },
+    {
+      "epoch": 44.64,
+      "learning_rate": 0.00017423076923076924,
+      "loss": 0.8776,
+      "step": 4866
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 0.00017420192307692305,
+      "loss": 0.8468,
+      "step": 4867
+    },
+    {
+      "epoch": 44.66,
+      "learning_rate": 0.00017417307692307692,
+      "loss": 0.759,
+      "step": 4868
+    },
+    {
+      "epoch": 44.67,
+      "learning_rate": 0.00017414423076923073,
+      "loss": 0.6937,
+      "step": 4869
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 0.0001741153846153846,
+      "loss": 0.7819,
+      "step": 4870
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 0.00017408653846153846,
+      "loss": 0.9208,
+      "step": 4871
+    },
+    {
+      "epoch": 44.7,
+      "learning_rate": 0.0001740576923076923,
+      "loss": 0.8843,
+      "step": 4872
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 0.00017402884615384614,
+      "loss": 0.8681,
+      "step": 4873
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 0.8381,
+      "step": 4874
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 0.00017397115384615384,
+      "loss": 0.763,
+      "step": 4875
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 0.0001739423076923077,
+      "loss": 0.8943,
+      "step": 4876
+    },
+    {
+      "epoch": 44.74,
+      "learning_rate": 0.00017391346153846152,
+      "loss": 0.7759,
+      "step": 4877
+    },
+    {
+      "epoch": 44.75,
+      "learning_rate": 0.00017388461538461538,
+      "loss": 0.952,
+      "step": 4878
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 0.0001738557692307692,
+      "loss": 0.8423,
+      "step": 4879
+    },
+    {
+      "epoch": 44.77,
+      "learning_rate": 0.00017382692307692306,
+      "loss": 0.6978,
+      "step": 4880
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 0.00017379807692307692,
+      "loss": 0.8438,
+      "step": 4881
+    },
+    {
+      "epoch": 44.79,
+      "learning_rate": 0.00017376923076923076,
+      "loss": 0.844,
+      "step": 4882
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 0.0001737403846153846,
+      "loss": 0.7166,
+      "step": 4883
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 0.00017371153846153844,
+      "loss": 0.812,
+      "step": 4884
+    },
+    {
+      "epoch": 44.82,
+      "learning_rate": 0.0001736826923076923,
+      "loss": 0.824,
+      "step": 4885
+    },
+    {
+      "epoch": 44.83,
+      "learning_rate": 0.00017365384615384617,
+      "loss": 0.6609,
+      "step": 4886
+    },
+    {
+      "epoch": 44.83,
+      "learning_rate": 0.00017362499999999998,
+      "loss": 0.8178,
+      "step": 4887
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 0.00017359615384615384,
+      "loss": 0.7726,
+      "step": 4888
+    },
+    {
+      "epoch": 44.85,
+      "learning_rate": 0.00017356730769230765,
+      "loss": 0.8066,
+      "step": 4889
+    },
+    {
+      "epoch": 44.86,
+      "learning_rate": 0.00017353846153846152,
+      "loss": 0.7658,
+      "step": 4890
+    },
+    {
+      "epoch": 44.87,
+      "learning_rate": 0.00017350961538461539,
+      "loss": 0.8384,
+      "step": 4891
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 0.00017348076923076922,
+      "loss": 0.7787,
+      "step": 4892
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 0.00017345192307692306,
+      "loss": 0.7692,
+      "step": 4893
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 0.0001734230769230769,
+      "loss": 0.83,
+      "step": 4894
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 0.00017339423076923076,
+      "loss": 0.8181,
+      "step": 4895
+    },
+    {
+      "epoch": 44.92,
+      "learning_rate": 0.00017336538461538458,
+      "loss": 0.8287,
+      "step": 4896
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 0.00017333653846153844,
+      "loss": 0.7433,
+      "step": 4897
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 0.0001733076923076923,
+      "loss": 0.8827,
+      "step": 4898
+    },
+    {
+      "epoch": 44.94,
+      "learning_rate": 0.00017327884615384612,
+      "loss": 0.7738,
+      "step": 4899
+    },
+    {
+      "epoch": 44.95,
+      "learning_rate": 0.00017324999999999998,
+      "loss": 0.7295,
+      "step": 4900
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 0.00017322115384615382,
+      "loss": 0.8134,
+      "step": 4901
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 0.00017319230769230769,
+      "loss": 0.7302,
+      "step": 4902
+    },
+    {
+      "epoch": 44.98,
+      "learning_rate": 0.00017316346153846152,
+      "loss": 0.8382,
+      "step": 4903
+    },
+    {
+      "epoch": 44.99,
+      "learning_rate": 0.00017313461538461536,
+      "loss": 0.8477,
+      "step": 4904
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 0.00017310576923076923,
+      "loss": 0.7077,
+      "step": 4905
+    },
+    {
+      "epoch": 45.01,
+      "learning_rate": 0.00017307692307692304,
+      "loss": 0.7049,
+      "step": 4906
+    },
+    {
+      "epoch": 45.02,
+      "learning_rate": 0.0001730480769230769,
+      "loss": 0.7413,
+      "step": 4907
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 0.00017301923076923077,
+      "loss": 0.7877,
+      "step": 4908
+    },
+    {
+      "epoch": 45.04,
+      "learning_rate": 0.00017299038461538458,
+      "loss": 0.7808,
+      "step": 4909
+    },
+    {
+      "epoch": 45.05,
+      "learning_rate": 0.00017296153846153845,
+      "loss": 0.7569,
+      "step": 4910
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 0.00017293269230769228,
+      "loss": 0.6505,
+      "step": 4911
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 0.00017290384615384615,
+      "loss": 0.794,
+      "step": 4912
+    },
+    {
+      "epoch": 45.07,
+      "learning_rate": 0.000172875,
+      "loss": 0.8083,
+      "step": 4913
+    },
+    {
+      "epoch": 45.08,
+      "learning_rate": 0.00017284615384615383,
+      "loss": 0.7519,
+      "step": 4914
+    },
+    {
+      "epoch": 45.09,
+      "learning_rate": 0.0001728173076923077,
+      "loss": 0.7604,
+      "step": 4915
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 0.0001727884615384615,
+      "loss": 0.6424,
+      "step": 4916
+    },
+    {
+      "epoch": 45.11,
+      "learning_rate": 0.00017275961538461537,
+      "loss": 0.6599,
+      "step": 4917
+    },
+    {
+      "epoch": 45.12,
+      "learning_rate": 0.00017273076923076923,
+      "loss": 0.8002,
+      "step": 4918
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 0.00017270192307692304,
+      "loss": 0.8975,
+      "step": 4919
+    },
+    {
+      "epoch": 45.14,
+      "learning_rate": 0.0001726730769230769,
+      "loss": 0.7993,
+      "step": 4920
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 0.00017264423076923075,
+      "loss": 0.8062,
+      "step": 4921
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 0.0001726153846153846,
+      "loss": 0.8326,
+      "step": 4922
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 0.00017258653846153845,
+      "loss": 0.903,
+      "step": 4923
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 0.0001725576923076923,
+      "loss": 0.7551,
+      "step": 4924
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 0.00017252884615384615,
+      "loss": 0.7968,
+      "step": 4925
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 0.00017249999999999996,
+      "loss": 0.7367,
+      "step": 4926
+    },
+    {
+      "epoch": 45.2,
+      "learning_rate": 0.00017247115384615383,
+      "loss": 0.7233,
+      "step": 4927
+    },
+    {
+      "epoch": 45.21,
+      "learning_rate": 0.0001724423076923077,
+      "loss": 0.7627,
+      "step": 4928
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 0.0001724134615384615,
+      "loss": 0.819,
+      "step": 4929
+    },
+    {
+      "epoch": 45.23,
+      "learning_rate": 0.00017238461538461537,
+      "loss": 0.8211,
+      "step": 4930
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 0.0001723557692307692,
+      "loss": 0.7335,
+      "step": 4931
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 0.00017232692307692307,
+      "loss": 0.7447,
+      "step": 4932
+    },
+    {
+      "epoch": 45.26,
+      "learning_rate": 0.0001722980769230769,
+      "loss": 0.8685,
+      "step": 4933
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 0.00017226923076923075,
+      "loss": 0.7324,
+      "step": 4934
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 0.00017224038461538462,
+      "loss": 0.8479,
+      "step": 4935
+    },
+    {
+      "epoch": 45.28,
+      "learning_rate": 0.00017221153846153843,
+      "loss": 0.7106,
+      "step": 4936
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 0.0001721826923076923,
+      "loss": 0.7919,
+      "step": 4937
+    },
+    {
+      "epoch": 45.3,
+      "learning_rate": 0.00017215384615384616,
+      "loss": 0.8242,
+      "step": 4938
+    },
+    {
+      "epoch": 45.31,
+      "learning_rate": 0.00017212499999999997,
+      "loss": 0.7197,
+      "step": 4939
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 0.00017209615384615383,
+      "loss": 0.8466,
+      "step": 4940
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 0.00017206730769230767,
+      "loss": 0.7396,
+      "step": 4941
+    },
+    {
+      "epoch": 45.34,
+      "learning_rate": 0.00017203846153846154,
+      "loss": 0.8013,
+      "step": 4942
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 0.00017200961538461538,
+      "loss": 0.7542,
+      "step": 4943
+    },
+    {
+      "epoch": 45.36,
+      "learning_rate": 0.00017198076923076921,
+      "loss": 0.8173,
+      "step": 4944
+    },
+    {
+      "epoch": 45.37,
+      "learning_rate": 0.00017195192307692308,
+      "loss": 0.8048,
+      "step": 4945
+    },
+    {
+      "epoch": 45.38,
+      "learning_rate": 0.0001719230769230769,
+      "loss": 0.7789,
+      "step": 4946
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 0.00017189423076923076,
+      "loss": 0.8069,
+      "step": 4947
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 0.00017186538461538462,
+      "loss": 0.8411,
+      "step": 4948
+    },
+    {
+      "epoch": 45.4,
+      "learning_rate": 0.00017183653846153843,
+      "loss": 0.8539,
+      "step": 4949
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 0.0001718076923076923,
+      "loss": 0.6917,
+      "step": 4950
+    },
+    {
+      "epoch": 45.42,
+      "learning_rate": 0.00017177884615384614,
+      "loss": 0.6932,
+      "step": 4951
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 0.00017175,
+      "loss": 0.7565,
+      "step": 4952
+    },
+    {
+      "epoch": 45.44,
+      "learning_rate": 0.00017172115384615384,
+      "loss": 0.7962,
+      "step": 4953
+    },
+    {
+      "epoch": 45.45,
+      "learning_rate": 0.00017169230769230768,
+      "loss": 0.6932,
+      "step": 4954
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 0.00017166346153846154,
+      "loss": 0.6844,
+      "step": 4955
+    },
+    {
+      "epoch": 45.47,
+      "learning_rate": 0.00017163461538461535,
+      "loss": 0.7413,
+      "step": 4956
+    },
+    {
+      "epoch": 45.48,
+      "learning_rate": 0.00017160576923076922,
+      "loss": 0.7712,
+      "step": 4957
+    },
+    {
+      "epoch": 45.49,
+      "learning_rate": 0.00017157692307692308,
+      "loss": 0.7675,
+      "step": 4958
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 0.0001715480769230769,
+      "loss": 0.8094,
+      "step": 4959
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 0.00017151923076923076,
+      "loss": 0.8021,
+      "step": 4960
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 0.0001714903846153846,
+      "loss": 0.7357,
+      "step": 4961
+    },
+    {
+      "epoch": 45.52,
+      "learning_rate": 0.00017146153846153846,
+      "loss": 0.7973,
+      "step": 4962
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 0.00017143269230769227,
+      "loss": 0.8093,
+      "step": 4963
+    },
+    {
+      "epoch": 45.54,
+      "learning_rate": 0.00017140384615384614,
+      "loss": 0.8248,
+      "step": 4964
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 0.000171375,
+      "loss": 0.8024,
+      "step": 4965
+    },
+    {
+      "epoch": 45.56,
+      "learning_rate": 0.00017134615384615382,
+      "loss": 0.6688,
+      "step": 4966
+    },
+    {
+      "epoch": 45.57,
+      "learning_rate": 0.00017131730769230768,
+      "loss": 0.8866,
+      "step": 4967
+    },
+    {
+      "epoch": 45.58,
+      "learning_rate": 0.00017128846153846152,
+      "loss": 0.844,
+      "step": 4968
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 0.00017125961538461536,
+      "loss": 0.8431,
+      "step": 4969
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 0.00017123076923076922,
+      "loss": 0.7266,
+      "step": 4970
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 0.00017120192307692306,
+      "loss": 0.9167,
+      "step": 4971
+    },
+    {
+      "epoch": 45.61,
+      "learning_rate": 0.00017117307692307693,
+      "loss": 0.7,
+      "step": 4972
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 0.00017114423076923074,
+      "loss": 0.756,
+      "step": 4973
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 0.0001711153846153846,
+      "loss": 0.797,
+      "step": 4974
+    },
+    {
+      "epoch": 45.64,
+      "learning_rate": 0.00017108653846153847,
+      "loss": 0.757,
+      "step": 4975
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 0.00017105769230769228,
+      "loss": 0.591,
+      "step": 4976
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 0.00017102884615384614,
+      "loss": 0.6729,
+      "step": 4977
+    },
+    {
+      "epoch": 45.67,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 0.6787,
+      "step": 4978
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 0.00017097115384615382,
+      "loss": 0.7875,
+      "step": 4979
+    },
+    {
+      "epoch": 45.69,
+      "learning_rate": 0.00017094230769230769,
+      "loss": 0.7384,
+      "step": 4980
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 0.00017091346153846152,
+      "loss": 0.7155,
+      "step": 4981
+    },
+    {
+      "epoch": 45.71,
+      "learning_rate": 0.00017088461538461536,
+      "loss": 0.6314,
+      "step": 4982
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 0.0001708557692307692,
+      "loss": 0.6926,
+      "step": 4983
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 0.00017082692307692307,
+      "loss": 0.931,
+      "step": 4984
+    },
+    {
+      "epoch": 45.73,
+      "learning_rate": 0.00017079807692307693,
+      "loss": 0.9104,
+      "step": 4985
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 0.00017076923076923074,
+      "loss": 0.5463,
+      "step": 4986
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 0.0001707403846153846,
+      "loss": 0.7641,
+      "step": 4987
+    },
+    {
+      "epoch": 45.76,
+      "learning_rate": 0.00017071153846153845,
+      "loss": 0.7954,
+      "step": 4988
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 0.00017068269230769228,
+      "loss": 0.8026,
+      "step": 4989
+    },
+    {
+      "epoch": 45.78,
+      "learning_rate": 0.00017065384615384615,
+      "loss": 0.7825,
+      "step": 4990
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 0.000170625,
+      "loss": 0.694,
+      "step": 4991
+    },
+    {
+      "epoch": 45.8,
+      "learning_rate": 0.00017059615384615382,
+      "loss": 0.7334,
+      "step": 4992
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 0.00017056730769230766,
+      "loss": 0.779,
+      "step": 4993
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 0.00017053846153846153,
+      "loss": 0.7374,
+      "step": 4994
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 0.0001705096153846154,
+      "loss": 0.8743,
+      "step": 4995
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 0.0001704807692307692,
+      "loss": 0.7998,
+      "step": 4996
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 0.00017045192307692307,
+      "loss": 0.7843,
+      "step": 4997
+    },
+    {
+      "epoch": 45.85,
+      "learning_rate": 0.0001704230769230769,
+      "loss": 0.8678,
+      "step": 4998
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 0.00017039423076923075,
+      "loss": 0.7869,
+      "step": 4999
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 0.0001703653846153846,
+      "loss": 0.774,
+      "step": 5000
+    },
+    {
+      "epoch": 45.87,
+      "eval_cer": 0.09707791721015308,
+      "eval_loss": 0.35577356815338135,
+      "eval_runtime": 13.9068,
+      "eval_samples_per_second": 118.432,
+      "eval_steps_per_second": 1.87,
+      "eval_wer": 0.3922990501480952,
+      "step": 5000
+    },
+    {
+      "epoch": 45.88,
+      "learning_rate": 0.00017033653846153845,
+      "loss": 0.7546,
+      "step": 5001
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 0.0001703076923076923,
+      "loss": 0.8216,
+      "step": 5002
+    },
+    {
+      "epoch": 45.9,
+      "learning_rate": 0.00017027884615384613,
+      "loss": 0.8245,
+      "step": 5003
+    },
+    {
+      "epoch": 45.91,
+      "learning_rate": 0.00017025,
+      "loss": 0.7721,
+      "step": 5004
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 0.00017022115384615386,
+      "loss": 0.6562,
+      "step": 5005
+    },
+    {
+      "epoch": 45.93,
+      "learning_rate": 0.00017019230769230767,
+      "loss": 0.8885,
+      "step": 5006
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 0.00017016346153846153,
+      "loss": 0.7578,
+      "step": 5007
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 0.00017013461538461537,
+      "loss": 0.7636,
+      "step": 5008
+    },
+    {
+      "epoch": 45.95,
+      "learning_rate": 0.0001701057692307692,
+      "loss": 0.7741,
+      "step": 5009
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 0.00017007692307692307,
+      "loss": 0.8179,
+      "step": 5010
+    },
+    {
+      "epoch": 45.97,
+      "learning_rate": 0.0001700480769230769,
+      "loss": 0.83,
+      "step": 5011
+    },
+    {
+      "epoch": 45.98,
+      "learning_rate": 0.00017001923076923075,
+      "loss": 0.7711,
+      "step": 5012
+    },
+    {
+      "epoch": 45.99,
+      "learning_rate": 0.0001699903846153846,
+      "loss": 0.7739,
+      "step": 5013
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 0.00016996153846153845,
+      "loss": 0.7556,
+      "step": 5014
+    },
+    {
+      "epoch": 46.01,
+      "learning_rate": 0.00016993269230769232,
+      "loss": 0.8378,
+      "step": 5015
+    },
+    {
+      "epoch": 46.02,
+      "learning_rate": 0.00016990384615384613,
+      "loss": 0.783,
+      "step": 5016
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 0.000169875,
+      "loss": 0.702,
+      "step": 5017
+    },
+    {
+      "epoch": 46.04,
+      "learning_rate": 0.00016984615384615383,
+      "loss": 0.7238,
+      "step": 5018
+    },
+    {
+      "epoch": 46.05,
+      "learning_rate": 0.00016981730769230767,
+      "loss": 0.6889,
+      "step": 5019
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 0.00016978846153846154,
+      "loss": 0.733,
+      "step": 5020
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 0.00016975961538461538,
+      "loss": 0.7523,
+      "step": 5021
+    },
+    {
+      "epoch": 46.07,
+      "learning_rate": 0.0001697307692307692,
+      "loss": 0.6732,
+      "step": 5022
+    },
+    {
+      "epoch": 46.08,
+      "learning_rate": 0.00016970192307692305,
+      "loss": 0.8168,
+      "step": 5023
+    },
+    {
+      "epoch": 46.09,
+      "learning_rate": 0.00016967307692307692,
+      "loss": 0.7618,
+      "step": 5024
+    },
+    {
+      "epoch": 46.1,
+      "learning_rate": 0.00016964423076923073,
+      "loss": 0.7499,
+      "step": 5025
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 0.0001696153846153846,
+      "loss": 0.7667,
+      "step": 5026
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 0.00016958653846153846,
+      "loss": 0.614,
+      "step": 5027
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 0.0001695576923076923,
+      "loss": 0.8102,
+      "step": 5028
+    },
+    {
+      "epoch": 46.14,
+      "learning_rate": 0.00016952884615384613,
+      "loss": 0.8765,
+      "step": 5029
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 0.00016949999999999997,
+      "loss": 0.7791,
+      "step": 5030
+    },
+    {
+      "epoch": 46.16,
+      "learning_rate": 0.00016947115384615384,
+      "loss": 0.8107,
+      "step": 5031
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 0.00016944230769230768,
+      "loss": 0.834,
+      "step": 5032
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 0.00016941346153846151,
+      "loss": 0.8339,
+      "step": 5033
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 0.00016938461538461538,
+      "loss": 0.7684,
+      "step": 5034
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 0.0001693557692307692,
+      "loss": 0.8174,
+      "step": 5035
+    },
+    {
+      "epoch": 46.2,
+      "learning_rate": 0.00016932692307692306,
+      "loss": 0.6863,
+      "step": 5036
+    },
+    {
+      "epoch": 46.21,
+      "learning_rate": 0.00016929807692307692,
+      "loss": 0.6887,
+      "step": 5037
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 0.00016926923076923073,
+      "loss": 0.719,
+      "step": 5038
+    },
+    {
+      "epoch": 46.23,
+      "learning_rate": 0.0001692403846153846,
+      "loss": 0.9394,
+      "step": 5039
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 0.00016921153846153844,
+      "loss": 0.7271,
+      "step": 5040
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 0.0001691826923076923,
+      "loss": 0.7002,
+      "step": 5041
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 0.00016915384615384614,
+      "loss": 0.7449,
+      "step": 5042
+    },
+    {
+      "epoch": 46.27,
+      "learning_rate": 0.00016912499999999998,
+      "loss": 0.8898,
+      "step": 5043
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 0.00016909615384615384,
+      "loss": 0.8541,
+      "step": 5044
+    },
+    {
+      "epoch": 46.28,
+      "learning_rate": 0.00016906730769230765,
+      "loss": 0.8173,
+      "step": 5045
+    },
+    {
+      "epoch": 46.29,
+      "learning_rate": 0.00016903846153846152,
+      "loss": 0.7063,
+      "step": 5046
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 0.00016900961538461538,
+      "loss": 0.705,
+      "step": 5047
+    },
+    {
+      "epoch": 46.31,
+      "learning_rate": 0.0001689807692307692,
+      "loss": 0.7752,
+      "step": 5048
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 0.00016895192307692306,
+      "loss": 0.7949,
+      "step": 5049
+    },
+    {
+      "epoch": 46.33,
+      "learning_rate": 0.0001689230769230769,
+      "loss": 0.8624,
+      "step": 5050
+    },
+    {
+      "epoch": 46.34,
+      "learning_rate": 0.00016889423076923076,
+      "loss": 0.792,
+      "step": 5051
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 0.0001688653846153846,
+      "loss": 0.744,
+      "step": 5052
+    },
+    {
+      "epoch": 46.36,
+      "learning_rate": 0.00016883653846153844,
+      "loss": 0.8409,
+      "step": 5053
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 0.0001688076923076923,
+      "loss": 0.7876,
+      "step": 5054
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 0.00016877884615384612,
+      "loss": 0.7452,
+      "step": 5055
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 0.00016874999999999998,
+      "loss": 0.8232,
+      "step": 5056
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 0.00016872115384615385,
+      "loss": 0.7044,
+      "step": 5057
+    },
+    {
+      "epoch": 46.4,
+      "learning_rate": 0.00016869230769230766,
+      "loss": 0.7553,
+      "step": 5058
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 0.00016866346153846152,
+      "loss": 0.7409,
+      "step": 5059
+    },
+    {
+      "epoch": 46.42,
+      "learning_rate": 0.00016863461538461536,
+      "loss": 0.7297,
+      "step": 5060
+    },
+    {
+      "epoch": 46.43,
+      "learning_rate": 0.00016860576923076923,
+      "loss": 0.8141,
+      "step": 5061
+    },
+    {
+      "epoch": 46.44,
+      "learning_rate": 0.00016857692307692306,
+      "loss": 0.7321,
+      "step": 5062
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 0.0001685480769230769,
+      "loss": 0.7416,
+      "step": 5063
+    },
+    {
+      "epoch": 46.46,
+      "learning_rate": 0.00016851923076923077,
+      "loss": 0.8214,
+      "step": 5064
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 0.00016849038461538458,
+      "loss": 0.8039,
+      "step": 5065
+    },
+    {
+      "epoch": 46.48,
+      "learning_rate": 0.00016846153846153844,
+      "loss": 0.7989,
+      "step": 5066
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 0.0001684326923076923,
+      "loss": 0.7032,
+      "step": 5067
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 0.00016840384615384612,
+      "loss": 0.735,
+      "step": 5068
+    },
+    {
+      "epoch": 46.5,
+      "learning_rate": 0.00016837499999999999,
+      "loss": 0.8637,
+      "step": 5069
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 0.00016834615384615382,
+      "loss": 0.8727,
+      "step": 5070
+    },
+    {
+      "epoch": 46.52,
+      "learning_rate": 0.0001683173076923077,
+      "loss": 0.7234,
+      "step": 5071
+    },
+    {
+      "epoch": 46.53,
+      "learning_rate": 0.00016828846153846153,
+      "loss": 0.7207,
+      "step": 5072
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 0.00016825961538461537,
+      "loss": 0.7873,
+      "step": 5073
+    },
+    {
+      "epoch": 46.55,
+      "learning_rate": 0.00016823076923076923,
+      "loss": 0.7456,
+      "step": 5074
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 0.00016820192307692304,
+      "loss": 0.7295,
+      "step": 5075
+    },
+    {
+      "epoch": 46.57,
+      "learning_rate": 0.0001681730769230769,
+      "loss": 0.7783,
+      "step": 5076
+    },
+    {
+      "epoch": 46.58,
+      "learning_rate": 0.00016814423076923077,
+      "loss": 0.9085,
+      "step": 5077
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 0.00016811538461538458,
+      "loss": 0.7708,
+      "step": 5078
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 0.00016808653846153845,
+      "loss": 0.7694,
+      "step": 5079
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 0.0001680576923076923,
+      "loss": 0.8843,
+      "step": 5080
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 0.00016802884615384615,
+      "loss": 0.7996,
+      "step": 5081
+    },
+    {
+      "epoch": 46.62,
+      "learning_rate": 0.000168,
+      "loss": 0.7991,
+      "step": 5082
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 0.00016797115384615383,
+      "loss": 0.7785,
+      "step": 5083
+    },
+    {
+      "epoch": 46.64,
+      "learning_rate": 0.0001679423076923077,
+      "loss": 0.8703,
+      "step": 5084
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 0.0001679134615384615,
+      "loss": 0.8393,
+      "step": 5085
+    },
+    {
+      "epoch": 46.66,
+      "learning_rate": 0.00016788461538461537,
+      "loss": 0.8454,
+      "step": 5086
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 0.00016785576923076924,
+      "loss": 0.9149,
+      "step": 5087
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 0.00016782692307692305,
+      "loss": 0.8245,
+      "step": 5088
+    },
+    {
+      "epoch": 46.69,
+      "learning_rate": 0.0001677980769230769,
+      "loss": 0.8214,
+      "step": 5089
+    },
+    {
+      "epoch": 46.7,
+      "learning_rate": 0.00016776923076923075,
+      "loss": 0.8104,
+      "step": 5090
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 0.00016774038461538462,
+      "loss": 0.7634,
+      "step": 5091
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 0.00016771153846153843,
+      "loss": 0.8008,
+      "step": 5092
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 0.0001676826923076923,
+      "loss": 0.733,
+      "step": 5093
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 0.00016765384615384616,
+      "loss": 0.8787,
+      "step": 5094
+    },
+    {
+      "epoch": 46.74,
+      "learning_rate": 0.00016762499999999997,
+      "loss": 0.6657,
+      "step": 5095
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 0.00016759615384615383,
+      "loss": 0.8023,
+      "step": 5096
+    },
+    {
+      "epoch": 46.76,
+      "learning_rate": 0.00016756730769230767,
+      "loss": 0.8376,
+      "step": 5097
+    },
+    {
+      "epoch": 46.77,
+      "learning_rate": 0.0001675384615384615,
+      "loss": 0.7871,
+      "step": 5098
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 0.00016750961538461537,
+      "loss": 0.6684,
+      "step": 5099
+    },
+    {
+      "epoch": 46.79,
+      "learning_rate": 0.0001674807692307692,
+      "loss": 0.7571,
+      "step": 5100
+    },
+    {
+      "epoch": 46.8,
+      "learning_rate": 0.00016745192307692308,
+      "loss": 0.8283,
+      "step": 5101
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 0.0001674230769230769,
+      "loss": 0.7427,
+      "step": 5102
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 0.00016739423076923075,
+      "loss": 0.7813,
+      "step": 5103
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 0.00016736538461538462,
+      "loss": 0.6944,
+      "step": 5104
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 0.00016733653846153843,
+      "loss": 0.7049,
+      "step": 5105
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 0.0001673076923076923,
+      "loss": 0.7111,
+      "step": 5106
+    },
+    {
+      "epoch": 46.85,
+      "learning_rate": 0.00016727884615384613,
+      "loss": 0.872,
+      "step": 5107
+    },
+    {
+      "epoch": 46.86,
+      "learning_rate": 0.00016724999999999997,
+      "loss": 0.7915,
+      "step": 5108
+    },
+    {
+      "epoch": 46.87,
+      "learning_rate": 0.00016722115384615384,
+      "loss": 0.6325,
+      "step": 5109
+    },
+    {
+      "epoch": 46.88,
+      "learning_rate": 0.00016719230769230768,
+      "loss": 0.7406,
+      "step": 5110
+    },
+    {
+      "epoch": 46.89,
+      "learning_rate": 0.00016716346153846154,
+      "loss": 0.707,
+      "step": 5111
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 0.00016713461538461535,
+      "loss": 0.7042,
+      "step": 5112
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 0.00016710576923076922,
+      "loss": 0.7741,
+      "step": 5113
+    },
+    {
+      "epoch": 46.92,
+      "learning_rate": 0.00016707692307692308,
+      "loss": 0.7985,
+      "step": 5114
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 0.0001670480769230769,
+      "loss": 0.7436,
+      "step": 5115
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 0.00016701923076923076,
+      "loss": 0.7816,
+      "step": 5116
+    },
+    {
+      "epoch": 46.94,
+      "learning_rate": 0.0001669903846153846,
+      "loss": 0.7422,
+      "step": 5117
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 0.00016696153846153844,
+      "loss": 0.8182,
+      "step": 5118
+    },
+    {
+      "epoch": 46.96,
+      "learning_rate": 0.0001669326923076923,
+      "loss": 0.8498,
+      "step": 5119
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 0.00016690384615384614,
+      "loss": 0.7121,
+      "step": 5120
+    },
+    {
+      "epoch": 46.98,
+      "learning_rate": 0.000166875,
+      "loss": 0.753,
+      "step": 5121
+    },
+    {
+      "epoch": 46.99,
+      "learning_rate": 0.00016684615384615382,
+      "loss": 0.8736,
+      "step": 5122
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 0.00016681730769230768,
+      "loss": 0.7586,
+      "step": 5123
+    },
+    {
+      "epoch": 47.01,
+      "learning_rate": 0.00016678846153846155,
+      "loss": 0.8853,
+      "step": 5124
+    },
+    {
+      "epoch": 47.02,
+      "learning_rate": 0.00016675961538461536,
+      "loss": 0.8201,
+      "step": 5125
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 0.00016673076923076922,
+      "loss": 0.8048,
+      "step": 5126
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 0.00016670192307692306,
+      "loss": 0.7471,
+      "step": 5127
+    },
+    {
+      "epoch": 47.05,
+      "learning_rate": 0.0001666730769230769,
+      "loss": 0.7586,
+      "step": 5128
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 0.00016664423076923076,
+      "loss": 0.8224,
+      "step": 5129
+    },
+    {
+      "epoch": 47.06,
+      "learning_rate": 0.0001666153846153846,
+      "loss": 0.7626,
+      "step": 5130
+    },
+    {
+      "epoch": 47.07,
+      "learning_rate": 0.00016658653846153847,
+      "loss": 0.7483,
+      "step": 5131
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 0.00016655769230769228,
+      "loss": 0.8156,
+      "step": 5132
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 0.00016652884615384614,
+      "loss": 0.7825,
+      "step": 5133
+    },
+    {
+      "epoch": 47.1,
+      "learning_rate": 0.0001665,
+      "loss": 0.8394,
+      "step": 5134
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 0.00016647115384615382,
+      "loss": 0.7736,
+      "step": 5135
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 0.00016644230769230768,
+      "loss": 0.8594,
+      "step": 5136
+    },
+    {
+      "epoch": 47.13,
+      "learning_rate": 0.00016641346153846152,
+      "loss": 0.7435,
+      "step": 5137
+    },
+    {
+      "epoch": 47.14,
+      "learning_rate": 0.00016638461538461536,
+      "loss": 0.7768,
+      "step": 5138
+    },
+    {
+      "epoch": 47.15,
+      "learning_rate": 0.00016635576923076923,
+      "loss": 0.8231,
+      "step": 5139
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 0.00016632692307692306,
+      "loss": 0.6788,
+      "step": 5140
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 0.00016629807692307693,
+      "loss": 0.7671,
+      "step": 5141
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 0.00016626923076923074,
+      "loss": 0.8659,
+      "step": 5142
+    },
+    {
+      "epoch": 47.18,
+      "learning_rate": 0.0001662403846153846,
+      "loss": 0.69,
+      "step": 5143
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 0.00016621153846153847,
+      "loss": 0.7183,
+      "step": 5144
+    },
+    {
+      "epoch": 47.2,
+      "learning_rate": 0.00016618269230769228,
+      "loss": 0.6922,
+      "step": 5145
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 0.00016615384615384615,
+      "loss": 0.8988,
+      "step": 5146
+    },
+    {
+      "epoch": 47.22,
+      "learning_rate": 0.00016612499999999999,
+      "loss": 0.8073,
+      "step": 5147
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 0.00016609615384615382,
+      "loss": 0.9407,
+      "step": 5148
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 0.0001660673076923077,
+      "loss": 0.685,
+      "step": 5149
+    },
+    {
+      "epoch": 47.25,
+      "learning_rate": 0.00016603846153846153,
+      "loss": 0.762,
+      "step": 5150
+    },
+    {
+      "epoch": 47.26,
+      "learning_rate": 0.0001660096153846154,
+      "loss": 0.7913,
+      "step": 5151
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 0.0001659807692307692,
+      "loss": 0.8134,
+      "step": 5152
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 0.00016595192307692307,
+      "loss": 0.7785,
+      "step": 5153
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 0.00016592307692307688,
+      "loss": 0.6165,
+      "step": 5154
+    },
+    {
+      "epoch": 47.29,
+      "learning_rate": 0.00016589423076923075,
+      "loss": 0.7235,
+      "step": 5155
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 0.0001658653846153846,
+      "loss": 0.7884,
+      "step": 5156
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 0.00016583653846153845,
+      "loss": 0.6673,
+      "step": 5157
+    },
+    {
+      "epoch": 47.32,
+      "learning_rate": 0.0001658076923076923,
+      "loss": 0.7916,
+      "step": 5158
+    },
+    {
+      "epoch": 47.33,
+      "learning_rate": 0.00016577884615384612,
+      "loss": 0.75,
+      "step": 5159
+    },
+    {
+      "epoch": 47.34,
+      "learning_rate": 0.00016575,
+      "loss": 0.819,
+      "step": 5160
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 0.00016572115384615386,
+      "loss": 0.7208,
+      "step": 5161
+    },
+    {
+      "epoch": 47.36,
+      "learning_rate": 0.00016569230769230767,
+      "loss": 0.7967,
+      "step": 5162
+    },
+    {
+      "epoch": 47.37,
+      "learning_rate": 0.00016566346153846153,
+      "loss": 0.6714,
+      "step": 5163
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 0.00016563461538461534,
+      "loss": 0.8687,
+      "step": 5164
+    },
+    {
+      "epoch": 47.39,
+      "learning_rate": 0.0001656057692307692,
+      "loss": 0.7527,
+      "step": 5165
+    },
+    {
+      "epoch": 47.39,
+      "learning_rate": 0.00016557692307692307,
+      "loss": 0.6487,
+      "step": 5166
+    },
+    {
+      "epoch": 47.4,
+      "learning_rate": 0.0001655480769230769,
+      "loss": 0.8661,
+      "step": 5167
+    },
+    {
+      "epoch": 47.41,
+      "learning_rate": 0.00016551923076923075,
+      "loss": 0.8007,
+      "step": 5168
+    },
+    {
+      "epoch": 47.42,
+      "learning_rate": 0.0001654903846153846,
+      "loss": 0.6612,
+      "step": 5169
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 0.00016546153846153845,
+      "loss": 0.6465,
+      "step": 5170
+    },
+    {
+      "epoch": 47.44,
+      "learning_rate": 0.00016543269230769232,
+      "loss": 0.7493,
+      "step": 5171
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 0.00016540384615384613,
+      "loss": 0.8826,
+      "step": 5172
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 0.000165375,
+      "loss": 0.7055,
+      "step": 5173
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 0.0001653461538461538,
+      "loss": 0.7324,
+      "step": 5174
+    },
+    {
+      "epoch": 47.48,
+      "learning_rate": 0.00016531730769230767,
+      "loss": 0.8374,
+      "step": 5175
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 0.00016528846153846154,
+      "loss": 0.8341,
+      "step": 5176
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 0.00016525961538461537,
+      "loss": 0.8305,
+      "step": 5177
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 0.0001652307692307692,
+      "loss": 0.8391,
+      "step": 5178
+    },
+    {
+      "epoch": 47.51,
+      "learning_rate": 0.00016520192307692305,
+      "loss": 0.7679,
+      "step": 5179
+    },
+    {
+      "epoch": 47.52,
+      "learning_rate": 0.00016517307692307692,
+      "loss": 0.8422,
+      "step": 5180
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 0.00016514423076923075,
+      "loss": 0.7257,
+      "step": 5181
+    },
+    {
+      "epoch": 47.54,
+      "learning_rate": 0.0001651153846153846,
+      "loss": 0.6361,
+      "step": 5182
+    },
+    {
+      "epoch": 47.55,
+      "learning_rate": 0.00016508653846153846,
+      "loss": 0.7741,
+      "step": 5183
+    },
+    {
+      "epoch": 47.56,
+      "learning_rate": 0.00016505769230769227,
+      "loss": 0.8191,
+      "step": 5184
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 0.00016502884615384613,
+      "loss": 0.9011,
+      "step": 5185
+    },
+    {
+      "epoch": 47.58,
+      "learning_rate": 0.000165,
+      "loss": 0.7447,
+      "step": 5186
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 0.00016497115384615384,
+      "loss": 0.7675,
+      "step": 5187
+    },
+    {
+      "epoch": 47.6,
+      "learning_rate": 0.00016494230769230768,
+      "loss": 0.7801,
+      "step": 5188
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 0.00016491346153846151,
+      "loss": 0.824,
+      "step": 5189
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 0.00016488461538461538,
+      "loss": 0.811,
+      "step": 5190
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 0.00016485576923076922,
+      "loss": 0.8007,
+      "step": 5191
+    },
+    {
+      "epoch": 47.63,
+      "learning_rate": 0.00016482692307692306,
+      "loss": 0.7115,
+      "step": 5192
+    },
+    {
+      "epoch": 47.64,
+      "learning_rate": 0.00016479807692307692,
+      "loss": 0.7492,
+      "step": 5193
+    },
+    {
+      "epoch": 47.65,
+      "learning_rate": 0.00016476923076923073,
+      "loss": 0.715,
+      "step": 5194
+    },
+    {
+      "epoch": 47.66,
+      "learning_rate": 0.0001647403846153846,
+      "loss": 0.8748,
+      "step": 5195
+    },
+    {
+      "epoch": 47.67,
+      "learning_rate": 0.00016471153846153846,
+      "loss": 0.6603,
+      "step": 5196
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 0.0001646826923076923,
+      "loss": 0.7562,
+      "step": 5197
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 0.00016465384615384614,
+      "loss": 0.8174,
+      "step": 5198
+    },
+    {
+      "epoch": 47.7,
+      "learning_rate": 0.00016462499999999998,
+      "loss": 0.8307,
+      "step": 5199
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 0.00016459615384615384,
+      "loss": 0.8279,
+      "step": 5200
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 0.00016456730769230768,
+      "loss": 0.835,
+      "step": 5201
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 0.00016453846153846152,
+      "loss": 0.8405,
+      "step": 5202
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 0.00016450961538461538,
+      "loss": 0.8045,
+      "step": 5203
+    },
+    {
+      "epoch": 47.74,
+      "learning_rate": 0.0001644807692307692,
+      "loss": 0.8126,
+      "step": 5204
+    },
+    {
+      "epoch": 47.75,
+      "learning_rate": 0.00016445192307692306,
+      "loss": 0.7654,
+      "step": 5205
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 0.00016442307692307692,
+      "loss": 0.6958,
+      "step": 5206
+    },
+    {
+      "epoch": 47.77,
+      "learning_rate": 0.00016439423076923076,
+      "loss": 0.7123,
+      "step": 5207
+    },
+    {
+      "epoch": 47.78,
+      "learning_rate": 0.0001643653846153846,
+      "loss": 0.7665,
+      "step": 5208
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 0.00016433653846153844,
+      "loss": 0.9309,
+      "step": 5209
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 0.0001643076923076923,
+      "loss": 0.9502,
+      "step": 5210
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 0.00016427884615384614,
+      "loss": 0.7176,
+      "step": 5211
+    },
+    {
+      "epoch": 47.82,
+      "learning_rate": 0.00016424999999999998,
+      "loss": 0.7381,
+      "step": 5212
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 0.00016422115384615385,
+      "loss": 0.7085,
+      "step": 5213
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 0.00016419230769230766,
+      "loss": 0.8174,
+      "step": 5214
+    },
+    {
+      "epoch": 47.84,
+      "learning_rate": 0.00016416346153846152,
+      "loss": 0.8275,
+      "step": 5215
+    },
+    {
+      "epoch": 47.85,
+      "learning_rate": 0.0001641346153846154,
+      "loss": 0.7643,
+      "step": 5216
+    },
+    {
+      "epoch": 47.86,
+      "learning_rate": 0.00016410576923076923,
+      "loss": 0.6938,
+      "step": 5217
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 0.00016407692307692306,
+      "loss": 0.8587,
+      "step": 5218
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 0.0001640480769230769,
+      "loss": 0.7019,
+      "step": 5219
+    },
+    {
+      "epoch": 47.89,
+      "learning_rate": 0.00016401923076923077,
+      "loss": 0.7947,
+      "step": 5220
+    },
+    {
+      "epoch": 47.9,
+      "learning_rate": 0.00016399038461538458,
+      "loss": 0.807,
+      "step": 5221
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 0.00016396153846153844,
+      "loss": 0.879,
+      "step": 5222
+    },
+    {
+      "epoch": 47.92,
+      "learning_rate": 0.0001639326923076923,
+      "loss": 0.739,
+      "step": 5223
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 0.00016390384615384612,
+      "loss": 0.8966,
+      "step": 5224
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 0.00016387499999999999,
+      "loss": 0.6776,
+      "step": 5225
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 0.00016384615384615382,
+      "loss": 0.8525,
+      "step": 5226
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 0.0001638173076923077,
+      "loss": 0.7318,
+      "step": 5227
+    },
+    {
+      "epoch": 47.96,
+      "learning_rate": 0.00016378846153846153,
+      "loss": 0.746,
+      "step": 5228
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 0.00016375961538461537,
+      "loss": 0.8123,
+      "step": 5229
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 0.00016373076923076923,
+      "loss": 0.7377,
+      "step": 5230
+    },
+    {
+      "epoch": 47.99,
+      "learning_rate": 0.00016370192307692304,
+      "loss": 0.7523,
+      "step": 5231
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 0.0001636730769230769,
+      "loss": 0.8698,
+      "step": 5232
+    },
+    {
+      "epoch": 48.01,
+      "learning_rate": 0.00016364423076923077,
+      "loss": 0.8477,
+      "step": 5233
+    },
+    {
+      "epoch": 48.02,
+      "learning_rate": 0.00016361538461538458,
+      "loss": 0.8375,
+      "step": 5234
+    },
+    {
+      "epoch": 48.03,
+      "learning_rate": 0.00016358653846153845,
+      "loss": 0.7527,
+      "step": 5235
+    },
+    {
+      "epoch": 48.04,
+      "learning_rate": 0.00016355769230769229,
+      "loss": 0.6872,
+      "step": 5236
+    },
+    {
+      "epoch": 48.05,
+      "learning_rate": 0.00016352884615384615,
+      "loss": 0.7407,
+      "step": 5237
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 0.0001635,
+      "loss": 0.8266,
+      "step": 5238
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 0.00016347115384615383,
+      "loss": 0.7256,
+      "step": 5239
+    },
+    {
+      "epoch": 48.07,
+      "learning_rate": 0.0001634423076923077,
+      "loss": 0.7337,
+      "step": 5240
+    },
+    {
+      "epoch": 48.08,
+      "learning_rate": 0.0001634134615384615,
+      "loss": 0.7485,
+      "step": 5241
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 0.00016338461538461537,
+      "loss": 0.797,
+      "step": 5242
+    },
+    {
+      "epoch": 48.1,
+      "learning_rate": 0.00016335576923076923,
+      "loss": 0.7008,
+      "step": 5243
+    },
+    {
+      "epoch": 48.11,
+      "learning_rate": 0.00016332692307692305,
+      "loss": 0.7809,
+      "step": 5244
+    },
+    {
+      "epoch": 48.12,
+      "learning_rate": 0.0001632980769230769,
+      "loss": 0.8491,
+      "step": 5245
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 0.00016326923076923075,
+      "loss": 0.8166,
+      "step": 5246
+    },
+    {
+      "epoch": 48.14,
+      "learning_rate": 0.0001632403846153846,
+      "loss": 0.6772,
+      "step": 5247
+    },
+    {
+      "epoch": 48.15,
+      "learning_rate": 0.00016321153846153845,
+      "loss": 0.7452,
+      "step": 5248
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 0.0001631826923076923,
+      "loss": 0.7249,
+      "step": 5249
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 0.00016315384615384616,
+      "loss": 0.7507,
+      "step": 5250
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 0.00016312499999999997,
+      "loss": 0.8199,
+      "step": 5251
+    },
+    {
+      "epoch": 48.18,
+      "learning_rate": 0.00016309615384615383,
+      "loss": 0.7995,
+      "step": 5252
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 0.0001630673076923077,
+      "loss": 0.8105,
+      "step": 5253
+    },
+    {
+      "epoch": 48.2,
+      "learning_rate": 0.0001630384615384615,
+      "loss": 0.7648,
+      "step": 5254
+    },
+    {
+      "epoch": 48.21,
+      "learning_rate": 0.00016300961538461537,
+      "loss": 0.8211,
+      "step": 5255
+    },
+    {
+      "epoch": 48.22,
+      "learning_rate": 0.0001629807692307692,
+      "loss": 0.8441,
+      "step": 5256
+    },
+    {
+      "epoch": 48.23,
+      "learning_rate": 0.00016295192307692305,
+      "loss": 0.7511,
+      "step": 5257
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 0.00016292307692307692,
+      "loss": 0.8939,
+      "step": 5258
+    },
+    {
+      "epoch": 48.25,
+      "learning_rate": 0.00016289423076923075,
+      "loss": 0.8138,
+      "step": 5259
+    },
+    {
+      "epoch": 48.26,
+      "learning_rate": 0.00016286538461538462,
+      "loss": 0.7437,
+      "step": 5260
+    },
+    {
+      "epoch": 48.27,
+      "learning_rate": 0.00016283653846153843,
+      "loss": 0.7852,
+      "step": 5261
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 0.0001628076923076923,
+      "loss": 0.7967,
+      "step": 5262
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 0.00016277884615384616,
+      "loss": 0.7012,
+      "step": 5263
+    },
+    {
+      "epoch": 48.29,
+      "learning_rate": 0.00016274999999999997,
+      "loss": 0.7002,
+      "step": 5264
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 0.00016272115384615384,
+      "loss": 0.6821,
+      "step": 5265
+    },
+    {
+      "epoch": 48.31,
+      "learning_rate": 0.00016269230769230767,
+      "loss": 0.8173,
+      "step": 5266
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 0.0001626634615384615,
+      "loss": 0.7566,
+      "step": 5267
+    },
+    {
+      "epoch": 48.33,
+      "learning_rate": 0.00016263461538461538,
+      "loss": 0.761,
+      "step": 5268
+    },
+    {
+      "epoch": 48.34,
+      "learning_rate": 0.00016260576923076922,
+      "loss": 0.8969,
+      "step": 5269
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 0.00016257692307692308,
+      "loss": 0.7799,
+      "step": 5270
+    },
+    {
+      "epoch": 48.36,
+      "learning_rate": 0.0001625480769230769,
+      "loss": 0.7552,
+      "step": 5271
+    },
+    {
+      "epoch": 48.37,
+      "learning_rate": 0.00016251923076923076,
+      "loss": 0.741,
+      "step": 5272
+    },
+    {
+      "epoch": 48.38,
+      "learning_rate": 0.00016249038461538462,
+      "loss": 0.7351,
+      "step": 5273
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 0.00016246153846153843,
+      "loss": 0.7936,
+      "step": 5274
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 0.0001624326923076923,
+      "loss": 0.773,
+      "step": 5275
+    },
+    {
+      "epoch": 48.4,
+      "learning_rate": 0.00016240384615384614,
+      "loss": 0.7809,
+      "step": 5276
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 0.00016237499999999998,
+      "loss": 0.7523,
+      "step": 5277
+    },
+    {
+      "epoch": 48.42,
+      "learning_rate": 0.00016234615384615384,
+      "loss": 0.655,
+      "step": 5278
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 0.00016231730769230768,
+      "loss": 0.757,
+      "step": 5279
+    },
+    {
+      "epoch": 48.44,
+      "learning_rate": 0.00016228846153846154,
+      "loss": 0.7389,
+      "step": 5280
+    },
+    {
+      "epoch": 48.45,
+      "learning_rate": 0.00016225961538461536,
+      "loss": 0.7286,
+      "step": 5281
+    },
+    {
+      "epoch": 48.46,
+      "learning_rate": 0.00016223076923076922,
+      "loss": 0.7958,
+      "step": 5282
+    },
+    {
+      "epoch": 48.47,
+      "learning_rate": 0.00016220192307692309,
+      "loss": 0.8128,
+      "step": 5283
+    },
+    {
+      "epoch": 48.48,
+      "learning_rate": 0.0001621730769230769,
+      "loss": 0.7603,
+      "step": 5284
+    },
+    {
+      "epoch": 48.49,
+      "learning_rate": 0.00016214423076923076,
+      "loss": 0.7794,
+      "step": 5285
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 0.0001621153846153846,
+      "loss": 0.7509,
+      "step": 5286
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 0.00016208653846153844,
+      "loss": 0.7689,
+      "step": 5287
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 0.00016205769230769228,
+      "loss": 0.7598,
+      "step": 5288
+    },
+    {
+      "epoch": 48.52,
+      "learning_rate": 0.00016202884615384614,
+      "loss": 0.8336,
+      "step": 5289
+    },
+    {
+      "epoch": 48.53,
+      "learning_rate": 0.000162,
+      "loss": 0.7637,
+      "step": 5290
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 0.00016197115384615382,
+      "loss": 0.7885,
+      "step": 5291
+    },
+    {
+      "epoch": 48.55,
+      "learning_rate": 0.00016194230769230768,
+      "loss": 0.7848,
+      "step": 5292
+    },
+    {
+      "epoch": 48.56,
+      "learning_rate": 0.00016191346153846152,
+      "loss": 0.7617,
+      "step": 5293
+    },
+    {
+      "epoch": 48.57,
+      "learning_rate": 0.00016188461538461536,
+      "loss": 0.7849,
+      "step": 5294
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 0.00016185576923076923,
+      "loss": 0.7929,
+      "step": 5295
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 0.00016182692307692306,
+      "loss": 0.6598,
+      "step": 5296
+    },
+    {
+      "epoch": 48.6,
+      "learning_rate": 0.0001617980769230769,
+      "loss": 0.8197,
+      "step": 5297
+    },
+    {
+      "epoch": 48.61,
+      "learning_rate": 0.00016176923076923074,
+      "loss": 0.8292,
+      "step": 5298
+    },
+    {
+      "epoch": 48.61,
+      "learning_rate": 0.0001617403846153846,
+      "loss": 0.8357,
+      "step": 5299
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 0.00016171153846153847,
+      "loss": 0.6624,
+      "step": 5300
+    },
+    {
+      "epoch": 48.63,
+      "learning_rate": 0.00016168269230769228,
+      "loss": 0.7312,
+      "step": 5301
+    },
+    {
+      "epoch": 48.64,
+      "learning_rate": 0.00016165384615384615,
+      "loss": 0.8359,
+      "step": 5302
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 0.00016162499999999996,
+      "loss": 0.6577,
+      "step": 5303
+    },
+    {
+      "epoch": 48.66,
+      "learning_rate": 0.00016159615384615382,
+      "loss": 0.8311,
+      "step": 5304
+    },
+    {
+      "epoch": 48.67,
+      "learning_rate": 0.0001615673076923077,
+      "loss": 0.7317,
+      "step": 5305
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 0.00016153846153846153,
+      "loss": 0.8544,
+      "step": 5306
+    },
+    {
+      "epoch": 48.69,
+      "learning_rate": 0.00016150961538461536,
+      "loss": 0.7185,
+      "step": 5307
+    },
+    {
+      "epoch": 48.7,
+      "learning_rate": 0.0001614807692307692,
+      "loss": 0.7663,
+      "step": 5308
+    },
+    {
+      "epoch": 48.71,
+      "learning_rate": 0.00016145192307692307,
+      "loss": 0.666,
+      "step": 5309
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 0.00016142307692307693,
+      "loss": 0.6997,
+      "step": 5310
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 0.00016139423076923074,
+      "loss": 0.723,
+      "step": 5311
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 0.0001613653846153846,
+      "loss": 0.8398,
+      "step": 5312
+    },
+    {
+      "epoch": 48.74,
+      "learning_rate": 0.00016133653846153842,
+      "loss": 0.8273,
+      "step": 5313
+    },
+    {
+      "epoch": 48.75,
+      "learning_rate": 0.00016130769230769229,
+      "loss": 0.7692,
+      "step": 5314
+    },
+    {
+      "epoch": 48.76,
+      "learning_rate": 0.00016127884615384615,
+      "loss": 0.704,
+      "step": 5315
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 0.00016125,
+      "loss": 0.6923,
+      "step": 5316
+    },
+    {
+      "epoch": 48.78,
+      "learning_rate": 0.00016122115384615383,
+      "loss": 0.9172,
+      "step": 5317
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 0.00016119230769230767,
+      "loss": 0.73,
+      "step": 5318
+    },
+    {
+      "epoch": 48.8,
+      "learning_rate": 0.00016116346153846153,
+      "loss": 0.7108,
+      "step": 5319
+    },
+    {
+      "epoch": 48.81,
+      "learning_rate": 0.0001611346153846154,
+      "loss": 0.8451,
+      "step": 5320
+    },
+    {
+      "epoch": 48.82,
+      "learning_rate": 0.0001611057692307692,
+      "loss": 0.734,
+      "step": 5321
+    },
+    {
+      "epoch": 48.83,
+      "learning_rate": 0.00016107692307692307,
+      "loss": 0.7284,
+      "step": 5322
+    },
+    {
+      "epoch": 48.83,
+      "learning_rate": 0.00016104807692307688,
+      "loss": 0.669,
+      "step": 5323
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 0.00016101923076923075,
+      "loss": 0.7805,
+      "step": 5324
+    },
+    {
+      "epoch": 48.85,
+      "learning_rate": 0.00016099038461538461,
+      "loss": 0.9443,
+      "step": 5325
+    },
+    {
+      "epoch": 48.86,
+      "learning_rate": 0.00016096153846153845,
+      "loss": 0.6901,
+      "step": 5326
+    },
+    {
+      "epoch": 48.87,
+      "learning_rate": 0.0001609326923076923,
+      "loss": 0.7696,
+      "step": 5327
+    },
+    {
+      "epoch": 48.88,
+      "learning_rate": 0.00016090384615384613,
+      "loss": 0.6597,
+      "step": 5328
+    },
+    {
+      "epoch": 48.89,
+      "learning_rate": 0.000160875,
+      "loss": 0.8059,
+      "step": 5329
+    },
+    {
+      "epoch": 48.9,
+      "learning_rate": 0.00016084615384615386,
+      "loss": 0.7397,
+      "step": 5330
+    },
+    {
+      "epoch": 48.91,
+      "learning_rate": 0.00016081730769230767,
+      "loss": 1.0439,
+      "step": 5331
+    },
+    {
+      "epoch": 48.92,
+      "learning_rate": 0.00016078846153846154,
+      "loss": 0.7457,
+      "step": 5332
+    },
+    {
+      "epoch": 48.93,
+      "learning_rate": 0.00016075961538461535,
+      "loss": 0.7794,
+      "step": 5333
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 0.0001607307692307692,
+      "loss": 0.7322,
+      "step": 5334
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 0.00016070192307692308,
+      "loss": 0.7582,
+      "step": 5335
+    },
+    {
+      "epoch": 48.95,
+      "learning_rate": 0.00016067307692307691,
+      "loss": 0.7596,
+      "step": 5336
+    },
+    {
+      "epoch": 48.96,
+      "learning_rate": 0.00016064423076923075,
+      "loss": 0.74,
+      "step": 5337
+    },
+    {
+      "epoch": 48.97,
+      "learning_rate": 0.0001606153846153846,
+      "loss": 0.8091,
+      "step": 5338
+    },
+    {
+      "epoch": 48.98,
+      "learning_rate": 0.00016058653846153846,
+      "loss": 0.7991,
+      "step": 5339
+    },
+    {
+      "epoch": 48.99,
+      "learning_rate": 0.00016055769230769232,
+      "loss": 0.8433,
+      "step": 5340
+    },
+    {
+      "epoch": 49.0,
+      "learning_rate": 0.00016052884615384613,
+      "loss": 0.697,
+      "step": 5341
+    },
+    {
+      "epoch": 49.01,
+      "learning_rate": 0.0001605,
+      "loss": 0.8329,
+      "step": 5342
+    },
+    {
+      "epoch": 49.02,
+      "learning_rate": 0.0001604711538461538,
+      "loss": 0.7041,
+      "step": 5343
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 0.00016044230769230767,
+      "loss": 0.7596,
+      "step": 5344
+    },
+    {
+      "epoch": 49.04,
+      "learning_rate": 0.00016041346153846154,
+      "loss": 0.7465,
+      "step": 5345
+    },
+    {
+      "epoch": 49.05,
+      "learning_rate": 0.00016038461538461538,
+      "loss": 0.7405,
+      "step": 5346
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 0.00016035576923076922,
+      "loss": 0.7605,
+      "step": 5347
+    },
+    {
+      "epoch": 49.06,
+      "learning_rate": 0.00016032692307692305,
+      "loss": 0.7439,
+      "step": 5348
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 0.00016029807692307692,
+      "loss": 0.7755,
+      "step": 5349
+    },
+    {
+      "epoch": 49.08,
+      "learning_rate": 0.00016026923076923073,
+      "loss": 0.715,
+      "step": 5350
+    },
+    {
+      "epoch": 49.09,
+      "learning_rate": 0.0001602403846153846,
+      "loss": 0.7341,
+      "step": 5351
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 0.00016021153846153846,
+      "loss": 0.7298,
+      "step": 5352
+    },
+    {
+      "epoch": 49.11,
+      "learning_rate": 0.00016018269230769227,
+      "loss": 0.8048,
+      "step": 5353
+    },
+    {
+      "epoch": 49.12,
+      "learning_rate": 0.00016015384615384614,
+      "loss": 0.7718,
+      "step": 5354
+    },
+    {
+      "epoch": 49.13,
+      "learning_rate": 0.00016012499999999998,
+      "loss": 0.6794,
+      "step": 5355
+    },
+    {
+      "epoch": 49.14,
+      "learning_rate": 0.00016009615384615384,
+      "loss": 0.7563,
+      "step": 5356
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 0.00016006730769230768,
+      "loss": 0.6324,
+      "step": 5357
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 0.00016003846153846152,
+      "loss": 0.7325,
+      "step": 5358
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 0.00016000961538461538,
+      "loss": 0.6856,
+      "step": 5359
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 0.0001599807692307692,
+      "loss": 0.6126,
+      "step": 5360
+    },
+    {
+      "epoch": 49.18,
+      "learning_rate": 0.00015995192307692306,
+      "loss": 0.7657,
+      "step": 5361
+    },
+    {
+      "epoch": 49.19,
+      "learning_rate": 0.00015992307692307692,
+      "loss": 0.8944,
+      "step": 5362
+    },
+    {
+      "epoch": 49.2,
+      "learning_rate": 0.00015989423076923073,
+      "loss": 0.6526,
+      "step": 5363
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 0.0001598653846153846,
+      "loss": 0.709,
+      "step": 5364
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 0.00015983653846153844,
+      "loss": 0.6991,
+      "step": 5365
+    },
+    {
+      "epoch": 49.23,
+      "learning_rate": 0.0001598076923076923,
+      "loss": 0.8572,
+      "step": 5366
+    },
+    {
+      "epoch": 49.24,
+      "learning_rate": 0.00015977884615384614,
+      "loss": 0.8425,
+      "step": 5367
+    },
+    {
+      "epoch": 49.25,
+      "learning_rate": 0.00015974999999999998,
+      "loss": 0.6634,
+      "step": 5368
+    },
+    {
+      "epoch": 49.26,
+      "learning_rate": 0.00015972115384615385,
+      "loss": 0.664,
+      "step": 5369
+    },
+    {
+      "epoch": 49.27,
+      "learning_rate": 0.00015969230769230766,
+      "loss": 0.7784,
+      "step": 5370
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 0.00015966346153846152,
+      "loss": 0.7189,
+      "step": 5371
+    },
+    {
+      "epoch": 49.28,
+      "learning_rate": 0.0001596346153846154,
+      "loss": 0.7285,
+      "step": 5372
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 0.0001596057692307692,
+      "loss": 0.7937,
+      "step": 5373
+    },
+    {
+      "epoch": 49.3,
+      "learning_rate": 0.00015957692307692306,
+      "loss": 0.617,
+      "step": 5374
+    },
+    {
+      "epoch": 49.31,
+      "learning_rate": 0.0001595480769230769,
+      "loss": 0.593,
+      "step": 5375
+    },
+    {
+      "epoch": 49.32,
+      "learning_rate": 0.00015951923076923077,
+      "loss": 0.6989,
+      "step": 5376
+    },
+    {
+      "epoch": 49.33,
+      "learning_rate": 0.0001594903846153846,
+      "loss": 0.845,
+      "step": 5377
+    },
+    {
+      "epoch": 49.34,
+      "learning_rate": 0.00015946153846153844,
+      "loss": 0.7435,
+      "step": 5378
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 0.0001594326923076923,
+      "loss": 0.7412,
+      "step": 5379
+    },
+    {
+      "epoch": 49.36,
+      "learning_rate": 0.00015940384615384612,
+      "loss": 0.661,
+      "step": 5380
+    },
+    {
+      "epoch": 49.37,
+      "learning_rate": 0.00015937499999999998,
+      "loss": 0.8003,
+      "step": 5381
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 0.00015934615384615385,
+      "loss": 0.8846,
+      "step": 5382
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 0.00015931730769230766,
+      "loss": 0.8178,
+      "step": 5383
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 0.00015928846153846153,
+      "loss": 0.8192,
+      "step": 5384
+    },
+    {
+      "epoch": 49.4,
+      "learning_rate": 0.00015925961538461536,
+      "loss": 1.0162,
+      "step": 5385
+    },
+    {
+      "epoch": 49.41,
+      "learning_rate": 0.00015923076923076923,
+      "loss": 0.7117,
+      "step": 5386
+    },
+    {
+      "epoch": 49.42,
+      "learning_rate": 0.00015920192307692307,
+      "loss": 0.8386,
+      "step": 5387
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 0.0001591730769230769,
+      "loss": 0.7483,
+      "step": 5388
+    },
+    {
+      "epoch": 49.44,
+      "learning_rate": 0.00015914423076923077,
+      "loss": 0.7708,
+      "step": 5389
+    },
+    {
+      "epoch": 49.45,
+      "learning_rate": 0.00015911538461538458,
+      "loss": 0.7461,
+      "step": 5390
+    },
+    {
+      "epoch": 49.46,
+      "learning_rate": 0.00015908653846153845,
+      "loss": 0.6555,
+      "step": 5391
+    },
+    {
+      "epoch": 49.47,
+      "learning_rate": 0.0001590576923076923,
+      "loss": 0.7256,
+      "step": 5392
+    },
+    {
+      "epoch": 49.48,
+      "learning_rate": 0.00015902884615384612,
+      "loss": 0.8205,
+      "step": 5393
+    },
+    {
+      "epoch": 49.49,
+      "learning_rate": 0.000159,
+      "loss": 0.7895,
+      "step": 5394
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 0.00015897115384615383,
+      "loss": 0.7215,
+      "step": 5395
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 0.0001589423076923077,
+      "loss": 0.7669,
+      "step": 5396
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 0.00015891346153846153,
+      "loss": 0.7468,
+      "step": 5397
+    },
+    {
+      "epoch": 49.52,
+      "learning_rate": 0.00015888461538461537,
+      "loss": 0.6327,
+      "step": 5398
+    },
+    {
+      "epoch": 49.53,
+      "learning_rate": 0.00015885576923076923,
+      "loss": 0.6658,
+      "step": 5399
+    },
+    {
+      "epoch": 49.54,
+      "learning_rate": 0.00015882692307692304,
+      "loss": 0.7754,
+      "step": 5400
+    },
+    {
+      "epoch": 49.55,
+      "learning_rate": 0.0001587980769230769,
+      "loss": 0.6803,
+      "step": 5401
+    },
+    {
+      "epoch": 49.56,
+      "learning_rate": 0.00015876923076923078,
+      "loss": 0.7391,
+      "step": 5402
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 0.00015874038461538459,
+      "loss": 0.7719,
+      "step": 5403
+    },
+    {
+      "epoch": 49.58,
+      "learning_rate": 0.00015871153846153845,
+      "loss": 0.8112,
+      "step": 5404
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 0.0001586826923076923,
+      "loss": 0.8292,
+      "step": 5405
+    },
+    {
+      "epoch": 49.6,
+      "learning_rate": 0.00015865384615384616,
+      "loss": 0.7514,
+      "step": 5406
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 0.000158625,
+      "loss": 0.8663,
+      "step": 5407
+    },
+    {
+      "epoch": 49.61,
+      "learning_rate": 0.00015859615384615383,
+      "loss": 0.7147,
+      "step": 5408
+    },
+    {
+      "epoch": 49.62,
+      "learning_rate": 0.0001585673076923077,
+      "loss": 0.7838,
+      "step": 5409
+    },
+    {
+      "epoch": 49.63,
+      "learning_rate": 0.0001585384615384615,
+      "loss": 0.8237,
+      "step": 5410
+    },
+    {
+      "epoch": 49.64,
+      "learning_rate": 0.00015850961538461537,
+      "loss": 0.6777,
+      "step": 5411
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 0.00015848076923076924,
+      "loss": 0.6978,
+      "step": 5412
+    },
+    {
+      "epoch": 49.66,
+      "learning_rate": 0.00015845192307692305,
+      "loss": 0.7888,
+      "step": 5413
+    },
+    {
+      "epoch": 49.67,
+      "learning_rate": 0.00015842307692307691,
+      "loss": 0.8403,
+      "step": 5414
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 0.00015839423076923075,
+      "loss": 0.7423,
+      "step": 5415
+    },
+    {
+      "epoch": 49.69,
+      "learning_rate": 0.00015836538461538462,
+      "loss": 0.811,
+      "step": 5416
+    },
+    {
+      "epoch": 49.7,
+      "learning_rate": 0.00015833653846153843,
+      "loss": 0.8522,
+      "step": 5417
+    },
+    {
+      "epoch": 49.71,
+      "learning_rate": 0.0001583076923076923,
+      "loss": 0.8408,
+      "step": 5418
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 0.00015827884615384616,
+      "loss": 0.7953,
+      "step": 5419
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 0.00015824999999999997,
+      "loss": 0.6165,
+      "step": 5420
+    },
+    {
+      "epoch": 49.73,
+      "learning_rate": 0.00015822115384615384,
+      "loss": 0.7146,
+      "step": 5421
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 0.00015819230769230767,
+      "loss": 0.7892,
+      "step": 5422
+    },
+    {
+      "epoch": 49.75,
+      "learning_rate": 0.0001581634615384615,
+      "loss": 0.7777,
+      "step": 5423
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 0.00015813461538461538,
+      "loss": 0.8023,
+      "step": 5424
+    },
+    {
+      "epoch": 49.77,
+      "learning_rate": 0.00015810576923076922,
+      "loss": 0.7463,
+      "step": 5425
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 0.00015807692307692308,
+      "loss": 0.7799,
+      "step": 5426
+    },
+    {
+      "epoch": 49.79,
+      "learning_rate": 0.0001580480769230769,
+      "loss": 0.701,
+      "step": 5427
+    },
+    {
+      "epoch": 49.8,
+      "learning_rate": 0.00015801923076923076,
+      "loss": 0.7952,
+      "step": 5428
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 0.00015799038461538462,
+      "loss": 0.8683,
+      "step": 5429
+    },
+    {
+      "epoch": 49.82,
+      "learning_rate": 0.00015796153846153843,
+      "loss": 0.7263,
+      "step": 5430
+    },
+    {
+      "epoch": 49.83,
+      "learning_rate": 0.0001579326923076923,
+      "loss": 0.6823,
+      "step": 5431
+    },
+    {
+      "epoch": 49.83,
+      "learning_rate": 0.00015790384615384614,
+      "loss": 0.7854,
+      "step": 5432
+    },
+    {
+      "epoch": 49.84,
+      "learning_rate": 0.00015787499999999997,
+      "loss": 0.7582,
+      "step": 5433
+    },
+    {
+      "epoch": 49.85,
+      "learning_rate": 0.00015784615384615384,
+      "loss": 0.7304,
+      "step": 5434
+    },
+    {
+      "epoch": 49.86,
+      "learning_rate": 0.00015781730769230768,
+      "loss": 0.7715,
+      "step": 5435
+    },
+    {
+      "epoch": 49.87,
+      "learning_rate": 0.00015778846153846154,
+      "loss": 0.8492,
+      "step": 5436
+    },
+    {
+      "epoch": 49.88,
+      "learning_rate": 0.00015775961538461535,
+      "loss": 0.7683,
+      "step": 5437
+    },
+    {
+      "epoch": 49.89,
+      "learning_rate": 0.00015773076923076922,
+      "loss": 0.8053,
+      "step": 5438
+    },
+    {
+      "epoch": 49.9,
+      "learning_rate": 0.00015770192307692309,
+      "loss": 0.7513,
+      "step": 5439
+    },
+    {
+      "epoch": 49.91,
+      "learning_rate": 0.0001576730769230769,
+      "loss": 0.7551,
+      "step": 5440
+    },
+    {
+      "epoch": 49.92,
+      "learning_rate": 0.00015764423076923076,
+      "loss": 0.7701,
+      "step": 5441
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 0.0001576153846153846,
+      "loss": 0.8417,
+      "step": 5442
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 0.00015758653846153844,
+      "loss": 0.7045,
+      "step": 5443
+    },
+    {
+      "epoch": 49.94,
+      "learning_rate": 0.0001575576923076923,
+      "loss": 0.849,
+      "step": 5444
+    },
+    {
+      "epoch": 49.95,
+      "learning_rate": 0.00015752884615384614,
+      "loss": 0.6919,
+      "step": 5445
+    },
+    {
+      "epoch": 49.96,
+      "learning_rate": 0.00015749999999999998,
+      "loss": 0.8337,
+      "step": 5446
+    },
+    {
+      "epoch": 49.97,
+      "learning_rate": 0.00015747115384615382,
+      "loss": 0.7014,
+      "step": 5447
+    },
+    {
+      "epoch": 49.98,
+      "learning_rate": 0.00015744230769230768,
+      "loss": 0.6753,
+      "step": 5448
+    },
+    {
+      "epoch": 49.99,
+      "learning_rate": 0.00015741346153846155,
+      "loss": 0.7791,
+      "step": 5449
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 0.00015738461538461536,
+      "loss": 0.8848,
+      "step": 5450
+    },
+    {
+      "epoch": 50.01,
+      "learning_rate": 0.00015735576923076922,
+      "loss": 0.8139,
+      "step": 5451
+    },
+    {
+      "epoch": 50.02,
+      "learning_rate": 0.00015732692307692306,
+      "loss": 0.6224,
+      "step": 5452
+    },
+    {
+      "epoch": 50.03,
+      "learning_rate": 0.0001572980769230769,
+      "loss": 0.6626,
+      "step": 5453
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 0.00015726923076923077,
+      "loss": 0.7677,
+      "step": 5454
+    },
+    {
+      "epoch": 50.05,
+      "learning_rate": 0.0001572403846153846,
+      "loss": 0.79,
+      "step": 5455
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 0.00015721153846153844,
+      "loss": 0.7129,
+      "step": 5456
+    },
+    {
+      "epoch": 50.06,
+      "learning_rate": 0.00015718269230769228,
+      "loss": 0.7511,
+      "step": 5457
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 0.00015715384615384615,
+      "loss": 0.716,
+      "step": 5458
+    },
+    {
+      "epoch": 50.08,
+      "learning_rate": 0.000157125,
+      "loss": 0.7071,
+      "step": 5459
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 0.00015709615384615382,
+      "loss": 0.6936,
+      "step": 5460
+    },
+    {
+      "epoch": 50.1,
+      "learning_rate": 0.0001570673076923077,
+      "loss": 0.7168,
+      "step": 5461
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 0.00015703846153846153,
+      "loss": 0.6846,
+      "step": 5462
+    },
+    {
+      "epoch": 50.12,
+      "learning_rate": 0.00015700961538461536,
+      "loss": 0.7639,
+      "step": 5463
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 0.00015698076923076923,
+      "loss": 0.9129,
+      "step": 5464
+    },
+    {
+      "epoch": 50.14,
+      "learning_rate": 0.00015695192307692307,
+      "loss": 0.7945,
+      "step": 5465
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 0.0001569230769230769,
+      "loss": 0.6784,
+      "step": 5466
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 0.00015689423076923074,
+      "loss": 0.7392,
+      "step": 5467
+    },
+    {
+      "epoch": 50.17,
+      "learning_rate": 0.0001568653846153846,
+      "loss": 0.7808,
+      "step": 5468
+    },
+    {
+      "epoch": 50.17,
+      "learning_rate": 0.00015683653846153847,
+      "loss": 0.9056,
+      "step": 5469
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 0.00015680769230769228,
+      "loss": 0.6253,
+      "step": 5470
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 0.00015677884615384615,
+      "loss": 0.6894,
+      "step": 5471
+    },
+    {
+      "epoch": 50.2,
+      "learning_rate": 0.00015675,
+      "loss": 0.612,
+      "step": 5472
+    },
+    {
+      "epoch": 50.21,
+      "learning_rate": 0.00015672115384615383,
+      "loss": 0.8367,
+      "step": 5473
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 0.0001566923076923077,
+      "loss": 0.8016,
+      "step": 5474
+    },
+    {
+      "epoch": 50.23,
+      "learning_rate": 0.00015666346153846153,
+      "loss": 0.718,
+      "step": 5475
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 0.00015663461538461537,
+      "loss": 0.7958,
+      "step": 5476
+    },
+    {
+      "epoch": 50.25,
+      "learning_rate": 0.0001566057692307692,
+      "loss": 0.8998,
+      "step": 5477
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 0.00015657692307692307,
+      "loss": 0.6612,
+      "step": 5478
+    },
+    {
+      "epoch": 50.27,
+      "learning_rate": 0.00015654807692307688,
+      "loss": 0.8196,
+      "step": 5479
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 0.00015651923076923075,
+      "loss": 0.7258,
+      "step": 5480
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 0.0001564903846153846,
+      "loss": 0.7364,
+      "step": 5481
+    },
+    {
+      "epoch": 50.29,
+      "learning_rate": 0.00015646153846153845,
+      "loss": 0.7564,
+      "step": 5482
+    },
+    {
+      "epoch": 50.3,
+      "learning_rate": 0.0001564326923076923,
+      "loss": 0.7333,
+      "step": 5483
+    },
+    {
+      "epoch": 50.31,
+      "learning_rate": 0.00015640384615384613,
+      "loss": 0.6898,
+      "step": 5484
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 0.000156375,
+      "loss": 0.7596,
+      "step": 5485
+    },
+    {
+      "epoch": 50.33,
+      "learning_rate": 0.00015634615384615383,
+      "loss": 0.7557,
+      "step": 5486
+    },
+    {
+      "epoch": 50.34,
+      "learning_rate": 0.00015631730769230767,
+      "loss": 0.7127,
+      "step": 5487
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 0.00015628846153846153,
+      "loss": 0.6478,
+      "step": 5488
+    },
+    {
+      "epoch": 50.36,
+      "learning_rate": 0.00015625961538461535,
+      "loss": 0.7162,
+      "step": 5489
+    },
+    {
+      "epoch": 50.37,
+      "learning_rate": 0.0001562307692307692,
+      "loss": 0.7212,
+      "step": 5490
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 0.00015620192307692308,
+      "loss": 0.7369,
+      "step": 5491
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 0.00015617307692307691,
+      "loss": 0.8534,
+      "step": 5492
+    },
+    {
+      "epoch": 50.39,
+      "learning_rate": 0.00015614423076923075,
+      "loss": 0.7183,
+      "step": 5493
+    },
+    {
+      "epoch": 50.4,
+      "learning_rate": 0.0001561153846153846,
+      "loss": 0.7712,
+      "step": 5494
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 0.00015608653846153846,
+      "loss": 0.7865,
+      "step": 5495
+    },
+    {
+      "epoch": 50.42,
+      "learning_rate": 0.0001560576923076923,
+      "loss": 0.6723,
+      "step": 5496
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 0.00015602884615384613,
+      "loss": 0.7891,
+      "step": 5497
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 0.000156,
+      "loss": 0.7588,
+      "step": 5498
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 0.0001559711538461538,
+      "loss": 0.7779,
+      "step": 5499
+    },
+    {
+      "epoch": 50.46,
+      "learning_rate": 0.00015594230769230767,
+      "loss": 0.8194,
+      "step": 5500
+    },
+    {
+      "epoch": 50.46,
+      "eval_cer": 0.11796278804779682,
+      "eval_loss": 0.37289363145828247,
+      "eval_runtime": 13.8331,
+      "eval_samples_per_second": 119.062,
+      "eval_steps_per_second": 1.88,
+      "eval_wer": 0.46032070268614034,
+      "step": 5500
+    },
+    {
+      "epoch": 50.47,
+      "learning_rate": 0.00015591346153846154,
+      "loss": 0.6899,
+      "step": 5501
+    },
+    {
+      "epoch": 50.48,
+      "learning_rate": 0.00015588461538461538,
+      "loss": 0.7828,
+      "step": 5502
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 0.00015585576923076922,
+      "loss": 0.6959,
+      "step": 5503
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 0.00015582692307692305,
+      "loss": 0.8391,
+      "step": 5504
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 0.00015579807692307692,
+      "loss": 0.8378,
+      "step": 5505
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 0.00015576923076923076,
+      "loss": 0.8523,
+      "step": 5506
+    },
+    {
+      "epoch": 50.52,
+      "learning_rate": 0.0001557403846153846,
+      "loss": 0.6957,
+      "step": 5507
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 0.00015571153846153846,
+      "loss": 0.8318,
+      "step": 5508
+    },
+    {
+      "epoch": 50.54,
+      "learning_rate": 0.00015568269230769227,
+      "loss": 0.7389,
+      "step": 5509
+    },
+    {
+      "epoch": 50.55,
+      "learning_rate": 0.00015565384615384614,
+      "loss": 0.8396,
+      "step": 5510
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 0.000155625,
+      "loss": 0.8761,
+      "step": 5511
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 0.0001555961538461538,
+      "loss": 0.8095,
+      "step": 5512
+    },
+    {
+      "epoch": 50.58,
+      "learning_rate": 0.00015556730769230768,
+      "loss": 0.8699,
+      "step": 5513
+    },
+    {
+      "epoch": 50.59,
+      "learning_rate": 0.00015553846153846152,
+      "loss": 0.808,
+      "step": 5514
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 0.00015550961538461538,
+      "loss": 0.6719,
+      "step": 5515
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 0.00015548076923076922,
+      "loss": 0.761,
+      "step": 5516
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 0.00015545192307692306,
+      "loss": 0.7795,
+      "step": 5517
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 0.00015542307692307692,
+      "loss": 0.8719,
+      "step": 5518
+    },
+    {
+      "epoch": 50.63,
+      "learning_rate": 0.00015539423076923073,
+      "loss": 0.6874,
+      "step": 5519
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 0.0001553653846153846,
+      "loss": 0.8274,
+      "step": 5520
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 0.00015533653846153846,
+      "loss": 0.7504,
+      "step": 5521
+    },
+    {
+      "epoch": 50.66,
+      "learning_rate": 0.00015530769230769228,
+      "loss": 0.6291,
+      "step": 5522
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 0.00015527884615384614,
+      "loss": 0.6841,
+      "step": 5523
+    },
+    {
+      "epoch": 50.68,
+      "learning_rate": 0.00015524999999999998,
+      "loss": 0.8822,
+      "step": 5524
+    },
+    {
+      "epoch": 50.69,
+      "learning_rate": 0.00015522115384615384,
+      "loss": 0.7393,
+      "step": 5525
+    },
+    {
+      "epoch": 50.7,
+      "learning_rate": 0.00015519230769230768,
+      "loss": 0.7458,
+      "step": 5526
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 0.00015516346153846152,
+      "loss": 0.6792,
+      "step": 5527
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 0.00015513461538461539,
+      "loss": 0.8458,
+      "step": 5528
+    },
+    {
+      "epoch": 50.72,
+      "learning_rate": 0.0001551057692307692,
+      "loss": 0.7248,
+      "step": 5529
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 0.00015507692307692306,
+      "loss": 0.8916,
+      "step": 5530
+    },
+    {
+      "epoch": 50.74,
+      "learning_rate": 0.00015504807692307693,
+      "loss": 0.7036,
+      "step": 5531
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 0.00015501923076923074,
+      "loss": 0.7527,
+      "step": 5532
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 0.0001549903846153846,
+      "loss": 0.7498,
+      "step": 5533
+    },
+    {
+      "epoch": 50.77,
+      "learning_rate": 0.00015496153846153844,
+      "loss": 0.8102,
+      "step": 5534
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 0.0001549326923076923,
+      "loss": 0.7146,
+      "step": 5535
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 0.00015490384615384615,
+      "loss": 0.7188,
+      "step": 5536
+    },
+    {
+      "epoch": 50.8,
+      "learning_rate": 0.00015487499999999998,
+      "loss": 0.7268,
+      "step": 5537
+    },
+    {
+      "epoch": 50.81,
+      "learning_rate": 0.00015484615384615385,
+      "loss": 0.7322,
+      "step": 5538
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 0.00015481730769230766,
+      "loss": 0.7417,
+      "step": 5539
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 0.00015478846153846152,
+      "loss": 0.7703,
+      "step": 5540
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 0.0001547596153846154,
+      "loss": 0.8078,
+      "step": 5541
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 0.0001547307692307692,
+      "loss": 0.7502,
+      "step": 5542
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 0.00015470192307692307,
+      "loss": 0.6942,
+      "step": 5543
+    },
+    {
+      "epoch": 50.86,
+      "learning_rate": 0.0001546730769230769,
+      "loss": 0.7175,
+      "step": 5544
+    },
+    {
+      "epoch": 50.87,
+      "learning_rate": 0.00015464423076923077,
+      "loss": 0.6766,
+      "step": 5545
+    },
+    {
+      "epoch": 50.88,
+      "learning_rate": 0.00015461538461538458,
+      "loss": 0.6171,
+      "step": 5546
+    },
+    {
+      "epoch": 50.89,
+      "learning_rate": 0.00015458653846153845,
+      "loss": 0.796,
+      "step": 5547
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 0.0001545576923076923,
+      "loss": 0.8995,
+      "step": 5548
+    },
+    {
+      "epoch": 50.91,
+      "learning_rate": 0.00015452884615384612,
+      "loss": 0.7952,
+      "step": 5549
+    },
+    {
+      "epoch": 50.92,
+      "learning_rate": 0.0001545,
+      "loss": 0.7845,
+      "step": 5550
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 0.00015447115384615383,
+      "loss": 0.8423,
+      "step": 5551
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 0.00015444230769230766,
+      "loss": 0.8366,
+      "step": 5552
+    },
+    {
+      "epoch": 50.94,
+      "learning_rate": 0.00015441346153846153,
+      "loss": 0.7117,
+      "step": 5553
+    },
+    {
+      "epoch": 50.95,
+      "learning_rate": 0.00015438461538461537,
+      "loss": 0.8352,
+      "step": 5554
+    },
+    {
+      "epoch": 50.96,
+      "learning_rate": 0.00015435576923076923,
+      "loss": 0.8352,
+      "step": 5555
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 0.00015432692307692304,
+      "loss": 0.7195,
+      "step": 5556
+    },
+    {
+      "epoch": 50.98,
+      "learning_rate": 0.0001542980769230769,
+      "loss": 0.7076,
+      "step": 5557
+    },
+    {
+      "epoch": 50.99,
+      "learning_rate": 0.00015426923076923077,
+      "loss": 0.7592,
+      "step": 5558
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 0.00015424038461538459,
+      "loss": 0.7708,
+      "step": 5559
+    },
+    {
+      "epoch": 51.01,
+      "learning_rate": 0.00015421153846153845,
+      "loss": 0.7738,
+      "step": 5560
+    },
+    {
+      "epoch": 51.02,
+      "learning_rate": 0.0001541826923076923,
+      "loss": 0.8917,
+      "step": 5561
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 0.00015415384615384613,
+      "loss": 0.6658,
+      "step": 5562
+    },
+    {
+      "epoch": 51.04,
+      "learning_rate": 0.000154125,
+      "loss": 0.7148,
+      "step": 5563
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 0.00015409615384615383,
+      "loss": 0.8684,
+      "step": 5564
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 0.0001540673076923077,
+      "loss": 0.8288,
+      "step": 5565
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 0.0001540384615384615,
+      "loss": 0.6942,
+      "step": 5566
+    },
+    {
+      "epoch": 51.07,
+      "learning_rate": 0.00015400961538461537,
+      "loss": 0.791,
+      "step": 5567
+    },
+    {
+      "epoch": 51.08,
+      "learning_rate": 0.00015398076923076924,
+      "loss": 0.7722,
+      "step": 5568
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 0.00015395192307692305,
+      "loss": 0.7225,
+      "step": 5569
+    },
+    {
+      "epoch": 51.1,
+      "learning_rate": 0.00015392307692307691,
+      "loss": 0.9038,
+      "step": 5570
+    },
+    {
+      "epoch": 51.11,
+      "learning_rate": 0.00015389423076923075,
+      "loss": 0.7109,
+      "step": 5571
+    },
+    {
+      "epoch": 51.12,
+      "learning_rate": 0.0001538653846153846,
+      "loss": 0.8749,
+      "step": 5572
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 0.00015383653846153846,
+      "loss": 0.7241,
+      "step": 5573
+    },
+    {
+      "epoch": 51.14,
+      "learning_rate": 0.0001538076923076923,
+      "loss": 0.8125,
+      "step": 5574
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 0.00015377884615384616,
+      "loss": 0.7322,
+      "step": 5575
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 0.00015374999999999997,
+      "loss": 0.7132,
+      "step": 5576
+    },
+    {
+      "epoch": 51.17,
+      "learning_rate": 0.00015372115384615383,
+      "loss": 0.7316,
+      "step": 5577
+    },
+    {
+      "epoch": 51.17,
+      "learning_rate": 0.0001536923076923077,
+      "loss": 0.7368,
+      "step": 5578
+    },
+    {
+      "epoch": 51.18,
+      "learning_rate": 0.0001536634615384615,
+      "loss": 0.7533,
+      "step": 5579
+    },
+    {
+      "epoch": 51.19,
+      "learning_rate": 0.00015363461538461538,
+      "loss": 0.7691,
+      "step": 5580
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 0.00015360576923076921,
+      "loss": 0.7079,
+      "step": 5581
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 0.00015357692307692305,
+      "loss": 0.7035,
+      "step": 5582
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 0.00015354807692307692,
+      "loss": 0.7053,
+      "step": 5583
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 0.00015351923076923076,
+      "loss": 0.8122,
+      "step": 5584
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 0.00015349038461538462,
+      "loss": 0.7878,
+      "step": 5585
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 0.00015346153846153843,
+      "loss": 0.7303,
+      "step": 5586
+    },
+    {
+      "epoch": 51.26,
+      "learning_rate": 0.0001534326923076923,
+      "loss": 0.7721,
+      "step": 5587
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 0.00015340384615384616,
+      "loss": 0.7977,
+      "step": 5588
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 0.00015337499999999997,
+      "loss": 0.7629,
+      "step": 5589
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 0.00015334615384615384,
+      "loss": 0.7148,
+      "step": 5590
+    },
+    {
+      "epoch": 51.29,
+      "learning_rate": 0.00015331730769230768,
+      "loss": 0.7853,
+      "step": 5591
+    },
+    {
+      "epoch": 51.3,
+      "learning_rate": 0.00015328846153846152,
+      "loss": 0.759,
+      "step": 5592
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 0.00015325961538461538,
+      "loss": 0.7502,
+      "step": 5593
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 0.00015323076923076922,
+      "loss": 0.7762,
+      "step": 5594
+    },
+    {
+      "epoch": 51.33,
+      "learning_rate": 0.00015320192307692308,
+      "loss": 0.9672,
+      "step": 5595
+    },
+    {
+      "epoch": 51.34,
+      "learning_rate": 0.0001531730769230769,
+      "loss": 0.7363,
+      "step": 5596
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 0.00015314423076923076,
+      "loss": 0.713,
+      "step": 5597
+    },
+    {
+      "epoch": 51.36,
+      "learning_rate": 0.00015311538461538463,
+      "loss": 0.6868,
+      "step": 5598
+    },
+    {
+      "epoch": 51.37,
+      "learning_rate": 0.00015308653846153844,
+      "loss": 0.7626,
+      "step": 5599
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 0.0001530576923076923,
+      "loss": 0.6997,
+      "step": 5600
+    },
+    {
+      "epoch": 51.39,
+      "learning_rate": 0.00015302884615384614,
+      "loss": 0.7399,
+      "step": 5601
+    },
+    {
+      "epoch": 51.39,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 0.8608,
+      "step": 5602
+    },
+    {
+      "epoch": 51.4,
+      "learning_rate": 0.00015297115384615384,
+      "loss": 0.7214,
+      "step": 5603
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 0.00015294230769230768,
+      "loss": 0.7598,
+      "step": 5604
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 0.00015291346153846155,
+      "loss": 0.8411,
+      "step": 5605
+    },
+    {
+      "epoch": 51.43,
+      "learning_rate": 0.00015288461538461536,
+      "loss": 0.7583,
+      "step": 5606
+    },
+    {
+      "epoch": 51.44,
+      "learning_rate": 0.00015285576923076922,
+      "loss": 0.7706,
+      "step": 5607
+    },
+    {
+      "epoch": 51.45,
+      "learning_rate": 0.0001528269230769231,
+      "loss": 0.6462,
+      "step": 5608
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 0.0001527980769230769,
+      "loss": 0.757,
+      "step": 5609
+    },
+    {
+      "epoch": 51.47,
+      "learning_rate": 0.00015276923076923077,
+      "loss": 0.609,
+      "step": 5610
+    },
+    {
+      "epoch": 51.48,
+      "learning_rate": 0.0001527403846153846,
+      "loss": 0.7273,
+      "step": 5611
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 0.00015271153846153844,
+      "loss": 0.8021,
+      "step": 5612
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 0.00015268269230769228,
+      "loss": 0.7853,
+      "step": 5613
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 0.00015265384615384614,
+      "loss": 0.8646,
+      "step": 5614
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 0.000152625,
+      "loss": 0.7091,
+      "step": 5615
+    },
+    {
+      "epoch": 51.52,
+      "learning_rate": 0.00015259615384615382,
+      "loss": 0.7175,
+      "step": 5616
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 0.00015256730769230769,
+      "loss": 0.6607,
+      "step": 5617
+    },
+    {
+      "epoch": 51.54,
+      "learning_rate": 0.0001525384615384615,
+      "loss": 0.8678,
+      "step": 5618
+    },
+    {
+      "epoch": 51.55,
+      "learning_rate": 0.00015250961538461536,
+      "loss": 0.7673,
+      "step": 5619
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 0.00015248076923076923,
+      "loss": 0.7681,
+      "step": 5620
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 0.00015245192307692307,
+      "loss": 0.7489,
+      "step": 5621
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 0.0001524230769230769,
+      "loss": 0.7817,
+      "step": 5622
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 0.00015239423076923074,
+      "loss": 0.7686,
+      "step": 5623
+    },
+    {
+      "epoch": 51.6,
+      "learning_rate": 0.0001523653846153846,
+      "loss": 0.6695,
+      "step": 5624
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 0.00015233653846153847,
+      "loss": 0.6106,
+      "step": 5625
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 0.00015230769230769228,
+      "loss": 0.7492,
+      "step": 5626
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 0.00015227884615384615,
+      "loss": 0.7683,
+      "step": 5627
+    },
+    {
+      "epoch": 51.63,
+      "learning_rate": 0.00015224999999999996,
+      "loss": 0.704,
+      "step": 5628
+    },
+    {
+      "epoch": 51.64,
+      "learning_rate": 0.00015222115384615383,
+      "loss": 0.7219,
+      "step": 5629
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 0.0001521923076923077,
+      "loss": 0.7186,
+      "step": 5630
+    },
+    {
+      "epoch": 51.66,
+      "learning_rate": 0.00015216346153846153,
+      "loss": 0.7456,
+      "step": 5631
+    },
+    {
+      "epoch": 51.67,
+      "learning_rate": 0.00015213461538461537,
+      "loss": 0.7596,
+      "step": 5632
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 0.0001521057692307692,
+      "loss": 0.6886,
+      "step": 5633
+    },
+    {
+      "epoch": 51.69,
+      "learning_rate": 0.00015207692307692307,
+      "loss": 0.728,
+      "step": 5634
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 0.00015204807692307694,
+      "loss": 0.7373,
+      "step": 5635
+    },
+    {
+      "epoch": 51.71,
+      "learning_rate": 0.00015201923076923075,
+      "loss": 0.7391,
+      "step": 5636
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 0.0001519903846153846,
+      "loss": 0.7973,
+      "step": 5637
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 0.00015196153846153842,
+      "loss": 0.6823,
+      "step": 5638
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 0.0001519326923076923,
+      "loss": 0.7984,
+      "step": 5639
+    },
+    {
+      "epoch": 51.74,
+      "learning_rate": 0.00015190384615384615,
+      "loss": 0.7277,
+      "step": 5640
+    },
+    {
+      "epoch": 51.75,
+      "learning_rate": 0.000151875,
+      "loss": 0.8809,
+      "step": 5641
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 0.00015184615384615383,
+      "loss": 0.7386,
+      "step": 5642
+    },
+    {
+      "epoch": 51.77,
+      "learning_rate": 0.00015181730769230767,
+      "loss": 0.8567,
+      "step": 5643
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 0.00015178846153846153,
+      "loss": 0.8371,
+      "step": 5644
+    },
+    {
+      "epoch": 51.79,
+      "learning_rate": 0.00015175961538461537,
+      "loss": 0.8845,
+      "step": 5645
+    },
+    {
+      "epoch": 51.8,
+      "learning_rate": 0.0001517307692307692,
+      "loss": 0.7846,
+      "step": 5646
+    },
+    {
+      "epoch": 51.81,
+      "learning_rate": 0.00015170192307692307,
+      "loss": 0.7736,
+      "step": 5647
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 0.00015167307692307689,
+      "loss": 0.6836,
+      "step": 5648
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 0.00015164423076923075,
+      "loss": 0.8167,
+      "step": 5649
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 0.00015161538461538462,
+      "loss": 0.824,
+      "step": 5650
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 0.00015158653846153845,
+      "loss": 0.7531,
+      "step": 5651
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 0.0001515576923076923,
+      "loss": 0.8154,
+      "step": 5652
+    },
+    {
+      "epoch": 51.86,
+      "learning_rate": 0.00015152884615384613,
+      "loss": 0.8224,
+      "step": 5653
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 0.0001515,
+      "loss": 0.7716,
+      "step": 5654
+    },
+    {
+      "epoch": 51.88,
+      "learning_rate": 0.00015147115384615383,
+      "loss": 0.7037,
+      "step": 5655
+    },
+    {
+      "epoch": 51.89,
+      "learning_rate": 0.00015144230769230767,
+      "loss": 0.7474,
+      "step": 5656
+    },
+    {
+      "epoch": 51.9,
+      "learning_rate": 0.00015141346153846154,
+      "loss": 0.8344,
+      "step": 5657
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 0.00015138461538461535,
+      "loss": 0.7821,
+      "step": 5658
+    },
+    {
+      "epoch": 51.92,
+      "learning_rate": 0.00015135576923076921,
+      "loss": 0.6258,
+      "step": 5659
+    },
+    {
+      "epoch": 51.93,
+      "learning_rate": 0.00015132692307692308,
+      "loss": 0.9157,
+      "step": 5660
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 0.00015129807692307692,
+      "loss": 0.6761,
+      "step": 5661
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 0.00015126923076923076,
+      "loss": 0.7171,
+      "step": 5662
+    },
+    {
+      "epoch": 51.95,
+      "learning_rate": 0.0001512403846153846,
+      "loss": 0.7339,
+      "step": 5663
+    },
+    {
+      "epoch": 51.96,
+      "learning_rate": 0.00015121153846153846,
+      "loss": 0.6538,
+      "step": 5664
+    },
+    {
+      "epoch": 51.97,
+      "learning_rate": 0.0001511826923076923,
+      "loss": 0.6225,
+      "step": 5665
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 0.00015115384615384614,
+      "loss": 0.6871,
+      "step": 5666
+    },
+    {
+      "epoch": 51.99,
+      "learning_rate": 0.000151125,
+      "loss": 0.8705,
+      "step": 5667
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 0.0001510961538461538,
+      "loss": 0.7655,
+      "step": 5668
+    },
+    {
+      "epoch": 52.01,
+      "learning_rate": 0.00015106730769230768,
+      "loss": 0.8422,
+      "step": 5669
+    },
+    {
+      "epoch": 52.02,
+      "learning_rate": 0.00015103846153846154,
+      "loss": 0.6608,
+      "step": 5670
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 0.00015100961538461538,
+      "loss": 0.8238,
+      "step": 5671
+    },
+    {
+      "epoch": 52.04,
+      "learning_rate": 0.00015098076923076922,
+      "loss": 0.8493,
+      "step": 5672
+    },
+    {
+      "epoch": 52.05,
+      "learning_rate": 0.00015095192307692306,
+      "loss": 0.7532,
+      "step": 5673
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 0.00015092307692307692,
+      "loss": 0.6434,
+      "step": 5674
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 0.00015089423076923073,
+      "loss": 0.7368,
+      "step": 5675
+    },
+    {
+      "epoch": 52.07,
+      "learning_rate": 0.0001508653846153846,
+      "loss": 0.7566,
+      "step": 5676
+    },
+    {
+      "epoch": 52.08,
+      "learning_rate": 0.00015083653846153846,
+      "loss": 0.6869,
+      "step": 5677
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 0.00015080769230769227,
+      "loss": 1.0038,
+      "step": 5678
+    },
+    {
+      "epoch": 52.1,
+      "learning_rate": 0.00015077884615384614,
+      "loss": 0.9043,
+      "step": 5679
+    },
+    {
+      "epoch": 52.11,
+      "learning_rate": 0.00015074999999999998,
+      "loss": 0.7614,
+      "step": 5680
+    },
+    {
+      "epoch": 52.12,
+      "learning_rate": 0.00015072115384615384,
+      "loss": 0.8554,
+      "step": 5681
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 0.00015069230769230768,
+      "loss": 0.7179,
+      "step": 5682
+    },
+    {
+      "epoch": 52.14,
+      "learning_rate": 0.00015066346153846152,
+      "loss": 0.682,
+      "step": 5683
+    },
+    {
+      "epoch": 52.15,
+      "learning_rate": 0.00015063461538461538,
+      "loss": 0.7739,
+      "step": 5684
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 0.0001506057692307692,
+      "loss": 0.7973,
+      "step": 5685
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 0.00015057692307692306,
+      "loss": 0.6869,
+      "step": 5686
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 0.00015054807692307693,
+      "loss": 0.7043,
+      "step": 5687
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 0.00015051923076923074,
+      "loss": 0.7829,
+      "step": 5688
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 0.0001504903846153846,
+      "loss": 0.6536,
+      "step": 5689
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 0.00015046153846153844,
+      "loss": 0.7371,
+      "step": 5690
+    },
+    {
+      "epoch": 52.21,
+      "learning_rate": 0.0001504326923076923,
+      "loss": 0.639,
+      "step": 5691
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 0.00015040384615384614,
+      "loss": 0.7374,
+      "step": 5692
+    },
+    {
+      "epoch": 52.23,
+      "learning_rate": 0.00015037499999999998,
+      "loss": 0.7524,
+      "step": 5693
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 0.00015034615384615385,
+      "loss": 0.8083,
+      "step": 5694
+    },
+    {
+      "epoch": 52.25,
+      "learning_rate": 0.00015031730769230766,
+      "loss": 0.7803,
+      "step": 5695
+    },
+    {
+      "epoch": 52.26,
+      "learning_rate": 0.00015028846153846152,
+      "loss": 0.8254,
+      "step": 5696
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 0.0001502596153846154,
+      "loss": 0.6849,
+      "step": 5697
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 0.0001502307692307692,
+      "loss": 0.7824,
+      "step": 5698
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 0.00015020192307692307,
+      "loss": 0.7397,
+      "step": 5699
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 0.0001501730769230769,
+      "loss": 0.6966,
+      "step": 5700
+    },
+    {
+      "epoch": 52.3,
+      "learning_rate": 0.00015014423076923077,
+      "loss": 0.6623,
+      "step": 5701
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 0.0001501153846153846,
+      "loss": 0.752,
+      "step": 5702
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 0.00015008653846153845,
+      "loss": 0.7526,
+      "step": 5703
+    },
+    {
+      "epoch": 52.33,
+      "learning_rate": 0.0001500576923076923,
+      "loss": 0.6141,
+      "step": 5704
+    },
+    {
+      "epoch": 52.34,
+      "learning_rate": 0.00015002884615384612,
+      "loss": 0.8417,
+      "step": 5705
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 0.00015,
+      "loss": 0.701,
+      "step": 5706
+    },
+    {
+      "epoch": 52.36,
+      "learning_rate": 0.00014997115384615383,
+      "loss": 0.6274,
+      "step": 5707
+    },
+    {
+      "epoch": 52.37,
+      "learning_rate": 0.00014994230769230766,
+      "loss": 0.6916,
+      "step": 5708
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 0.00014991346153846153,
+      "loss": 0.6726,
+      "step": 5709
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 0.00014988461538461537,
+      "loss": 0.6602,
+      "step": 5710
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 0.0001498557692307692,
+      "loss": 0.7075,
+      "step": 5711
+    },
+    {
+      "epoch": 52.4,
+      "learning_rate": 0.00014982692307692307,
+      "loss": 0.7681,
+      "step": 5712
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 0.0001497980769230769,
+      "loss": 0.6933,
+      "step": 5713
+    },
+    {
+      "epoch": 52.42,
+      "learning_rate": 0.00014976923076923077,
+      "loss": 0.8095,
+      "step": 5714
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 0.0001497403846153846,
+      "loss": 0.6574,
+      "step": 5715
+    },
+    {
+      "epoch": 52.44,
+      "learning_rate": 0.00014971153846153845,
+      "loss": 0.7448,
+      "step": 5716
+    },
+    {
+      "epoch": 52.45,
+      "learning_rate": 0.0001496826923076923,
+      "loss": 0.8188,
+      "step": 5717
+    },
+    {
+      "epoch": 52.46,
+      "learning_rate": 0.00014965384615384613,
+      "loss": 0.8671,
+      "step": 5718
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 0.000149625,
+      "loss": 0.8905,
+      "step": 5719
+    },
+    {
+      "epoch": 52.48,
+      "learning_rate": 0.00014959615384615383,
+      "loss": 0.8355,
+      "step": 5720
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 0.00014956730769230767,
+      "loss": 0.5866,
+      "step": 5721
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 0.00014953846153846153,
+      "loss": 0.7516,
+      "step": 5722
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 0.00014950961538461537,
+      "loss": 0.7778,
+      "step": 5723
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 0.0001494807692307692,
+      "loss": 0.6679,
+      "step": 5724
+    },
+    {
+      "epoch": 52.52,
+      "learning_rate": 0.00014945192307692307,
+      "loss": 0.8653,
+      "step": 5725
+    },
+    {
+      "epoch": 52.53,
+      "learning_rate": 0.0001494230769230769,
+      "loss": 0.6678,
+      "step": 5726
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 0.00014939423076923075,
+      "loss": 0.7352,
+      "step": 5727
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 0.0001493653846153846,
+      "loss": 0.6479,
+      "step": 5728
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 0.00014933653846153845,
+      "loss": 0.7326,
+      "step": 5729
+    },
+    {
+      "epoch": 52.57,
+      "learning_rate": 0.0001493076923076923,
+      "loss": 0.8676,
+      "step": 5730
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 0.00014927884615384613,
+      "loss": 0.748,
+      "step": 5731
+    },
+    {
+      "epoch": 52.59,
+      "learning_rate": 0.00014925,
+      "loss": 0.7741,
+      "step": 5732
+    },
+    {
+      "epoch": 52.6,
+      "learning_rate": 0.00014922115384615383,
+      "loss": 0.7988,
+      "step": 5733
+    },
+    {
+      "epoch": 52.61,
+      "learning_rate": 0.00014919230769230767,
+      "loss": 0.6643,
+      "step": 5734
+    },
+    {
+      "epoch": 52.61,
+      "learning_rate": 0.00014916346153846154,
+      "loss": 0.7987,
+      "step": 5735
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 0.00014913461538461538,
+      "loss": 0.611,
+      "step": 5736
+    },
+    {
+      "epoch": 52.63,
+      "learning_rate": 0.00014910576923076921,
+      "loss": 0.6926,
+      "step": 5737
+    },
+    {
+      "epoch": 52.64,
+      "learning_rate": 0.00014907692307692305,
+      "loss": 0.6671,
+      "step": 5738
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 0.0001490480769230769,
+      "loss": 0.8026,
+      "step": 5739
+    },
+    {
+      "epoch": 52.66,
+      "learning_rate": 0.00014901923076923076,
+      "loss": 0.7828,
+      "step": 5740
+    },
+    {
+      "epoch": 52.67,
+      "learning_rate": 0.0001489903846153846,
+      "loss": 0.8589,
+      "step": 5741
+    },
+    {
+      "epoch": 52.68,
+      "learning_rate": 0.00014896153846153846,
+      "loss": 0.7421,
+      "step": 5742
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 0.0001489326923076923,
+      "loss": 0.8142,
+      "step": 5743
+    },
+    {
+      "epoch": 52.7,
+      "learning_rate": 0.00014890384615384613,
+      "loss": 0.7855,
+      "step": 5744
+    },
+    {
+      "epoch": 52.71,
+      "learning_rate": 0.000148875,
+      "loss": 0.7564,
+      "step": 5745
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 0.00014884615384615384,
+      "loss": 0.767,
+      "step": 5746
+    },
+    {
+      "epoch": 52.72,
+      "learning_rate": 0.00014881730769230768,
+      "loss": 0.7159,
+      "step": 5747
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 0.00014878846153846151,
+      "loss": 0.6731,
+      "step": 5748
+    },
+    {
+      "epoch": 52.74,
+      "learning_rate": 0.00014875961538461535,
+      "loss": 0.7437,
+      "step": 5749
+    },
+    {
+      "epoch": 52.75,
+      "learning_rate": 0.00014873076923076922,
+      "loss": 0.8162,
+      "step": 5750
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 0.00014870192307692306,
+      "loss": 0.6884,
+      "step": 5751
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 0.00014867307692307692,
+      "loss": 0.6953,
+      "step": 5752
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 0.00014864423076923076,
+      "loss": 0.7868,
+      "step": 5753
+    },
+    {
+      "epoch": 52.79,
+      "learning_rate": 0.0001486153846153846,
+      "loss": 0.8248,
+      "step": 5754
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 0.00014858653846153844,
+      "loss": 0.7768,
+      "step": 5755
+    },
+    {
+      "epoch": 52.81,
+      "learning_rate": 0.0001485576923076923,
+      "loss": 0.7204,
+      "step": 5756
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 0.00014852884615384614,
+      "loss": 0.7608,
+      "step": 5757
+    },
+    {
+      "epoch": 52.83,
+      "learning_rate": 0.00014849999999999998,
+      "loss": 0.7309,
+      "step": 5758
+    },
+    {
+      "epoch": 52.83,
+      "learning_rate": 0.00014847115384615382,
+      "loss": 0.655,
+      "step": 5759
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 0.00014844230769230768,
+      "loss": 0.7986,
+      "step": 5760
+    },
+    {
+      "epoch": 52.85,
+      "learning_rate": 0.00014841346153846152,
+      "loss": 0.7545,
+      "step": 5761
+    },
+    {
+      "epoch": 52.86,
+      "learning_rate": 0.00014838461538461538,
+      "loss": 0.6192,
+      "step": 5762
+    },
+    {
+      "epoch": 52.87,
+      "learning_rate": 0.00014835576923076922,
+      "loss": 0.7232,
+      "step": 5763
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 0.00014832692307692306,
+      "loss": 0.7519,
+      "step": 5764
+    },
+    {
+      "epoch": 52.89,
+      "learning_rate": 0.0001482980769230769,
+      "loss": 0.7867,
+      "step": 5765
+    },
+    {
+      "epoch": 52.9,
+      "learning_rate": 0.00014826923076923076,
+      "loss": 0.7324,
+      "step": 5766
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 0.0001482403846153846,
+      "loss": 0.7679,
+      "step": 5767
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 0.00014821153846153844,
+      "loss": 0.683,
+      "step": 5768
+    },
+    {
+      "epoch": 52.93,
+      "learning_rate": 0.00014818269230769228,
+      "loss": 0.7489,
+      "step": 5769
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 0.00014815384615384614,
+      "loss": 0.7094,
+      "step": 5770
+    },
+    {
+      "epoch": 52.94,
+      "learning_rate": 0.00014812499999999998,
+      "loss": 0.8602,
+      "step": 5771
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 0.00014809615384615385,
+      "loss": 0.7458,
+      "step": 5772
+    },
+    {
+      "epoch": 52.96,
+      "learning_rate": 0.00014806730769230769,
+      "loss": 0.7199,
+      "step": 5773
+    },
+    {
+      "epoch": 52.97,
+      "learning_rate": 0.00014803846153846152,
+      "loss": 0.8921,
+      "step": 5774
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 0.00014800961538461536,
+      "loss": 0.7711,
+      "step": 5775
+    },
+    {
+      "epoch": 52.99,
+      "learning_rate": 0.00014798076923076923,
+      "loss": 0.7687,
+      "step": 5776
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 0.00014795192307692307,
+      "loss": 0.6505,
+      "step": 5777
+    },
+    {
+      "epoch": 53.01,
+      "learning_rate": 0.0001479230769230769,
+      "loss": 0.8131,
+      "step": 5778
+    },
+    {
+      "epoch": 53.02,
+      "learning_rate": 0.00014789423076923074,
+      "loss": 0.748,
+      "step": 5779
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 0.0001478653846153846,
+      "loss": 0.6959,
+      "step": 5780
+    },
+    {
+      "epoch": 53.04,
+      "learning_rate": 0.00014783653846153844,
+      "loss": 0.7904,
+      "step": 5781
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 0.0001478076923076923,
+      "loss": 0.6323,
+      "step": 5782
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 0.00014777884615384615,
+      "loss": 0.8103,
+      "step": 5783
+    },
+    {
+      "epoch": 53.06,
+      "learning_rate": 0.00014774999999999999,
+      "loss": 0.7795,
+      "step": 5784
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 0.00014772115384615382,
+      "loss": 0.732,
+      "step": 5785
+    },
+    {
+      "epoch": 53.08,
+      "learning_rate": 0.0001476923076923077,
+      "loss": 0.7279,
+      "step": 5786
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 0.00014766346153846153,
+      "loss": 0.7816,
+      "step": 5787
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 0.00014763461538461537,
+      "loss": 0.7934,
+      "step": 5788
+    },
+    {
+      "epoch": 53.11,
+      "learning_rate": 0.0001476057692307692,
+      "loss": 0.682,
+      "step": 5789
+    },
+    {
+      "epoch": 53.12,
+      "learning_rate": 0.00014757692307692307,
+      "loss": 0.7091,
+      "step": 5790
+    },
+    {
+      "epoch": 53.13,
+      "learning_rate": 0.0001475480769230769,
+      "loss": 0.7674,
+      "step": 5791
+    },
+    {
+      "epoch": 53.14,
+      "learning_rate": 0.00014751923076923077,
+      "loss": 0.7043,
+      "step": 5792
+    },
+    {
+      "epoch": 53.15,
+      "learning_rate": 0.0001474903846153846,
+      "loss": 0.7939,
+      "step": 5793
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 0.00014746153846153845,
+      "loss": 0.8071,
+      "step": 5794
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 0.0001474326923076923,
+      "loss": 0.7439,
+      "step": 5795
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 0.00014740384615384613,
+      "loss": 0.8088,
+      "step": 5796
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 0.000147375,
+      "loss": 0.7739,
+      "step": 5797
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 0.00014734615384615383,
+      "loss": 0.8271,
+      "step": 5798
+    },
+    {
+      "epoch": 53.2,
+      "learning_rate": 0.00014731730769230767,
+      "loss": 0.7521,
+      "step": 5799
+    },
+    {
+      "epoch": 53.21,
+      "learning_rate": 0.00014728846153846153,
+      "loss": 0.7668,
+      "step": 5800
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 0.00014725961538461537,
+      "loss": 0.6913,
+      "step": 5801
+    },
+    {
+      "epoch": 53.23,
+      "learning_rate": 0.00014723076923076924,
+      "loss": 0.647,
+      "step": 5802
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 0.00014720192307692307,
+      "loss": 0.7913,
+      "step": 5803
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 0.0001471730769230769,
+      "loss": 0.6876,
+      "step": 5804
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 0.00014714423076923075,
+      "loss": 0.7164,
+      "step": 5805
+    },
+    {
+      "epoch": 53.27,
+      "learning_rate": 0.0001471153846153846,
+      "loss": 0.6768,
+      "step": 5806
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 0.00014708653846153845,
+      "loss": 0.7685,
+      "step": 5807
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 0.0001470576923076923,
+      "loss": 0.6897,
+      "step": 5808
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 0.00014702884615384613,
+      "loss": 0.6367,
+      "step": 5809
+    },
+    {
+      "epoch": 53.3,
+      "learning_rate": 0.000147,
+      "loss": 0.7467,
+      "step": 5810
+    },
+    {
+      "epoch": 53.31,
+      "learning_rate": 0.00014697115384615383,
+      "loss": 0.7975,
+      "step": 5811
+    },
+    {
+      "epoch": 53.32,
+      "learning_rate": 0.0001469423076923077,
+      "loss": 0.7699,
+      "step": 5812
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 0.00014691346153846154,
+      "loss": 0.7649,
+      "step": 5813
+    },
+    {
+      "epoch": 53.34,
+      "learning_rate": 0.00014688461538461537,
+      "loss": 0.8369,
+      "step": 5814
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 0.0001468557692307692,
+      "loss": 0.7092,
+      "step": 5815
+    },
+    {
+      "epoch": 53.36,
+      "learning_rate": 0.00014682692307692305,
+      "loss": 0.7331,
+      "step": 5816
+    },
+    {
+      "epoch": 53.37,
+      "learning_rate": 0.00014679807692307692,
+      "loss": 0.7332,
+      "step": 5817
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 0.00014676923076923075,
+      "loss": 0.7266,
+      "step": 5818
+    },
+    {
+      "epoch": 53.39,
+      "learning_rate": 0.0001467403846153846,
+      "loss": 0.7526,
+      "step": 5819
+    },
+    {
+      "epoch": 53.39,
+      "learning_rate": 0.00014671153846153846,
+      "loss": 0.8817,
+      "step": 5820
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 0.0001466826923076923,
+      "loss": 0.7071,
+      "step": 5821
+    },
+    {
+      "epoch": 53.41,
+      "learning_rate": 0.00014665384615384613,
+      "loss": 0.8914,
+      "step": 5822
+    },
+    {
+      "epoch": 53.42,
+      "learning_rate": 0.000146625,
+      "loss": 0.6475,
+      "step": 5823
+    },
+    {
+      "epoch": 53.43,
+      "learning_rate": 0.00014659615384615384,
+      "loss": 0.7728,
+      "step": 5824
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 0.00014656730769230768,
+      "loss": 0.7614,
+      "step": 5825
+    },
+    {
+      "epoch": 53.45,
+      "learning_rate": 0.00014653846153846151,
+      "loss": 0.7284,
+      "step": 5826
+    },
+    {
+      "epoch": 53.46,
+      "learning_rate": 0.00014650961538461535,
+      "loss": 0.7214,
+      "step": 5827
+    },
+    {
+      "epoch": 53.47,
+      "learning_rate": 0.00014648076923076922,
+      "loss": 0.7413,
+      "step": 5828
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 0.00014645192307692306,
+      "loss": 0.6898,
+      "step": 5829
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 0.00014642307692307692,
+      "loss": 0.6787,
+      "step": 5830
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 0.00014639423076923076,
+      "loss": 0.783,
+      "step": 5831
+    },
+    {
+      "epoch": 53.5,
+      "learning_rate": 0.0001463653846153846,
+      "loss": 0.7623,
+      "step": 5832
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 0.00014633653846153846,
+      "loss": 0.7736,
+      "step": 5833
+    },
+    {
+      "epoch": 53.52,
+      "learning_rate": 0.0001463076923076923,
+      "loss": 0.8084,
+      "step": 5834
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 0.00014627884615384614,
+      "loss": 0.6876,
+      "step": 5835
+    },
+    {
+      "epoch": 53.54,
+      "learning_rate": 0.00014624999999999998,
+      "loss": 0.8138,
+      "step": 5836
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 0.00014622115384615382,
+      "loss": 0.8069,
+      "step": 5837
+    },
+    {
+      "epoch": 53.56,
+      "learning_rate": 0.00014619230769230768,
+      "loss": 0.7703,
+      "step": 5838
+    },
+    {
+      "epoch": 53.57,
+      "learning_rate": 0.00014616346153846152,
+      "loss": 0.7926,
+      "step": 5839
+    },
+    {
+      "epoch": 53.58,
+      "learning_rate": 0.00014613461538461538,
+      "loss": 0.6854,
+      "step": 5840
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 0.00014610576923076922,
+      "loss": 0.8466,
+      "step": 5841
+    },
+    {
+      "epoch": 53.6,
+      "learning_rate": 0.00014607692307692306,
+      "loss": 0.7711,
+      "step": 5842
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 0.00014604807692307693,
+      "loss": 0.8695,
+      "step": 5843
+    },
+    {
+      "epoch": 53.61,
+      "learning_rate": 0.00014601923076923076,
+      "loss": 0.8399,
+      "step": 5844
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 0.0001459903846153846,
+      "loss": 0.6377,
+      "step": 5845
+    },
+    {
+      "epoch": 53.63,
+      "learning_rate": 0.00014596153846153844,
+      "loss": 0.5844,
+      "step": 5846
+    },
+    {
+      "epoch": 53.64,
+      "learning_rate": 0.00014593269230769228,
+      "loss": 0.8029,
+      "step": 5847
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 0.00014590384615384614,
+      "loss": 0.8158,
+      "step": 5848
+    },
+    {
+      "epoch": 53.66,
+      "learning_rate": 0.00014587499999999998,
+      "loss": 0.746,
+      "step": 5849
+    },
+    {
+      "epoch": 53.67,
+      "learning_rate": 0.00014584615384615385,
+      "loss": 0.735,
+      "step": 5850
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 0.00014581730769230768,
+      "loss": 0.8466,
+      "step": 5851
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 0.00014578846153846152,
+      "loss": 0.6672,
+      "step": 5852
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 0.00014575961538461536,
+      "loss": 0.7159,
+      "step": 5853
+    },
+    {
+      "epoch": 53.71,
+      "learning_rate": 0.00014573076923076923,
+      "loss": 0.68,
+      "step": 5854
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 0.00014570192307692306,
+      "loss": 0.7996,
+      "step": 5855
+    },
+    {
+      "epoch": 53.72,
+      "learning_rate": 0.0001456730769230769,
+      "loss": 0.7731,
+      "step": 5856
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 0.00014564423076923074,
+      "loss": 0.7547,
+      "step": 5857
+    },
+    {
+      "epoch": 53.74,
+      "learning_rate": 0.0001456153846153846,
+      "loss": 0.6695,
+      "step": 5858
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 0.00014558653846153844,
+      "loss": 0.7821,
+      "step": 5859
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 0.0001455576923076923,
+      "loss": 0.7066,
+      "step": 5860
+    },
+    {
+      "epoch": 53.77,
+      "learning_rate": 0.00014552884615384615,
+      "loss": 0.8864,
+      "step": 5861
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 0.00014549999999999999,
+      "loss": 0.654,
+      "step": 5862
+    },
+    {
+      "epoch": 53.79,
+      "learning_rate": 0.00014547115384615382,
+      "loss": 0.7576,
+      "step": 5863
+    },
+    {
+      "epoch": 53.8,
+      "learning_rate": 0.0001454423076923077,
+      "loss": 0.7172,
+      "step": 5864
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 0.00014541346153846153,
+      "loss": 0.7574,
+      "step": 5865
+    },
+    {
+      "epoch": 53.82,
+      "learning_rate": 0.00014538461538461537,
+      "loss": 0.6774,
+      "step": 5866
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 0.0001453557692307692,
+      "loss": 0.617,
+      "step": 5867
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 0.00014532692307692307,
+      "loss": 0.7921,
+      "step": 5868
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 0.0001452980769230769,
+      "loss": 0.8141,
+      "step": 5869
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 0.00014526923076923077,
+      "loss": 0.7896,
+      "step": 5870
+    },
+    {
+      "epoch": 53.86,
+      "learning_rate": 0.0001452403846153846,
+      "loss": 0.7894,
+      "step": 5871
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 0.00014521153846153845,
+      "loss": 0.6867,
+      "step": 5872
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 0.0001451826923076923,
+      "loss": 0.8106,
+      "step": 5873
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 0.00014515384615384615,
+      "loss": 0.7538,
+      "step": 5874
+    },
+    {
+      "epoch": 53.9,
+      "learning_rate": 0.000145125,
+      "loss": 0.6941,
+      "step": 5875
+    },
+    {
+      "epoch": 53.91,
+      "learning_rate": 0.00014509615384615383,
+      "loss": 0.7229,
+      "step": 5876
+    },
+    {
+      "epoch": 53.92,
+      "learning_rate": 0.00014506730769230767,
+      "loss": 0.7136,
+      "step": 5877
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 0.00014503846153846153,
+      "loss": 0.7899,
+      "step": 5878
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 0.00014500961538461537,
+      "loss": 0.7752,
+      "step": 5879
+    },
+    {
+      "epoch": 53.94,
+      "learning_rate": 0.00014498076923076924,
+      "loss": 0.7348,
+      "step": 5880
+    },
+    {
+      "epoch": 53.95,
+      "learning_rate": 0.00014495192307692307,
+      "loss": 0.7205,
+      "step": 5881
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 0.0001449230769230769,
+      "loss": 0.676,
+      "step": 5882
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 0.00014489423076923075,
+      "loss": 0.9084,
+      "step": 5883
+    },
+    {
+      "epoch": 53.98,
+      "learning_rate": 0.00014486538461538462,
+      "loss": 0.7015,
+      "step": 5884
+    },
+    {
+      "epoch": 53.99,
+      "learning_rate": 0.00014483653846153845,
+      "loss": 0.7023,
+      "step": 5885
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 0.0001448076923076923,
+      "loss": 0.7335,
+      "step": 5886
+    },
+    {
+      "epoch": 54.01,
+      "learning_rate": 0.00014477884615384613,
+      "loss": 0.7511,
+      "step": 5887
+    },
+    {
+      "epoch": 54.02,
+      "learning_rate": 0.00014475,
+      "loss": 0.6411,
+      "step": 5888
+    },
+    {
+      "epoch": 54.03,
+      "learning_rate": 0.00014472115384615383,
+      "loss": 0.6893,
+      "step": 5889
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 0.0001446923076923077,
+      "loss": 0.6639,
+      "step": 5890
+    },
+    {
+      "epoch": 54.05,
+      "learning_rate": 0.00014466346153846154,
+      "loss": 0.8033,
+      "step": 5891
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 0.00014463461538461537,
+      "loss": 0.7104,
+      "step": 5892
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 0.0001446057692307692,
+      "loss": 0.7893,
+      "step": 5893
+    },
+    {
+      "epoch": 54.07,
+      "learning_rate": 0.00014457692307692305,
+      "loss": 0.7041,
+      "step": 5894
+    },
+    {
+      "epoch": 54.08,
+      "learning_rate": 0.00014454807692307692,
+      "loss": 0.711,
+      "step": 5895
+    },
+    {
+      "epoch": 54.09,
+      "learning_rate": 0.00014451923076923075,
+      "loss": 0.6298,
+      "step": 5896
+    },
+    {
+      "epoch": 54.1,
+      "learning_rate": 0.0001444903846153846,
+      "loss": 0.7239,
+      "step": 5897
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 0.00014446153846153846,
+      "loss": 0.5854,
+      "step": 5898
+    },
+    {
+      "epoch": 54.12,
+      "learning_rate": 0.0001444326923076923,
+      "loss": 0.7612,
+      "step": 5899
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 0.00014440384615384616,
+      "loss": 0.6483,
+      "step": 5900
+    },
+    {
+      "epoch": 54.14,
+      "learning_rate": 0.000144375,
+      "loss": 0.557,
+      "step": 5901
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 0.00014434615384615384,
+      "loss": 0.7607,
+      "step": 5902
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 0.00014431730769230768,
+      "loss": 0.6418,
+      "step": 5903
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 0.00014428846153846151,
+      "loss": 0.7118,
+      "step": 5904
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 0.00014425961538461538,
+      "loss": 0.8032,
+      "step": 5905
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 0.00014423076923076922,
+      "loss": 0.7192,
+      "step": 5906
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 0.00014420192307692306,
+      "loss": 0.7631,
+      "step": 5907
+    },
+    {
+      "epoch": 54.2,
+      "learning_rate": 0.00014417307692307692,
+      "loss": 0.8019,
+      "step": 5908
+    },
+    {
+      "epoch": 54.21,
+      "learning_rate": 0.00014414423076923076,
+      "loss": 0.6162,
+      "step": 5909
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 0.0001441153846153846,
+      "loss": 0.6865,
+      "step": 5910
+    },
+    {
+      "epoch": 54.23,
+      "learning_rate": 0.00014408653846153846,
+      "loss": 0.8468,
+      "step": 5911
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 0.0001440576923076923,
+      "loss": 0.7951,
+      "step": 5912
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 0.00014402884615384614,
+      "loss": 0.6623,
+      "step": 5913
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 0.6268,
+      "step": 5914
+    },
+    {
+      "epoch": 54.27,
+      "learning_rate": 0.00014397115384615384,
+      "loss": 0.7375,
+      "step": 5915
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 0.00014394230769230768,
+      "loss": 0.7062,
+      "step": 5916
+    },
+    {
+      "epoch": 54.28,
+      "learning_rate": 0.00014391346153846152,
+      "loss": 0.7894,
+      "step": 5917
+    },
+    {
+      "epoch": 54.29,
+      "learning_rate": 0.00014388461538461538,
+      "loss": 0.8114,
+      "step": 5918
+    },
+    {
+      "epoch": 54.3,
+      "learning_rate": 0.00014385576923076922,
+      "loss": 0.7454,
+      "step": 5919
+    },
+    {
+      "epoch": 54.31,
+      "learning_rate": 0.00014382692307692306,
+      "loss": 0.7045,
+      "step": 5920
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 0.00014379807692307692,
+      "loss": 0.7118,
+      "step": 5921
+    },
+    {
+      "epoch": 54.33,
+      "learning_rate": 0.00014376923076923076,
+      "loss": 0.7424,
+      "step": 5922
+    },
+    {
+      "epoch": 54.34,
+      "learning_rate": 0.0001437403846153846,
+      "loss": 0.7747,
+      "step": 5923
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 0.00014371153846153844,
+      "loss": 0.6267,
+      "step": 5924
+    },
+    {
+      "epoch": 54.36,
+      "learning_rate": 0.00014368269230769228,
+      "loss": 0.8123,
+      "step": 5925
+    },
+    {
+      "epoch": 54.37,
+      "learning_rate": 0.00014365384615384614,
+      "loss": 0.6694,
+      "step": 5926
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 0.00014362499999999998,
+      "loss": 0.7349,
+      "step": 5927
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 0.00014359615384615385,
+      "loss": 0.6674,
+      "step": 5928
+    },
+    {
+      "epoch": 54.39,
+      "learning_rate": 0.00014356730769230768,
+      "loss": 0.7774,
+      "step": 5929
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 0.00014353846153846152,
+      "loss": 0.8248,
+      "step": 5930
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 0.0001435096153846154,
+      "loss": 0.7926,
+      "step": 5931
+    },
+    {
+      "epoch": 54.42,
+      "learning_rate": 0.00014348076923076923,
+      "loss": 0.7084,
+      "step": 5932
+    },
+    {
+      "epoch": 54.43,
+      "learning_rate": 0.00014345192307692306,
+      "loss": 0.7055,
+      "step": 5933
+    },
+    {
+      "epoch": 54.44,
+      "learning_rate": 0.0001434230769230769,
+      "loss": 0.6472,
+      "step": 5934
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 0.00014339423076923074,
+      "loss": 0.8031,
+      "step": 5935
+    },
+    {
+      "epoch": 54.46,
+      "learning_rate": 0.0001433653846153846,
+      "loss": 0.7845,
+      "step": 5936
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 0.00014333653846153844,
+      "loss": 0.7728,
+      "step": 5937
+    },
+    {
+      "epoch": 54.48,
+      "learning_rate": 0.0001433076923076923,
+      "loss": 0.6303,
+      "step": 5938
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 0.00014327884615384615,
+      "loss": 0.6868,
+      "step": 5939
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 0.00014324999999999999,
+      "loss": 0.796,
+      "step": 5940
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 0.00014322115384615385,
+      "loss": 0.7073,
+      "step": 5941
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 0.0001431923076923077,
+      "loss": 0.7256,
+      "step": 5942
+    },
+    {
+      "epoch": 54.52,
+      "learning_rate": 0.00014316346153846153,
+      "loss": 0.731,
+      "step": 5943
+    },
+    {
+      "epoch": 54.53,
+      "learning_rate": 0.00014313461538461537,
+      "loss": 0.6958,
+      "step": 5944
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 0.0001431057692307692,
+      "loss": 0.614,
+      "step": 5945
+    },
+    {
+      "epoch": 54.55,
+      "learning_rate": 0.00014307692307692307,
+      "loss": 0.858,
+      "step": 5946
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 0.0001430480769230769,
+      "loss": 0.727,
+      "step": 5947
+    },
+    {
+      "epoch": 54.57,
+      "learning_rate": 0.00014301923076923074,
+      "loss": 0.8251,
+      "step": 5948
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 0.0001429903846153846,
+      "loss": 0.6658,
+      "step": 5949
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 0.00014296153846153845,
+      "loss": 0.7118,
+      "step": 5950
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 0.00014293269230769229,
+      "loss": 0.805,
+      "step": 5951
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 0.00014290384615384615,
+      "loss": 0.6551,
+      "step": 5952
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 0.000142875,
+      "loss": 0.7207,
+      "step": 5953
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 0.00014284615384615383,
+      "loss": 0.9177,
+      "step": 5954
+    },
+    {
+      "epoch": 54.63,
+      "learning_rate": 0.00014281730769230767,
+      "loss": 0.7139,
+      "step": 5955
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 0.00014278846153846153,
+      "loss": 0.761,
+      "step": 5956
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 0.00014275961538461537,
+      "loss": 0.7442,
+      "step": 5957
+    },
+    {
+      "epoch": 54.66,
+      "learning_rate": 0.0001427307692307692,
+      "loss": 0.7375,
+      "step": 5958
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 0.00014270192307692307,
+      "loss": 0.7339,
+      "step": 5959
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 0.0001426730769230769,
+      "loss": 0.6401,
+      "step": 5960
+    },
+    {
+      "epoch": 54.69,
+      "learning_rate": 0.00014264423076923075,
+      "loss": 0.7456,
+      "step": 5961
+    },
+    {
+      "epoch": 54.7,
+      "learning_rate": 0.00014261538461538461,
+      "loss": 0.7452,
+      "step": 5962
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 0.00014258653846153845,
+      "loss": 0.7436,
+      "step": 5963
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 0.0001425576923076923,
+      "loss": 0.7751,
+      "step": 5964
+    },
+    {
+      "epoch": 54.72,
+      "learning_rate": 0.00014252884615384613,
+      "loss": 0.6908,
+      "step": 5965
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 0.0001425,
+      "loss": 0.7991,
+      "step": 5966
+    },
+    {
+      "epoch": 54.74,
+      "learning_rate": 0.00014247115384615383,
+      "loss": 0.7929,
+      "step": 5967
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 0.00014244230769230767,
+      "loss": 0.6872,
+      "step": 5968
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 0.00014241346153846154,
+      "loss": 0.7665,
+      "step": 5969
+    },
+    {
+      "epoch": 54.77,
+      "learning_rate": 0.00014238461538461537,
+      "loss": 0.7415,
+      "step": 5970
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 0.0001423557692307692,
+      "loss": 0.7301,
+      "step": 5971
+    },
+    {
+      "epoch": 54.79,
+      "learning_rate": 0.00014232692307692308,
+      "loss": 0.6926,
+      "step": 5972
+    },
+    {
+      "epoch": 54.8,
+      "learning_rate": 0.00014229807692307692,
+      "loss": 0.7497,
+      "step": 5973
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 0.00014226923076923075,
+      "loss": 0.8134,
+      "step": 5974
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 0.0001422403846153846,
+      "loss": 0.7013,
+      "step": 5975
+    },
+    {
+      "epoch": 54.83,
+      "learning_rate": 0.00014221153846153843,
+      "loss": 0.6576,
+      "step": 5976
+    },
+    {
+      "epoch": 54.83,
+      "learning_rate": 0.0001421826923076923,
+      "loss": 0.745,
+      "step": 5977
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 0.00014215384615384613,
+      "loss": 0.7765,
+      "step": 5978
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 0.000142125,
+      "loss": 0.7641,
+      "step": 5979
+    },
+    {
+      "epoch": 54.86,
+      "learning_rate": 0.00014209615384615384,
+      "loss": 0.7261,
+      "step": 5980
+    },
+    {
+      "epoch": 54.87,
+      "learning_rate": 0.00014206730769230768,
+      "loss": 0.6414,
+      "step": 5981
+    },
+    {
+      "epoch": 54.88,
+      "learning_rate": 0.00014203846153846154,
+      "loss": 0.6847,
+      "step": 5982
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 0.00014200961538461538,
+      "loss": 0.7021,
+      "step": 5983
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 0.00014198076923076922,
+      "loss": 0.8498,
+      "step": 5984
+    },
+    {
+      "epoch": 54.91,
+      "learning_rate": 0.00014195192307692305,
+      "loss": 0.8168,
+      "step": 5985
+    },
+    {
+      "epoch": 54.92,
+      "learning_rate": 0.0001419230769230769,
+      "loss": 0.6851,
+      "step": 5986
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 0.00014189423076923076,
+      "loss": 0.7651,
+      "step": 5987
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 0.0001418653846153846,
+      "loss": 0.5846,
+      "step": 5988
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 0.00014183653846153846,
+      "loss": 0.7203,
+      "step": 5989
+    },
+    {
+      "epoch": 54.95,
+      "learning_rate": 0.0001418076923076923,
+      "loss": 0.6096,
+      "step": 5990
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 0.00014177884615384614,
+      "loss": 0.8451,
+      "step": 5991
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 0.00014174999999999998,
+      "loss": 0.772,
+      "step": 5992
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 0.00014172115384615384,
+      "loss": 0.6076,
+      "step": 5993
+    },
+    {
+      "epoch": 54.99,
+      "learning_rate": 0.00014169230769230768,
+      "loss": 0.6944,
+      "step": 5994
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 0.00014166346153846152,
+      "loss": 0.7239,
+      "step": 5995
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 0.00014163461538461536,
+      "loss": 0.6041,
+      "step": 5996
+    },
+    {
+      "epoch": 55.02,
+      "learning_rate": 0.00014160576923076922,
+      "loss": 0.7009,
+      "step": 5997
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 0.00014157692307692306,
+      "loss": 0.7262,
+      "step": 5998
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 0.00014154807692307692,
+      "loss": 0.6563,
+      "step": 5999
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 0.00014151923076923076,
+      "loss": 0.8616,
+      "step": 6000
+    },
+    {
+      "epoch": 55.05,
+      "eval_cer": 0.09628330259823985,
+      "eval_loss": 0.3616411089897156,
+      "eval_runtime": 14.0591,
+      "eval_samples_per_second": 117.148,
+      "eval_steps_per_second": 1.849,
+      "eval_wer": 0.3907670309467879,
+      "step": 6000
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 0.0001414903846153846,
+      "loss": 0.6863,
+      "step": 6001
+    },
+    {
+      "epoch": 55.06,
+      "learning_rate": 0.00014146153846153844,
+      "loss": 0.689,
+      "step": 6002
+    },
+    {
+      "epoch": 55.07,
+      "learning_rate": 0.0001414326923076923,
+      "loss": 0.7246,
+      "step": 6003
+    },
+    {
+      "epoch": 55.08,
+      "learning_rate": 0.00014140384615384614,
+      "loss": 0.6813,
+      "step": 6004
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 0.00014137499999999998,
+      "loss": 0.8061,
+      "step": 6005
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 0.00014134615384615382,
+      "loss": 0.7315,
+      "step": 6006
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 0.00014131730769230768,
+      "loss": 0.6669,
+      "step": 6007
+    },
+    {
+      "epoch": 55.12,
+      "learning_rate": 0.00014128846153846152,
+      "loss": 0.6148,
+      "step": 6008
+    },
+    {
+      "epoch": 55.13,
+      "learning_rate": 0.0001412596153846154,
+      "loss": 0.8097,
+      "step": 6009
+    },
+    {
+      "epoch": 55.14,
+      "learning_rate": 0.00014123076923076923,
+      "loss": 0.6407,
+      "step": 6010
+    },
+    {
+      "epoch": 55.15,
+      "learning_rate": 0.00014120192307692306,
+      "loss": 0.7244,
+      "step": 6011
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 0.0001411730769230769,
+      "loss": 0.7372,
+      "step": 6012
+    },
+    {
+      "epoch": 55.17,
+      "learning_rate": 0.00014114423076923077,
+      "loss": 0.8218,
+      "step": 6013
+    },
+    {
+      "epoch": 55.17,
+      "learning_rate": 0.0001411153846153846,
+      "loss": 0.7256,
+      "step": 6014
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 0.00014108653846153844,
+      "loss": 0.7795,
+      "step": 6015
+    },
+    {
+      "epoch": 55.19,
+      "learning_rate": 0.00014105769230769228,
+      "loss": 0.7022,
+      "step": 6016
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 0.00014102884615384615,
+      "loss": 0.7651,
+      "step": 6017
+    },
+    {
+      "epoch": 55.21,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 0.7651,
+      "step": 6018
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 0.00014097115384615385,
+      "loss": 0.6051,
+      "step": 6019
+    },
+    {
+      "epoch": 55.23,
+      "learning_rate": 0.0001409423076923077,
+      "loss": 0.7414,
+      "step": 6020
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 0.00014091346153846153,
+      "loss": 0.667,
+      "step": 6021
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 0.00014088461538461536,
+      "loss": 0.7405,
+      "step": 6022
+    },
+    {
+      "epoch": 55.26,
+      "learning_rate": 0.0001408557692307692,
+      "loss": 0.8073,
+      "step": 6023
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 0.00014082692307692307,
+      "loss": 0.6372,
+      "step": 6024
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 0.0001407980769230769,
+      "loss": 0.8077,
+      "step": 6025
+    },
+    {
+      "epoch": 55.28,
+      "learning_rate": 0.00014076923076923074,
+      "loss": 0.6091,
+      "step": 6026
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 0.0001407403846153846,
+      "loss": 0.7759,
+      "step": 6027
+    },
+    {
+      "epoch": 55.3,
+      "learning_rate": 0.00014071153846153845,
+      "loss": 0.6125,
+      "step": 6028
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 0.0001406826923076923,
+      "loss": 0.7112,
+      "step": 6029
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 0.00014065384615384615,
+      "loss": 0.8072,
+      "step": 6030
+    },
+    {
+      "epoch": 55.33,
+      "learning_rate": 0.000140625,
+      "loss": 0.7551,
+      "step": 6031
+    },
+    {
+      "epoch": 55.34,
+      "learning_rate": 0.00014059615384615383,
+      "loss": 0.7686,
+      "step": 6032
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 0.00014056730769230767,
+      "loss": 0.6571,
+      "step": 6033
+    },
+    {
+      "epoch": 55.36,
+      "learning_rate": 0.00014053846153846153,
+      "loss": 0.7454,
+      "step": 6034
+    },
+    {
+      "epoch": 55.37,
+      "learning_rate": 0.00014050961538461537,
+      "loss": 0.7014,
+      "step": 6035
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 0.0001404807692307692,
+      "loss": 0.7563,
+      "step": 6036
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 0.00014045192307692307,
+      "loss": 0.7366,
+      "step": 6037
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 0.0001404230769230769,
+      "loss": 0.8438,
+      "step": 6038
+    },
+    {
+      "epoch": 55.4,
+      "learning_rate": 0.00014039423076923078,
+      "loss": 0.6675,
+      "step": 6039
+    },
+    {
+      "epoch": 55.41,
+      "learning_rate": 0.00014036538461538461,
+      "loss": 0.7216,
+      "step": 6040
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 0.00014033653846153845,
+      "loss": 0.7352,
+      "step": 6041
+    },
+    {
+      "epoch": 55.43,
+      "learning_rate": 0.00014033653846153845,
+      "loss": 0.6719,
+      "step": 6042
+    },
+    {
+      "epoch": 55.44,
+      "learning_rate": 0.0001403076923076923,
+      "loss": 0.7585,
+      "step": 6043
+    },
+    {
+      "epoch": 55.45,
+      "learning_rate": 0.00014027884615384613,
+      "loss": 0.7586,
+      "step": 6044
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 0.00014025,
+      "loss": 0.5733,
+      "step": 6045
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 0.00014022115384615383,
+      "loss": 0.68,
+      "step": 6046
+    },
+    {
+      "epoch": 55.48,
+      "learning_rate": 0.00014019230769230767,
+      "loss": 0.844,
+      "step": 6047
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 0.00014016346153846154,
+      "loss": 0.8355,
+      "step": 6048
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 0.00014013461538461537,
+      "loss": 0.6665,
+      "step": 6049
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 0.0001401057692307692,
+      "loss": 0.7824,
+      "step": 6050
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 0.00014007692307692308,
+      "loss": 0.8471,
+      "step": 6051
+    },
+    {
+      "epoch": 55.52,
+      "learning_rate": 0.00014004807692307692,
+      "loss": 0.7797,
+      "step": 6052
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 0.00014001923076923075,
+      "loss": 0.6043,
+      "step": 6053
+    },
+    {
+      "epoch": 55.54,
+      "learning_rate": 0.0001399903846153846,
+      "loss": 0.6564,
+      "step": 6054
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 0.00013996153846153843,
+      "loss": 0.66,
+      "step": 6055
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 0.0001399326923076923,
+      "loss": 0.6664,
+      "step": 6056
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 0.00013990384615384613,
+      "loss": 0.6763,
+      "step": 6057
+    },
+    {
+      "epoch": 55.58,
+      "learning_rate": 0.000139875,
+      "loss": 0.7862,
+      "step": 6058
+    },
+    {
+      "epoch": 55.59,
+      "learning_rate": 0.00013984615384615384,
+      "loss": 0.7537,
+      "step": 6059
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 0.00013981730769230767,
+      "loss": 0.73,
+      "step": 6060
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 0.00013978846153846154,
+      "loss": 0.6928,
+      "step": 6061
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 0.00013975961538461538,
+      "loss": 0.673,
+      "step": 6062
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 0.00013973076923076922,
+      "loss": 0.6838,
+      "step": 6063
+    },
+    {
+      "epoch": 55.63,
+      "learning_rate": 0.00013970192307692305,
+      "loss": 0.6799,
+      "step": 6064
+    },
+    {
+      "epoch": 55.64,
+      "learning_rate": 0.0001396730769230769,
+      "loss": 0.6097,
+      "step": 6065
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 0.00013964423076923076,
+      "loss": 0.8047,
+      "step": 6066
+    },
+    {
+      "epoch": 55.66,
+      "learning_rate": 0.0001396153846153846,
+      "loss": 0.7257,
+      "step": 6067
+    },
+    {
+      "epoch": 55.67,
+      "learning_rate": 0.00013958653846153846,
+      "loss": 0.7407,
+      "step": 6068
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 0.0001395576923076923,
+      "loss": 0.647,
+      "step": 6069
+    },
+    {
+      "epoch": 55.69,
+      "learning_rate": 0.00013952884615384614,
+      "loss": 0.6539,
+      "step": 6070
+    },
+    {
+      "epoch": 55.7,
+      "learning_rate": 0.0001395,
+      "loss": 0.7919,
+      "step": 6071
+    },
+    {
+      "epoch": 55.71,
+      "learning_rate": 0.00013947115384615384,
+      "loss": 0.8506,
+      "step": 6072
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 0.00013944230769230768,
+      "loss": 0.8536,
+      "step": 6073
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 0.00013941346153846152,
+      "loss": 0.74,
+      "step": 6074
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 0.00013938461538461536,
+      "loss": 0.7647,
+      "step": 6075
+    },
+    {
+      "epoch": 55.74,
+      "learning_rate": 0.00013935576923076922,
+      "loss": 0.7407,
+      "step": 6076
+    },
+    {
+      "epoch": 55.75,
+      "learning_rate": 0.00013932692307692306,
+      "loss": 0.6365,
+      "step": 6077
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 0.00013929807692307692,
+      "loss": 0.8183,
+      "step": 6078
+    },
+    {
+      "epoch": 55.77,
+      "learning_rate": 0.00013926923076923076,
+      "loss": 0.6404,
+      "step": 6079
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 0.0001392403846153846,
+      "loss": 0.6332,
+      "step": 6080
+    },
+    {
+      "epoch": 55.79,
+      "learning_rate": 0.00013921153846153844,
+      "loss": 0.7521,
+      "step": 6081
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 0.0001391826923076923,
+      "loss": 0.8206,
+      "step": 6082
+    },
+    {
+      "epoch": 55.81,
+      "learning_rate": 0.00013915384615384614,
+      "loss": 0.6288,
+      "step": 6083
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 0.00013912499999999998,
+      "loss": 0.7162,
+      "step": 6084
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 0.00013909615384615382,
+      "loss": 0.6398,
+      "step": 6085
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 0.00013906730769230768,
+      "loss": 0.6097,
+      "step": 6086
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 0.00013903846153846152,
+      "loss": 0.7508,
+      "step": 6087
+    },
+    {
+      "epoch": 55.85,
+      "learning_rate": 0.0001390096153846154,
+      "loss": 0.7586,
+      "step": 6088
+    },
+    {
+      "epoch": 55.86,
+      "learning_rate": 0.00013898076923076923,
+      "loss": 0.7414,
+      "step": 6089
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 0.00013895192307692306,
+      "loss": 0.7095,
+      "step": 6090
+    },
+    {
+      "epoch": 55.88,
+      "learning_rate": 0.0001389230769230769,
+      "loss": 0.7082,
+      "step": 6091
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 0.00013889423076923077,
+      "loss": 0.7732,
+      "step": 6092
+    },
+    {
+      "epoch": 55.9,
+      "learning_rate": 0.0001388653846153846,
+      "loss": 0.6454,
+      "step": 6093
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 0.00013883653846153844,
+      "loss": 0.7883,
+      "step": 6094
+    },
+    {
+      "epoch": 55.92,
+      "learning_rate": 0.00013880769230769228,
+      "loss": 0.719,
+      "step": 6095
+    },
+    {
+      "epoch": 55.93,
+      "learning_rate": 0.00013877884615384615,
+      "loss": 0.7008,
+      "step": 6096
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 0.00013874999999999998,
+      "loss": 0.7201,
+      "step": 6097
+    },
+    {
+      "epoch": 55.94,
+      "learning_rate": 0.00013872115384615385,
+      "loss": 0.6721,
+      "step": 6098
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 0.0001386923076923077,
+      "loss": 0.7379,
+      "step": 6099
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 0.00013866346153846153,
+      "loss": 0.8201,
+      "step": 6100
+    },
+    {
+      "epoch": 55.97,
+      "learning_rate": 0.00013863461538461536,
+      "loss": 0.6661,
+      "step": 6101
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 0.00013860576923076923,
+      "loss": 0.6904,
+      "step": 6102
+    },
+    {
+      "epoch": 55.99,
+      "learning_rate": 0.00013857692307692307,
+      "loss": 0.71,
+      "step": 6103
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 0.0001385480769230769,
+      "loss": 0.7579,
+      "step": 6104
+    },
+    {
+      "epoch": 56.01,
+      "learning_rate": 0.00013851923076923074,
+      "loss": 0.7874,
+      "step": 6105
+    },
+    {
+      "epoch": 56.02,
+      "learning_rate": 0.0001384903846153846,
+      "loss": 0.783,
+      "step": 6106
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 0.00013846153846153845,
+      "loss": 0.7202,
+      "step": 6107
+    },
+    {
+      "epoch": 56.04,
+      "learning_rate": 0.0001384326923076923,
+      "loss": 0.7547,
+      "step": 6108
+    },
+    {
+      "epoch": 56.05,
+      "learning_rate": 0.00013840384615384615,
+      "loss": 0.7316,
+      "step": 6109
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 0.000138375,
+      "loss": 0.7399,
+      "step": 6110
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 0.00013834615384615383,
+      "loss": 0.7195,
+      "step": 6111
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 0.0001383173076923077,
+      "loss": 0.7357,
+      "step": 6112
+    },
+    {
+      "epoch": 56.08,
+      "learning_rate": 0.00013828846153846153,
+      "loss": 0.6981,
+      "step": 6113
+    },
+    {
+      "epoch": 56.09,
+      "learning_rate": 0.00013825961538461537,
+      "loss": 0.6252,
+      "step": 6114
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 0.0001382307692307692,
+      "loss": 0.7641,
+      "step": 6115
+    },
+    {
+      "epoch": 56.11,
+      "learning_rate": 0.00013820192307692307,
+      "loss": 0.7034,
+      "step": 6116
+    },
+    {
+      "epoch": 56.12,
+      "learning_rate": 0.0001381730769230769,
+      "loss": 0.6401,
+      "step": 6117
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 0.00013814423076923078,
+      "loss": 0.6345,
+      "step": 6118
+    },
+    {
+      "epoch": 56.14,
+      "learning_rate": 0.00013811538461538461,
+      "loss": 0.67,
+      "step": 6119
+    },
+    {
+      "epoch": 56.15,
+      "learning_rate": 0.00013808653846153845,
+      "loss": 0.7445,
+      "step": 6120
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 0.0001380576923076923,
+      "loss": 0.6525,
+      "step": 6121
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 0.00013802884615384613,
+      "loss": 0.6922,
+      "step": 6122
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 0.000138,
+      "loss": 0.6206,
+      "step": 6123
+    },
+    {
+      "epoch": 56.18,
+      "learning_rate": 0.00013797115384615383,
+      "loss": 0.8053,
+      "step": 6124
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 0.00013794230769230767,
+      "loss": 0.7353,
+      "step": 6125
+    },
+    {
+      "epoch": 56.2,
+      "learning_rate": 0.00013791346153846153,
+      "loss": 0.7385,
+      "step": 6126
+    },
+    {
+      "epoch": 56.21,
+      "learning_rate": 0.00013788461538461537,
+      "loss": 0.6298,
+      "step": 6127
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 0.00013785576923076924,
+      "loss": 0.7766,
+      "step": 6128
+    },
+    {
+      "epoch": 56.23,
+      "learning_rate": 0.00013782692307692308,
+      "loss": 0.5864,
+      "step": 6129
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 0.00013779807692307691,
+      "loss": 0.7616,
+      "step": 6130
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 0.00013776923076923075,
+      "loss": 0.7752,
+      "step": 6131
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 0.0001377403846153846,
+      "loss": 0.5992,
+      "step": 6132
+    },
+    {
+      "epoch": 56.27,
+      "learning_rate": 0.00013771153846153846,
+      "loss": 0.6192,
+      "step": 6133
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 0.0001376826923076923,
+      "loss": 0.6542,
+      "step": 6134
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 0.00013765384615384613,
+      "loss": 0.6364,
+      "step": 6135
+    },
+    {
+      "epoch": 56.29,
+      "learning_rate": 0.000137625,
+      "loss": 0.839,
+      "step": 6136
+    },
+    {
+      "epoch": 56.3,
+      "learning_rate": 0.00013759615384615384,
+      "loss": 0.6632,
+      "step": 6137
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 0.0001375673076923077,
+      "loss": 0.6972,
+      "step": 6138
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 0.00013753846153846154,
+      "loss": 0.7997,
+      "step": 6139
+    },
+    {
+      "epoch": 56.33,
+      "learning_rate": 0.00013750961538461538,
+      "loss": 0.7099,
+      "step": 6140
+    },
+    {
+      "epoch": 56.34,
+      "learning_rate": 0.00013748076923076922,
+      "loss": 0.8259,
+      "step": 6141
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 0.00013745192307692305,
+      "loss": 0.7569,
+      "step": 6142
+    },
+    {
+      "epoch": 56.36,
+      "learning_rate": 0.00013742307692307692,
+      "loss": 0.6914,
+      "step": 6143
+    },
+    {
+      "epoch": 56.37,
+      "learning_rate": 0.00013739423076923076,
+      "loss": 0.7781,
+      "step": 6144
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 0.0001373653846153846,
+      "loss": 0.6826,
+      "step": 6145
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 0.00013733653846153846,
+      "loss": 0.758,
+      "step": 6146
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 0.0001373076923076923,
+      "loss": 0.7115,
+      "step": 6147
+    },
+    {
+      "epoch": 56.4,
+      "learning_rate": 0.00013727884615384614,
+      "loss": 0.7212,
+      "step": 6148
+    },
+    {
+      "epoch": 56.41,
+      "learning_rate": 0.00013725,
+      "loss": 0.657,
+      "step": 6149
+    },
+    {
+      "epoch": 56.42,
+      "learning_rate": 0.00013722115384615384,
+      "loss": 0.6132,
+      "step": 6150
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 0.00013719230769230768,
+      "loss": 0.7013,
+      "step": 6151
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 0.00013716346153846152,
+      "loss": 0.7712,
+      "step": 6152
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 0.00013713461538461535,
+      "loss": 0.6868,
+      "step": 6153
+    },
+    {
+      "epoch": 56.46,
+      "learning_rate": 0.00013710576923076922,
+      "loss": 0.6185,
+      "step": 6154
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 0.00013707692307692306,
+      "loss": 0.7508,
+      "step": 6155
+    },
+    {
+      "epoch": 56.48,
+      "learning_rate": 0.00013704807692307692,
+      "loss": 0.66,
+      "step": 6156
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 0.00013701923076923076,
+      "loss": 0.6682,
+      "step": 6157
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 0.0001369903846153846,
+      "loss": 0.6843,
+      "step": 6158
+    },
+    {
+      "epoch": 56.5,
+      "learning_rate": 0.00013696153846153847,
+      "loss": 0.7053,
+      "step": 6159
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 0.0001369326923076923,
+      "loss": 0.7094,
+      "step": 6160
+    },
+    {
+      "epoch": 56.52,
+      "learning_rate": 0.00013690384615384614,
+      "loss": 0.6319,
+      "step": 6161
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 0.00013687499999999998,
+      "loss": 0.6565,
+      "step": 6162
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 0.00013684615384615382,
+      "loss": 0.6278,
+      "step": 6163
+    },
+    {
+      "epoch": 56.55,
+      "learning_rate": 0.00013681730769230768,
+      "loss": 0.6651,
+      "step": 6164
+    },
+    {
+      "epoch": 56.56,
+      "learning_rate": 0.00013678846153846152,
+      "loss": 0.6921,
+      "step": 6165
+    },
+    {
+      "epoch": 56.57,
+      "learning_rate": 0.00013675961538461539,
+      "loss": 0.861,
+      "step": 6166
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 0.00013673076923076922,
+      "loss": 0.6424,
+      "step": 6167
+    },
+    {
+      "epoch": 56.59,
+      "learning_rate": 0.00013670192307692306,
+      "loss": 0.8173,
+      "step": 6168
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 0.00013667307692307693,
+      "loss": 0.5613,
+      "step": 6169
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 0.00013664423076923077,
+      "loss": 0.7724,
+      "step": 6170
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 0.0001366153846153846,
+      "loss": 0.6247,
+      "step": 6171
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 0.00013658653846153844,
+      "loss": 0.7161,
+      "step": 6172
+    },
+    {
+      "epoch": 56.63,
+      "learning_rate": 0.00013655769230769228,
+      "loss": 0.7977,
+      "step": 6173
+    },
+    {
+      "epoch": 56.64,
+      "learning_rate": 0.00013652884615384615,
+      "loss": 0.5772,
+      "step": 6174
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 0.00013649999999999998,
+      "loss": 0.6883,
+      "step": 6175
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 0.00013647115384615382,
+      "loss": 0.7058,
+      "step": 6176
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 0.0001364423076923077,
+      "loss": 0.7666,
+      "step": 6177
+    },
+    {
+      "epoch": 56.68,
+      "learning_rate": 0.00013641346153846153,
+      "loss": 0.684,
+      "step": 6178
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 0.00013638461538461536,
+      "loss": 0.6725,
+      "step": 6179
+    },
+    {
+      "epoch": 56.7,
+      "learning_rate": 0.00013635576923076923,
+      "loss": 0.6595,
+      "step": 6180
+    },
+    {
+      "epoch": 56.71,
+      "learning_rate": 0.00013632692307692307,
+      "loss": 0.7367,
+      "step": 6181
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 0.0001362980769230769,
+      "loss": 0.6686,
+      "step": 6182
+    },
+    {
+      "epoch": 56.72,
+      "learning_rate": 0.00013626923076923074,
+      "loss": 0.7521,
+      "step": 6183
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 0.0001362403846153846,
+      "loss": 0.6819,
+      "step": 6184
+    },
+    {
+      "epoch": 56.74,
+      "learning_rate": 0.00013621153846153845,
+      "loss": 0.7033,
+      "step": 6185
+    },
+    {
+      "epoch": 56.75,
+      "learning_rate": 0.00013618269230769228,
+      "loss": 0.7964,
+      "step": 6186
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 0.00013615384615384615,
+      "loss": 0.628,
+      "step": 6187
+    },
+    {
+      "epoch": 56.77,
+      "learning_rate": 0.000136125,
+      "loss": 0.558,
+      "step": 6188
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 0.00013609615384615383,
+      "loss": 0.8112,
+      "step": 6189
+    },
+    {
+      "epoch": 56.79,
+      "learning_rate": 0.0001360673076923077,
+      "loss": 0.6239,
+      "step": 6190
+    },
+    {
+      "epoch": 56.8,
+      "learning_rate": 0.00013603846153846153,
+      "loss": 0.7915,
+      "step": 6191
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 0.00013600961538461537,
+      "loss": 0.6436,
+      "step": 6192
+    },
+    {
+      "epoch": 56.82,
+      "learning_rate": 0.0001359807692307692,
+      "loss": 0.7857,
+      "step": 6193
+    },
+    {
+      "epoch": 56.83,
+      "learning_rate": 0.00013595192307692307,
+      "loss": 0.7021,
+      "step": 6194
+    },
+    {
+      "epoch": 56.83,
+      "learning_rate": 0.0001359230769230769,
+      "loss": 0.782,
+      "step": 6195
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 0.00013589423076923075,
+      "loss": 0.673,
+      "step": 6196
+    },
+    {
+      "epoch": 56.85,
+      "learning_rate": 0.0001358653846153846,
+      "loss": 0.6316,
+      "step": 6197
+    },
+    {
+      "epoch": 56.86,
+      "learning_rate": 0.00013583653846153845,
+      "loss": 0.7581,
+      "step": 6198
+    },
+    {
+      "epoch": 56.87,
+      "learning_rate": 0.0001358076923076923,
+      "loss": 0.706,
+      "step": 6199
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 0.00013577884615384615,
+      "loss": 0.7716,
+      "step": 6200
+    },
+    {
+      "epoch": 56.89,
+      "learning_rate": 0.00013575,
+      "loss": 0.6337,
+      "step": 6201
+    },
+    {
+      "epoch": 56.9,
+      "learning_rate": 0.00013572115384615383,
+      "loss": 0.7029,
+      "step": 6202
+    },
+    {
+      "epoch": 56.91,
+      "learning_rate": 0.00013569230769230767,
+      "loss": 0.5836,
+      "step": 6203
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 0.00013566346153846153,
+      "loss": 0.6598,
+      "step": 6204
+    },
+    {
+      "epoch": 56.93,
+      "learning_rate": 0.00013563461538461537,
+      "loss": 0.7323,
+      "step": 6205
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 0.0001356057692307692,
+      "loss": 0.7452,
+      "step": 6206
+    },
+    {
+      "epoch": 56.94,
+      "learning_rate": 0.00013557692307692308,
+      "loss": 0.7027,
+      "step": 6207
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 0.00013554807692307691,
+      "loss": 0.692,
+      "step": 6208
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 0.00013551923076923075,
+      "loss": 0.8282,
+      "step": 6209
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 0.00013549038461538462,
+      "loss": 0.6404,
+      "step": 6210
+    },
+    {
+      "epoch": 56.98,
+      "learning_rate": 0.00013546153846153846,
+      "loss": 0.6194,
+      "step": 6211
+    },
+    {
+      "epoch": 56.99,
+      "learning_rate": 0.0001354326923076923,
+      "loss": 0.7177,
+      "step": 6212
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 0.00013540384615384613,
+      "loss": 0.665,
+      "step": 6213
+    },
+    {
+      "epoch": 57.01,
+      "learning_rate": 0.00013537499999999997,
+      "loss": 0.799,
+      "step": 6214
+    },
+    {
+      "epoch": 57.02,
+      "learning_rate": 0.00013534615384615384,
+      "loss": 0.734,
+      "step": 6215
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 0.00013531730769230767,
+      "loss": 0.5981,
+      "step": 6216
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 0.00013528846153846154,
+      "loss": 0.6532,
+      "step": 6217
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 0.00013525961538461538,
+      "loss": 0.7727,
+      "step": 6218
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 0.00013523076923076922,
+      "loss": 0.6284,
+      "step": 6219
+    },
+    {
+      "epoch": 57.06,
+      "learning_rate": 0.00013520192307692305,
+      "loss": 0.7326,
+      "step": 6220
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 0.00013517307692307692,
+      "loss": 0.6131,
+      "step": 6221
+    },
+    {
+      "epoch": 57.08,
+      "learning_rate": 0.00013514423076923076,
+      "loss": 0.6515,
+      "step": 6222
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 0.0001351153846153846,
+      "loss": 0.7219,
+      "step": 6223
+    },
+    {
+      "epoch": 57.1,
+      "learning_rate": 0.00013508653846153843,
+      "loss": 0.8251,
+      "step": 6224
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 0.0001350576923076923,
+      "loss": 0.7216,
+      "step": 6225
+    },
+    {
+      "epoch": 57.12,
+      "learning_rate": 0.00013502884615384614,
+      "loss": 0.6944,
+      "step": 6226
+    },
+    {
+      "epoch": 57.13,
+      "learning_rate": 0.000135,
+      "loss": 0.6941,
+      "step": 6227
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 0.00013497115384615384,
+      "loss": 0.6759,
+      "step": 6228
+    },
+    {
+      "epoch": 57.15,
+      "learning_rate": 0.00013494230769230768,
+      "loss": 0.7191,
+      "step": 6229
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 0.00013491346153846152,
+      "loss": 0.786,
+      "step": 6230
+    },
+    {
+      "epoch": 57.17,
+      "learning_rate": 0.00013488461538461538,
+      "loss": 0.6762,
+      "step": 6231
+    },
+    {
+      "epoch": 57.17,
+      "learning_rate": 0.00013485576923076922,
+      "loss": 0.7476,
+      "step": 6232
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 0.00013482692307692306,
+      "loss": 0.7614,
+      "step": 6233
+    },
+    {
+      "epoch": 57.19,
+      "learning_rate": 0.0001347980769230769,
+      "loss": 0.6163,
+      "step": 6234
+    },
+    {
+      "epoch": 57.2,
+      "learning_rate": 0.00013476923076923076,
+      "loss": 0.7878,
+      "step": 6235
+    },
+    {
+      "epoch": 57.21,
+      "learning_rate": 0.0001347403846153846,
+      "loss": 0.7082,
+      "step": 6236
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 0.00013471153846153846,
+      "loss": 0.7939,
+      "step": 6237
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 0.0001346826923076923,
+      "loss": 0.6703,
+      "step": 6238
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 0.00013465384615384614,
+      "loss": 0.8448,
+      "step": 6239
+    },
+    {
+      "epoch": 57.25,
+      "learning_rate": 0.00013462499999999998,
+      "loss": 0.6895,
+      "step": 6240
+    },
+    {
+      "epoch": 57.26,
+      "learning_rate": 0.00013459615384615384,
+      "loss": 0.6787,
+      "step": 6241
+    },
+    {
+      "epoch": 57.27,
+      "learning_rate": 0.00013456730769230768,
+      "loss": 0.7309,
+      "step": 6242
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 0.00013453846153846152,
+      "loss": 0.8462,
+      "step": 6243
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 0.00013450961538461536,
+      "loss": 0.7112,
+      "step": 6244
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 0.00013448076923076922,
+      "loss": 0.7226,
+      "step": 6245
+    },
+    {
+      "epoch": 57.3,
+      "learning_rate": 0.00013445192307692306,
+      "loss": 0.7211,
+      "step": 6246
+    },
+    {
+      "epoch": 57.31,
+      "learning_rate": 0.00013442307692307693,
+      "loss": 0.778,
+      "step": 6247
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 0.00013439423076923077,
+      "loss": 0.6746,
+      "step": 6248
+    },
+    {
+      "epoch": 57.33,
+      "learning_rate": 0.0001343653846153846,
+      "loss": 0.7087,
+      "step": 6249
+    },
+    {
+      "epoch": 57.34,
+      "learning_rate": 0.00013433653846153844,
+      "loss": 0.6826,
+      "step": 6250
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 0.00013430769230769228,
+      "loss": 0.7902,
+      "step": 6251
+    },
+    {
+      "epoch": 57.36,
+      "learning_rate": 0.00013427884615384615,
+      "loss": 0.7261,
+      "step": 6252
+    },
+    {
+      "epoch": 57.37,
+      "learning_rate": 0.00013424999999999998,
+      "loss": 0.7182,
+      "step": 6253
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 0.00013422115384615382,
+      "loss": 0.6996,
+      "step": 6254
+    },
+    {
+      "epoch": 57.39,
+      "learning_rate": 0.0001341923076923077,
+      "loss": 0.6662,
+      "step": 6255
+    },
+    {
+      "epoch": 57.39,
+      "learning_rate": 0.00013416346153846153,
+      "loss": 0.799,
+      "step": 6256
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 0.0001341346153846154,
+      "loss": 0.7881,
+      "step": 6257
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 0.00013410576923076923,
+      "loss": 0.644,
+      "step": 6258
+    },
+    {
+      "epoch": 57.42,
+      "learning_rate": 0.00013407692307692307,
+      "loss": 0.754,
+      "step": 6259
+    },
+    {
+      "epoch": 57.43,
+      "learning_rate": 0.0001340480769230769,
+      "loss": 0.6542,
+      "step": 6260
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 0.00013401923076923074,
+      "loss": 0.9827,
+      "step": 6261
+    },
+    {
+      "epoch": 57.45,
+      "learning_rate": 0.0001339903846153846,
+      "loss": 0.7242,
+      "step": 6262
+    },
+    {
+      "epoch": 57.46,
+      "learning_rate": 0.00013396153846153845,
+      "loss": 0.6873,
+      "step": 6263
+    },
+    {
+      "epoch": 57.47,
+      "learning_rate": 0.00013393269230769228,
+      "loss": 0.6601,
+      "step": 6264
+    },
+    {
+      "epoch": 57.48,
+      "learning_rate": 0.00013390384615384615,
+      "loss": 0.6599,
+      "step": 6265
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 0.000133875,
+      "loss": 0.5892,
+      "step": 6266
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 0.00013384615384615385,
+      "loss": 0.7995,
+      "step": 6267
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 0.0001338173076923077,
+      "loss": 0.7766,
+      "step": 6268
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 0.00013378846153846153,
+      "loss": 0.7996,
+      "step": 6269
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 0.00013375961538461537,
+      "loss": 0.7952,
+      "step": 6270
+    },
+    {
+      "epoch": 57.53,
+      "learning_rate": 0.0001337307692307692,
+      "loss": 0.607,
+      "step": 6271
+    },
+    {
+      "epoch": 57.54,
+      "learning_rate": 0.00013370192307692307,
+      "loss": 0.6727,
+      "step": 6272
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 0.0001336730769230769,
+      "loss": 0.6137,
+      "step": 6273
+    },
+    {
+      "epoch": 57.56,
+      "learning_rate": 0.00013364423076923075,
+      "loss": 0.6547,
+      "step": 6274
+    },
+    {
+      "epoch": 57.57,
+      "learning_rate": 0.0001336153846153846,
+      "loss": 0.6748,
+      "step": 6275
+    },
+    {
+      "epoch": 57.58,
+      "learning_rate": 0.00013358653846153845,
+      "loss": 0.759,
+      "step": 6276
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 0.0001335576923076923,
+      "loss": 0.813,
+      "step": 6277
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 0.00013352884615384615,
+      "loss": 0.686,
+      "step": 6278
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 0.0001335,
+      "loss": 0.6764,
+      "step": 6279
+    },
+    {
+      "epoch": 57.61,
+      "learning_rate": 0.00013347115384615383,
+      "loss": 0.6904,
+      "step": 6280
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 0.00013344230769230767,
+      "loss": 0.8033,
+      "step": 6281
+    },
+    {
+      "epoch": 57.63,
+      "learning_rate": 0.0001334134615384615,
+      "loss": 0.7402,
+      "step": 6282
+    },
+    {
+      "epoch": 57.64,
+      "learning_rate": 0.00013338461538461537,
+      "loss": 0.7071,
+      "step": 6283
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 0.0001333557692307692,
+      "loss": 0.5966,
+      "step": 6284
+    },
+    {
+      "epoch": 57.66,
+      "learning_rate": 0.00013332692307692308,
+      "loss": 0.6489,
+      "step": 6285
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 0.00013329807692307691,
+      "loss": 0.7626,
+      "step": 6286
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 0.00013326923076923075,
+      "loss": 0.7227,
+      "step": 6287
+    },
+    {
+      "epoch": 57.69,
+      "learning_rate": 0.00013324038461538462,
+      "loss": 0.6527,
+      "step": 6288
+    },
+    {
+      "epoch": 57.7,
+      "learning_rate": 0.00013321153846153846,
+      "loss": 0.7644,
+      "step": 6289
+    },
+    {
+      "epoch": 57.71,
+      "learning_rate": 0.0001331826923076923,
+      "loss": 0.6428,
+      "step": 6290
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 0.00013315384615384613,
+      "loss": 0.7269,
+      "step": 6291
+    },
+    {
+      "epoch": 57.72,
+      "learning_rate": 0.00013312499999999997,
+      "loss": 0.7768,
+      "step": 6292
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 0.00013309615384615383,
+      "loss": 0.5932,
+      "step": 6293
+    },
+    {
+      "epoch": 57.74,
+      "learning_rate": 0.00013306730769230767,
+      "loss": 0.7291,
+      "step": 6294
+    },
+    {
+      "epoch": 57.75,
+      "learning_rate": 0.00013303846153846154,
+      "loss": 0.6835,
+      "step": 6295
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 0.00013300961538461538,
+      "loss": 0.6866,
+      "step": 6296
+    },
+    {
+      "epoch": 57.77,
+      "learning_rate": 0.00013298076923076921,
+      "loss": 0.7277,
+      "step": 6297
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 0.00013295192307692308,
+      "loss": 0.8515,
+      "step": 6298
+    },
+    {
+      "epoch": 57.79,
+      "learning_rate": 0.00013292307692307692,
+      "loss": 0.6286,
+      "step": 6299
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 0.00013289423076923076,
+      "loss": 0.748,
+      "step": 6300
+    },
+    {
+      "epoch": 57.81,
+      "learning_rate": 0.0001328653846153846,
+      "loss": 0.7169,
+      "step": 6301
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 0.00013283653846153843,
+      "loss": 0.6906,
+      "step": 6302
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 0.0001328076923076923,
+      "loss": 0.6058,
+      "step": 6303
+    },
+    {
+      "epoch": 57.83,
+      "learning_rate": 0.00013277884615384614,
+      "loss": 0.6032,
+      "step": 6304
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 0.00013275,
+      "loss": 0.7454,
+      "step": 6305
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 0.00013272115384615384,
+      "loss": 0.5659,
+      "step": 6306
+    },
+    {
+      "epoch": 57.86,
+      "learning_rate": 0.00013269230769230768,
+      "loss": 0.6968,
+      "step": 6307
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 0.00013266346153846154,
+      "loss": 0.6911,
+      "step": 6308
+    },
+    {
+      "epoch": 57.88,
+      "learning_rate": 0.00013263461538461538,
+      "loss": 0.6311,
+      "step": 6309
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 0.00013260576923076922,
+      "loss": 0.6708,
+      "step": 6310
+    },
+    {
+      "epoch": 57.9,
+      "learning_rate": 0.00013257692307692306,
+      "loss": 0.687,
+      "step": 6311
+    },
+    {
+      "epoch": 57.91,
+      "learning_rate": 0.0001325480769230769,
+      "loss": 0.6705,
+      "step": 6312
+    },
+    {
+      "epoch": 57.92,
+      "learning_rate": 0.00013251923076923076,
+      "loss": 0.6116,
+      "step": 6313
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 0.0001324903846153846,
+      "loss": 0.7275,
+      "step": 6314
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 0.00013246153846153846,
+      "loss": 0.5947,
+      "step": 6315
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 0.0001324326923076923,
+      "loss": 0.6485,
+      "step": 6316
+    },
+    {
+      "epoch": 57.95,
+      "learning_rate": 0.00013240384615384614,
+      "loss": 0.6689,
+      "step": 6317
+    },
+    {
+      "epoch": 57.96,
+      "learning_rate": 0.00013237499999999998,
+      "loss": 0.7052,
+      "step": 6318
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 0.00013234615384615384,
+      "loss": 0.6788,
+      "step": 6319
+    },
+    {
+      "epoch": 57.98,
+      "learning_rate": 0.00013231730769230768,
+      "loss": 0.8048,
+      "step": 6320
+    },
+    {
+      "epoch": 57.99,
+      "learning_rate": 0.00013228846153846152,
+      "loss": 0.7002,
+      "step": 6321
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 0.00013225961538461536,
+      "loss": 0.591,
+      "step": 6322
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 0.00013223076923076922,
+      "loss": 0.7052,
+      "step": 6323
+    },
+    {
+      "epoch": 58.02,
+      "learning_rate": 0.00013220192307692306,
+      "loss": 0.738,
+      "step": 6324
+    },
+    {
+      "epoch": 58.03,
+      "learning_rate": 0.00013217307692307693,
+      "loss": 0.6836,
+      "step": 6325
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 0.00013214423076923077,
+      "loss": 0.8392,
+      "step": 6326
+    },
+    {
+      "epoch": 58.05,
+      "learning_rate": 0.0001321153846153846,
+      "loss": 0.7378,
+      "step": 6327
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 0.00013208653846153844,
+      "loss": 0.5858,
+      "step": 6328
+    },
+    {
+      "epoch": 58.06,
+      "learning_rate": 0.0001320576923076923,
+      "loss": 0.6358,
+      "step": 6329
+    },
+    {
+      "epoch": 58.07,
+      "learning_rate": 0.00013202884615384614,
+      "loss": 0.7049,
+      "step": 6330
+    },
+    {
+      "epoch": 58.08,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 0.58,
+      "step": 6331
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 0.00013197115384615382,
+      "loss": 0.7951,
+      "step": 6332
+    },
+    {
+      "epoch": 58.1,
+      "learning_rate": 0.00013194230769230769,
+      "loss": 0.7781,
+      "step": 6333
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 0.00013191346153846152,
+      "loss": 0.7464,
+      "step": 6334
+    },
+    {
+      "epoch": 58.12,
+      "learning_rate": 0.0001318846153846154,
+      "loss": 0.6443,
+      "step": 6335
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 0.00013185576923076923,
+      "loss": 0.7405,
+      "step": 6336
+    },
+    {
+      "epoch": 58.14,
+      "learning_rate": 0.00013182692307692307,
+      "loss": 0.6673,
+      "step": 6337
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 0.0001317980769230769,
+      "loss": 0.5664,
+      "step": 6338
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 0.00013176923076923077,
+      "loss": 0.6869,
+      "step": 6339
+    },
+    {
+      "epoch": 58.17,
+      "learning_rate": 0.0001317403846153846,
+      "loss": 0.7746,
+      "step": 6340
+    },
+    {
+      "epoch": 58.17,
+      "learning_rate": 0.00013171153846153845,
+      "loss": 0.7701,
+      "step": 6341
+    },
+    {
+      "epoch": 58.18,
+      "learning_rate": 0.00013168269230769228,
+      "loss": 0.7385,
+      "step": 6342
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 0.00013165384615384615,
+      "loss": 0.6631,
+      "step": 6343
+    },
+    {
+      "epoch": 58.2,
+      "learning_rate": 0.000131625,
+      "loss": 0.7147,
+      "step": 6344
+    },
+    {
+      "epoch": 58.21,
+      "learning_rate": 0.00013159615384615385,
+      "loss": 0.8587,
+      "step": 6345
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 0.0001315673076923077,
+      "loss": 0.6437,
+      "step": 6346
+    },
+    {
+      "epoch": 58.23,
+      "learning_rate": 0.00013153846153846153,
+      "loss": 0.6948,
+      "step": 6347
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 0.00013150961538461537,
+      "loss": 0.7295,
+      "step": 6348
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 0.0001314807692307692,
+      "loss": 0.6502,
+      "step": 6349
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 0.00013145192307692307,
+      "loss": 0.7448,
+      "step": 6350
+    },
+    {
+      "epoch": 58.27,
+      "learning_rate": 0.0001314230769230769,
+      "loss": 0.7156,
+      "step": 6351
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 0.00013139423076923075,
+      "loss": 0.8379,
+      "step": 6352
+    },
+    {
+      "epoch": 58.28,
+      "learning_rate": 0.0001313653846153846,
+      "loss": 0.7074,
+      "step": 6353
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 0.00013133653846153845,
+      "loss": 0.7099,
+      "step": 6354
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 0.00013130769230769232,
+      "loss": 0.7238,
+      "step": 6355
+    },
+    {
+      "epoch": 58.31,
+      "learning_rate": 0.00013127884615384615,
+      "loss": 0.7203,
+      "step": 6356
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 0.00013125,
+      "loss": 0.7497,
+      "step": 6357
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 0.00013122115384615383,
+      "loss": 0.6929,
+      "step": 6358
+    },
+    {
+      "epoch": 58.34,
+      "learning_rate": 0.00013119230769230767,
+      "loss": 0.6957,
+      "step": 6359
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 0.00013116346153846153,
+      "loss": 0.6222,
+      "step": 6360
+    },
+    {
+      "epoch": 58.36,
+      "learning_rate": 0.00013113461538461537,
+      "loss": 0.8128,
+      "step": 6361
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 0.0001311057692307692,
+      "loss": 0.6897,
+      "step": 6362
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 0.00013107692307692308,
+      "loss": 0.6984,
+      "step": 6363
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 0.0001310480769230769,
+      "loss": 0.7634,
+      "step": 6364
+    },
+    {
+      "epoch": 58.39,
+      "learning_rate": 0.00013101923076923078,
+      "loss": 0.6676,
+      "step": 6365
+    },
+    {
+      "epoch": 58.4,
+      "learning_rate": 0.00013099038461538462,
+      "loss": 0.667,
+      "step": 6366
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 0.00013096153846153845,
+      "loss": 0.7936,
+      "step": 6367
+    },
+    {
+      "epoch": 58.42,
+      "learning_rate": 0.0001309326923076923,
+      "loss": 0.6801,
+      "step": 6368
+    },
+    {
+      "epoch": 58.43,
+      "learning_rate": 0.00013090384615384613,
+      "loss": 0.6797,
+      "step": 6369
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 0.000130875,
+      "loss": 0.6342,
+      "step": 6370
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 0.00013084615384615383,
+      "loss": 0.6187,
+      "step": 6371
+    },
+    {
+      "epoch": 58.46,
+      "learning_rate": 0.00013081730769230767,
+      "loss": 0.5399,
+      "step": 6372
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 0.00013078846153846154,
+      "loss": 0.7584,
+      "step": 6373
+    },
+    {
+      "epoch": 58.48,
+      "learning_rate": 0.00013075961538461538,
+      "loss": 0.6786,
+      "step": 6374
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 0.00013073076923076921,
+      "loss": 0.7984,
+      "step": 6375
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 0.00013070192307692308,
+      "loss": 0.7479,
+      "step": 6376
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 0.00013067307692307692,
+      "loss": 0.9003,
+      "step": 6377
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 0.00013064423076923076,
+      "loss": 0.7163,
+      "step": 6378
+    },
+    {
+      "epoch": 58.52,
+      "learning_rate": 0.0001306153846153846,
+      "loss": 0.7857,
+      "step": 6379
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 0.00013058653846153843,
+      "loss": 0.6131,
+      "step": 6380
+    },
+    {
+      "epoch": 58.54,
+      "learning_rate": 0.0001305576923076923,
+      "loss": 0.4631,
+      "step": 6381
+    },
+    {
+      "epoch": 58.55,
+      "learning_rate": 0.00013052884615384614,
+      "loss": 0.827,
+      "step": 6382
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 0.0001305,
+      "loss": 0.7242,
+      "step": 6383
+    },
+    {
+      "epoch": 58.57,
+      "learning_rate": 0.00013047115384615384,
+      "loss": 0.7411,
+      "step": 6384
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 0.00013044230769230768,
+      "loss": 0.7731,
+      "step": 6385
+    },
+    {
+      "epoch": 58.59,
+      "learning_rate": 0.00013041346153846154,
+      "loss": 0.6555,
+      "step": 6386
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 0.00013038461538461538,
+      "loss": 0.6981,
+      "step": 6387
+    },
+    {
+      "epoch": 58.61,
+      "learning_rate": 0.00013035576923076922,
+      "loss": 0.6721,
+      "step": 6388
+    },
+    {
+      "epoch": 58.61,
+      "learning_rate": 0.00013032692307692306,
+      "loss": 0.7052,
+      "step": 6389
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 0.0001302980769230769,
+      "loss": 0.6893,
+      "step": 6390
+    },
+    {
+      "epoch": 58.63,
+      "learning_rate": 0.00013026923076923076,
+      "loss": 0.7866,
+      "step": 6391
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 0.0001302403846153846,
+      "loss": 0.6832,
+      "step": 6392
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 0.00013021153846153846,
+      "loss": 0.6611,
+      "step": 6393
+    },
+    {
+      "epoch": 58.66,
+      "learning_rate": 0.0001301826923076923,
+      "loss": 0.6241,
+      "step": 6394
+    },
+    {
+      "epoch": 58.67,
+      "learning_rate": 0.00013015384615384614,
+      "loss": 0.696,
+      "step": 6395
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 0.000130125,
+      "loss": 0.7644,
+      "step": 6396
+    },
+    {
+      "epoch": 58.69,
+      "learning_rate": 0.00013009615384615384,
+      "loss": 0.6059,
+      "step": 6397
+    },
+    {
+      "epoch": 58.7,
+      "learning_rate": 0.00013006730769230768,
+      "loss": 0.7211,
+      "step": 6398
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 0.00013003846153846152,
+      "loss": 0.6769,
+      "step": 6399
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 0.00013000961538461536,
+      "loss": 0.708,
+      "step": 6400
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 0.00012998076923076922,
+      "loss": 0.6853,
+      "step": 6401
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 0.00012995192307692306,
+      "loss": 0.8181,
+      "step": 6402
+    },
+    {
+      "epoch": 58.74,
+      "learning_rate": 0.00012992307692307693,
+      "loss": 0.7849,
+      "step": 6403
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 0.00012989423076923076,
+      "loss": 0.6688,
+      "step": 6404
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 0.0001298653846153846,
+      "loss": 0.6751,
+      "step": 6405
+    },
+    {
+      "epoch": 58.77,
+      "learning_rate": 0.00012983653846153844,
+      "loss": 0.5928,
+      "step": 6406
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 0.0001298076923076923,
+      "loss": 0.7552,
+      "step": 6407
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 0.00012977884615384614,
+      "loss": 0.7238,
+      "step": 6408
+    },
+    {
+      "epoch": 58.8,
+      "learning_rate": 0.00012974999999999998,
+      "loss": 0.8016,
+      "step": 6409
+    },
+    {
+      "epoch": 58.81,
+      "learning_rate": 0.00012972115384615382,
+      "loss": 0.6794,
+      "step": 6410
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 0.00012969230769230769,
+      "loss": 0.7144,
+      "step": 6411
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 0.00012966346153846152,
+      "loss": 0.6353,
+      "step": 6412
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 0.00012963461538461536,
+      "loss": 0.721,
+      "step": 6413
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 0.00012960576923076923,
+      "loss": 0.6879,
+      "step": 6414
+    },
+    {
+      "epoch": 58.85,
+      "learning_rate": 0.00012957692307692307,
+      "loss": 0.7669,
+      "step": 6415
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 0.0001295480769230769,
+      "loss": 0.7196,
+      "step": 6416
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 0.00012951923076923077,
+      "loss": 0.6782,
+      "step": 6417
+    },
+    {
+      "epoch": 58.88,
+      "learning_rate": 0.0001294903846153846,
+      "loss": 0.7009,
+      "step": 6418
+    },
+    {
+      "epoch": 58.89,
+      "learning_rate": 0.00012946153846153845,
+      "loss": 0.8264,
+      "step": 6419
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 0.00012943269230769228,
+      "loss": 0.7412,
+      "step": 6420
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 0.00012940384615384615,
+      "loss": 0.8151,
+      "step": 6421
+    },
+    {
+      "epoch": 58.92,
+      "learning_rate": 0.000129375,
+      "loss": 0.6919,
+      "step": 6422
+    },
+    {
+      "epoch": 58.93,
+      "learning_rate": 0.00012934615384615383,
+      "loss": 0.7991,
+      "step": 6423
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 0.0001293173076923077,
+      "loss": 0.7674,
+      "step": 6424
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 0.00012928846153846153,
+      "loss": 0.8105,
+      "step": 6425
+    },
+    {
+      "epoch": 58.95,
+      "learning_rate": 0.00012925961538461537,
+      "loss": 0.7344,
+      "step": 6426
+    },
+    {
+      "epoch": 58.96,
+      "learning_rate": 0.00012923076923076923,
+      "loss": 0.6499,
+      "step": 6427
+    },
+    {
+      "epoch": 58.97,
+      "learning_rate": 0.00012920192307692307,
+      "loss": 0.773,
+      "step": 6428
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 0.0001291730769230769,
+      "loss": 0.5908,
+      "step": 6429
+    },
+    {
+      "epoch": 58.99,
+      "learning_rate": 0.00012914423076923075,
+      "loss": 0.616,
+      "step": 6430
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 0.0001291153846153846,
+      "loss": 0.6376,
+      "step": 6431
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 0.00012908653846153845,
+      "loss": 0.6712,
+      "step": 6432
+    },
+    {
+      "epoch": 59.02,
+      "learning_rate": 0.0001290576923076923,
+      "loss": 0.7875,
+      "step": 6433
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 0.00012902884615384615,
+      "loss": 0.6567,
+      "step": 6434
+    },
+    {
+      "epoch": 59.04,
+      "learning_rate": 0.000129,
+      "loss": 0.7469,
+      "step": 6435
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 0.00012897115384615383,
+      "loss": 0.7347,
+      "step": 6436
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 0.0001289423076923077,
+      "loss": 0.7266,
+      "step": 6437
+    },
+    {
+      "epoch": 59.06,
+      "learning_rate": 0.00012891346153846153,
+      "loss": 0.6752,
+      "step": 6438
+    },
+    {
+      "epoch": 59.07,
+      "learning_rate": 0.00012888461538461537,
+      "loss": 0.5967,
+      "step": 6439
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 0.0001288557692307692,
+      "loss": 0.782,
+      "step": 6440
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 0.00012882692307692305,
+      "loss": 0.6663,
+      "step": 6441
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 0.0001287980769230769,
+      "loss": 0.6654,
+      "step": 6442
+    },
+    {
+      "epoch": 59.11,
+      "learning_rate": 0.00012876923076923075,
+      "loss": 0.7903,
+      "step": 6443
+    },
+    {
+      "epoch": 59.12,
+      "learning_rate": 0.00012874038461538462,
+      "loss": 0.8259,
+      "step": 6444
+    },
+    {
+      "epoch": 59.13,
+      "learning_rate": 0.00012871153846153845,
+      "loss": 0.7494,
+      "step": 6445
+    },
+    {
+      "epoch": 59.14,
+      "learning_rate": 0.0001286826923076923,
+      "loss": 0.7194,
+      "step": 6446
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 0.00012865384615384613,
+      "loss": 0.6607,
+      "step": 6447
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 0.000128625,
+      "loss": 0.7064,
+      "step": 6448
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 0.00012859615384615383,
+      "loss": 0.6429,
+      "step": 6449
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 0.00012856730769230767,
+      "loss": 0.7299,
+      "step": 6450
+    },
+    {
+      "epoch": 59.18,
+      "learning_rate": 0.0001285384615384615,
+      "loss": 0.5782,
+      "step": 6451
+    },
+    {
+      "epoch": 59.19,
+      "learning_rate": 0.00012850961538461538,
+      "loss": 0.758,
+      "step": 6452
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 0.00012848076923076921,
+      "loss": 0.7028,
+      "step": 6453
+    },
+    {
+      "epoch": 59.21,
+      "learning_rate": 0.00012845192307692308,
+      "loss": 0.6892,
+      "step": 6454
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 0.00012842307692307692,
+      "loss": 0.7273,
+      "step": 6455
+    },
+    {
+      "epoch": 59.23,
+      "learning_rate": 0.00012839423076923076,
+      "loss": 0.6908,
+      "step": 6456
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 0.0001283653846153846,
+      "loss": 0.7227,
+      "step": 6457
+    },
+    {
+      "epoch": 59.25,
+      "learning_rate": 0.00012833653846153846,
+      "loss": 0.6142,
+      "step": 6458
+    },
+    {
+      "epoch": 59.26,
+      "learning_rate": 0.0001283076923076923,
+      "loss": 0.6864,
+      "step": 6459
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 0.00012827884615384614,
+      "loss": 0.6779,
+      "step": 6460
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 0.00012824999999999997,
+      "loss": 0.7541,
+      "step": 6461
+    },
+    {
+      "epoch": 59.28,
+      "learning_rate": 0.00012822115384615384,
+      "loss": 0.7938,
+      "step": 6462
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 0.00012819230769230768,
+      "loss": 0.73,
+      "step": 6463
+    },
+    {
+      "epoch": 59.3,
+      "learning_rate": 0.00012816346153846154,
+      "loss": 0.6656,
+      "step": 6464
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 0.00012813461538461538,
+      "loss": 0.7116,
+      "step": 6465
+    },
+    {
+      "epoch": 59.32,
+      "learning_rate": 0.00012810576923076922,
+      "loss": 0.6756,
+      "step": 6466
+    },
+    {
+      "epoch": 59.33,
+      "learning_rate": 0.00012807692307692306,
+      "loss": 0.6261,
+      "step": 6467
+    },
+    {
+      "epoch": 59.34,
+      "learning_rate": 0.00012804807692307692,
+      "loss": 0.7057,
+      "step": 6468
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 0.00012801923076923076,
+      "loss": 0.7788,
+      "step": 6469
+    },
+    {
+      "epoch": 59.36,
+      "learning_rate": 0.0001279903846153846,
+      "loss": 0.72,
+      "step": 6470
+    },
+    {
+      "epoch": 59.37,
+      "learning_rate": 0.00012796153846153844,
+      "loss": 0.6644,
+      "step": 6471
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 0.0001279326923076923,
+      "loss": 0.7034,
+      "step": 6472
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 0.00012790384615384614,
+      "loss": 0.7159,
+      "step": 6473
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 0.000127875,
+      "loss": 0.6316,
+      "step": 6474
+    },
+    {
+      "epoch": 59.4,
+      "learning_rate": 0.00012784615384615384,
+      "loss": 0.6463,
+      "step": 6475
+    },
+    {
+      "epoch": 59.41,
+      "learning_rate": 0.00012781730769230768,
+      "loss": 0.706,
+      "step": 6476
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 0.00012778846153846152,
+      "loss": 0.6967,
+      "step": 6477
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 0.00012775961538461536,
+      "loss": 0.651,
+      "step": 6478
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 0.00012773076923076922,
+      "loss": 0.756,
+      "step": 6479
+    },
+    {
+      "epoch": 59.45,
+      "learning_rate": 0.00012770192307692306,
+      "loss": 0.6835,
+      "step": 6480
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 0.0001276730769230769,
+      "loss": 0.7407,
+      "step": 6481
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 0.00012764423076923076,
+      "loss": 0.6482,
+      "step": 6482
+    },
+    {
+      "epoch": 59.48,
+      "learning_rate": 0.0001276153846153846,
+      "loss": 0.6816,
+      "step": 6483
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 0.00012758653846153847,
+      "loss": 0.5678,
+      "step": 6484
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 0.0001275576923076923,
+      "loss": 0.8117,
+      "step": 6485
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 0.00012752884615384614,
+      "loss": 0.651,
+      "step": 6486
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 0.00012749999999999998,
+      "loss": 0.6699,
+      "step": 6487
+    },
+    {
+      "epoch": 59.52,
+      "learning_rate": 0.00012747115384615382,
+      "loss": 0.7557,
+      "step": 6488
+    },
+    {
+      "epoch": 59.53,
+      "learning_rate": 0.00012744230769230769,
+      "loss": 0.7253,
+      "step": 6489
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 0.00012741346153846152,
+      "loss": 0.8648,
+      "step": 6490
+    },
+    {
+      "epoch": 59.55,
+      "learning_rate": 0.00012738461538461536,
+      "loss": 0.6079,
+      "step": 6491
+    },
+    {
+      "epoch": 59.56,
+      "learning_rate": 0.00012735576923076923,
+      "loss": 0.6596,
+      "step": 6492
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 0.00012732692307692307,
+      "loss": 0.7785,
+      "step": 6493
+    },
+    {
+      "epoch": 59.58,
+      "learning_rate": 0.00012729807692307693,
+      "loss": 0.7771,
+      "step": 6494
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 0.00012726923076923077,
+      "loss": 0.726,
+      "step": 6495
+    },
+    {
+      "epoch": 59.6,
+      "learning_rate": 0.0001272403846153846,
+      "loss": 0.6247,
+      "step": 6496
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 0.00012721153846153844,
+      "loss": 0.7426,
+      "step": 6497
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 0.00012718269230769228,
+      "loss": 0.7596,
+      "step": 6498
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 0.00012715384615384615,
+      "loss": 0.7857,
+      "step": 6499
+    },
+    {
+      "epoch": 59.63,
+      "learning_rate": 0.000127125,
+      "loss": 0.7901,
+      "step": 6500
+    },
+    {
+      "epoch": 59.63,
+      "eval_cer": 0.09520382614432,
+      "eval_loss": 0.3574734628200531,
+      "eval_runtime": 14.1374,
+      "eval_samples_per_second": 116.499,
+      "eval_steps_per_second": 1.839,
+      "eval_wer": 0.3837197426207742,
+      "step": 6500
+    },
+    {
+      "epoch": 59.64,
+      "learning_rate": 0.00012709615384615382,
+      "loss": 0.7247,
+      "step": 6501
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 0.0001270673076923077,
+      "loss": 0.6624,
+      "step": 6502
+    },
+    {
+      "epoch": 59.66,
+      "learning_rate": 0.00012703846153846153,
+      "loss": 0.6397,
+      "step": 6503
+    },
+    {
+      "epoch": 59.67,
+      "learning_rate": 0.00012700961538461537,
+      "loss": 0.7047,
+      "step": 6504
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 0.00012698076923076923,
+      "loss": 0.7246,
+      "step": 6505
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 0.00012695192307692307,
+      "loss": 0.6059,
+      "step": 6506
+    },
+    {
+      "epoch": 59.7,
+      "learning_rate": 0.0001269230769230769,
+      "loss": 0.766,
+      "step": 6507
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 0.00012689423076923075,
+      "loss": 0.6585,
+      "step": 6508
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 0.00012686538461538458,
+      "loss": 0.7029,
+      "step": 6509
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 0.00012683653846153845,
+      "loss": 0.8076,
+      "step": 6510
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 0.0001268076923076923,
+      "loss": 0.797,
+      "step": 6511
+    },
+    {
+      "epoch": 59.74,
+      "learning_rate": 0.00012677884615384615,
+      "loss": 0.7833,
+      "step": 6512
+    },
+    {
+      "epoch": 59.75,
+      "learning_rate": 0.00012675,
+      "loss": 0.6196,
+      "step": 6513
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 0.00012672115384615383,
+      "loss": 0.6785,
+      "step": 6514
+    },
+    {
+      "epoch": 59.77,
+      "learning_rate": 0.0001266923076923077,
+      "loss": 0.6953,
+      "step": 6515
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 0.00012666346153846153,
+      "loss": 0.6978,
+      "step": 6516
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 0.00012663461538461537,
+      "loss": 0.7463,
+      "step": 6517
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 0.0001266057692307692,
+      "loss": 0.6475,
+      "step": 6518
+    },
+    {
+      "epoch": 59.81,
+      "learning_rate": 0.00012657692307692305,
+      "loss": 0.7471,
+      "step": 6519
+    },
+    {
+      "epoch": 59.82,
+      "learning_rate": 0.0001265480769230769,
+      "loss": 0.7578,
+      "step": 6520
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 0.00012651923076923075,
+      "loss": 0.6959,
+      "step": 6521
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 0.00012649038461538462,
+      "loss": 0.6277,
+      "step": 6522
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 0.00012646153846153845,
+      "loss": 0.7087,
+      "step": 6523
+    },
+    {
+      "epoch": 59.85,
+      "learning_rate": 0.0001264326923076923,
+      "loss": 0.608,
+      "step": 6524
+    },
+    {
+      "epoch": 59.86,
+      "learning_rate": 0.00012640384615384616,
+      "loss": 0.6694,
+      "step": 6525
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 0.000126375,
+      "loss": 0.7149,
+      "step": 6526
+    },
+    {
+      "epoch": 59.88,
+      "learning_rate": 0.00012634615384615383,
+      "loss": 0.6089,
+      "step": 6527
+    },
+    {
+      "epoch": 59.89,
+      "learning_rate": 0.00012631730769230767,
+      "loss": 0.6392,
+      "step": 6528
+    },
+    {
+      "epoch": 59.9,
+      "learning_rate": 0.0001262884615384615,
+      "loss": 0.6999,
+      "step": 6529
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 0.00012625961538461538,
+      "loss": 0.7487,
+      "step": 6530
+    },
+    {
+      "epoch": 59.92,
+      "learning_rate": 0.0001262307692307692,
+      "loss": 0.7272,
+      "step": 6531
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 0.00012620192307692308,
+      "loss": 0.7052,
+      "step": 6532
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 0.00012617307692307692,
+      "loss": 0.7039,
+      "step": 6533
+    },
+    {
+      "epoch": 59.94,
+      "learning_rate": 0.00012614423076923075,
+      "loss": 0.6691,
+      "step": 6534
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 0.00012611538461538462,
+      "loss": 0.6021,
+      "step": 6535
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 0.00012608653846153846,
+      "loss": 0.6678,
+      "step": 6536
+    },
+    {
+      "epoch": 59.97,
+      "learning_rate": 0.0001260576923076923,
+      "loss": 0.7213,
+      "step": 6537
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 0.00012602884615384613,
+      "loss": 0.6979,
+      "step": 6538
+    },
+    {
+      "epoch": 59.99,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 0.6299,
+      "step": 6539
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 0.00012597115384615384,
+      "loss": 0.7577,
+      "step": 6540
+    },
+    {
+      "epoch": 60.01,
+      "learning_rate": 0.00012594230769230768,
+      "loss": 0.7508,
+      "step": 6541
+    },
+    {
+      "epoch": 60.02,
+      "learning_rate": 0.00012591346153846154,
+      "loss": 0.6513,
+      "step": 6542
+    },
+    {
+      "epoch": 60.03,
+      "learning_rate": 0.00012588461538461538,
+      "loss": 0.6136,
+      "step": 6543
+    },
+    {
+      "epoch": 60.04,
+      "learning_rate": 0.00012585576923076922,
+      "loss": 0.6489,
+      "step": 6544
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 0.00012582692307692306,
+      "loss": 0.624,
+      "step": 6545
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 0.00012579807692307692,
+      "loss": 0.7016,
+      "step": 6546
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 0.00012576923076923076,
+      "loss": 0.6663,
+      "step": 6547
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 0.0001257403846153846,
+      "loss": 0.5995,
+      "step": 6548
+    },
+    {
+      "epoch": 60.08,
+      "learning_rate": 0.00012571153846153844,
+      "loss": 0.6822,
+      "step": 6549
+    },
+    {
+      "epoch": 60.09,
+      "learning_rate": 0.0001256826923076923,
+      "loss": 0.608,
+      "step": 6550
+    },
+    {
+      "epoch": 60.1,
+      "learning_rate": 0.00012565384615384614,
+      "loss": 0.6925,
+      "step": 6551
+    },
+    {
+      "epoch": 60.11,
+      "learning_rate": 0.000125625,
+      "loss": 0.7113,
+      "step": 6552
+    },
+    {
+      "epoch": 60.12,
+      "learning_rate": 0.00012559615384615384,
+      "loss": 0.7161,
+      "step": 6553
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 0.00012556730769230768,
+      "loss": 0.6933,
+      "step": 6554
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 0.00012553846153846152,
+      "loss": 0.5676,
+      "step": 6555
+    },
+    {
+      "epoch": 60.15,
+      "learning_rate": 0.00012550961538461538,
+      "loss": 0.623,
+      "step": 6556
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 0.00012548076923076922,
+      "loss": 0.6608,
+      "step": 6557
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 0.00012545192307692306,
+      "loss": 0.7452,
+      "step": 6558
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 0.0001254230769230769,
+      "loss": 0.6451,
+      "step": 6559
+    },
+    {
+      "epoch": 60.18,
+      "learning_rate": 0.00012539423076923076,
+      "loss": 0.667,
+      "step": 6560
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 0.0001253653846153846,
+      "loss": 0.8104,
+      "step": 6561
+    },
+    {
+      "epoch": 60.2,
+      "learning_rate": 0.00012533653846153847,
+      "loss": 0.6148,
+      "step": 6562
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 0.0001253076923076923,
+      "loss": 0.6917,
+      "step": 6563
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 0.00012527884615384614,
+      "loss": 0.7217,
+      "step": 6564
+    },
+    {
+      "epoch": 60.23,
+      "learning_rate": 0.00012524999999999998,
+      "loss": 0.6838,
+      "step": 6565
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 0.00012522115384615385,
+      "loss": 0.7249,
+      "step": 6566
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 0.00012519230769230768,
+      "loss": 0.757,
+      "step": 6567
+    },
+    {
+      "epoch": 60.26,
+      "learning_rate": 0.00012516346153846152,
+      "loss": 0.7745,
+      "step": 6568
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 0.00012513461538461536,
+      "loss": 0.6906,
+      "step": 6569
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 0.00012510576923076923,
+      "loss": 0.6351,
+      "step": 6570
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 0.00012507692307692306,
+      "loss": 0.5973,
+      "step": 6571
+    },
+    {
+      "epoch": 60.29,
+      "learning_rate": 0.00012504807692307693,
+      "loss": 0.6632,
+      "step": 6572
+    },
+    {
+      "epoch": 60.3,
+      "learning_rate": 0.00012501923076923077,
+      "loss": 0.6711,
+      "step": 6573
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 0.0001249903846153846,
+      "loss": 0.633,
+      "step": 6574
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 0.00012496153846153844,
+      "loss": 0.6673,
+      "step": 6575
+    },
+    {
+      "epoch": 60.33,
+      "learning_rate": 0.00012493269230769228,
+      "loss": 0.6909,
+      "step": 6576
+    },
+    {
+      "epoch": 60.34,
+      "learning_rate": 0.00012490384615384615,
+      "loss": 0.5827,
+      "step": 6577
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 0.00012487499999999999,
+      "loss": 0.6951,
+      "step": 6578
+    },
+    {
+      "epoch": 60.36,
+      "learning_rate": 0.00012484615384615382,
+      "loss": 0.7493,
+      "step": 6579
+    },
+    {
+      "epoch": 60.37,
+      "learning_rate": 0.0001248173076923077,
+      "loss": 0.712,
+      "step": 6580
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 0.00012478846153846153,
+      "loss": 0.7818,
+      "step": 6581
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 0.0001247596153846154,
+      "loss": 0.5968,
+      "step": 6582
+    },
+    {
+      "epoch": 60.39,
+      "learning_rate": 0.00012473076923076923,
+      "loss": 0.5573,
+      "step": 6583
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 0.00012470192307692307,
+      "loss": 0.7596,
+      "step": 6584
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 0.0001246730769230769,
+      "loss": 0.6763,
+      "step": 6585
+    },
+    {
+      "epoch": 60.42,
+      "learning_rate": 0.00012464423076923075,
+      "loss": 0.6933,
+      "step": 6586
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 0.0001246153846153846,
+      "loss": 0.6632,
+      "step": 6587
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 0.00012458653846153845,
+      "loss": 0.6344,
+      "step": 6588
+    },
+    {
+      "epoch": 60.45,
+      "learning_rate": 0.0001245576923076923,
+      "loss": 0.8274,
+      "step": 6589
+    },
+    {
+      "epoch": 60.46,
+      "learning_rate": 0.00012452884615384615,
+      "loss": 0.6461,
+      "step": 6590
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 0.0001245,
+      "loss": 0.7277,
+      "step": 6591
+    },
+    {
+      "epoch": 60.48,
+      "learning_rate": 0.00012447115384615386,
+      "loss": 0.646,
+      "step": 6592
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 0.0001244423076923077,
+      "loss": 0.6845,
+      "step": 6593
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 0.00012441346153846153,
+      "loss": 0.7404,
+      "step": 6594
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 0.00012438461538461537,
+      "loss": 0.6145,
+      "step": 6595
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 0.0001243557692307692,
+      "loss": 0.6669,
+      "step": 6596
+    },
+    {
+      "epoch": 60.52,
+      "learning_rate": 0.00012432692307692307,
+      "loss": 0.7084,
+      "step": 6597
+    },
+    {
+      "epoch": 60.53,
+      "learning_rate": 0.0001242980769230769,
+      "loss": 0.7259,
+      "step": 6598
+    },
+    {
+      "epoch": 60.54,
+      "learning_rate": 0.00012426923076923075,
+      "loss": 0.6725,
+      "step": 6599
+    },
+    {
+      "epoch": 60.55,
+      "learning_rate": 0.00012424038461538462,
+      "loss": 0.6712,
+      "step": 6600
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 0.00012421153846153845,
+      "loss": 0.6907,
+      "step": 6601
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 0.0001241826923076923,
+      "loss": 0.7118,
+      "step": 6602
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 0.00012415384615384616,
+      "loss": 0.7078,
+      "step": 6603
+    },
+    {
+      "epoch": 60.59,
+      "learning_rate": 0.000124125,
+      "loss": 0.7318,
+      "step": 6604
+    },
+    {
+      "epoch": 60.6,
+      "learning_rate": 0.00012409615384615383,
+      "loss": 0.668,
+      "step": 6605
+    },
+    {
+      "epoch": 60.61,
+      "learning_rate": 0.00012406730769230767,
+      "loss": 0.6543,
+      "step": 6606
+    },
+    {
+      "epoch": 60.61,
+      "learning_rate": 0.0001240384615384615,
+      "loss": 0.6047,
+      "step": 6607
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 0.00012400961538461537,
+      "loss": 0.714,
+      "step": 6608
+    },
+    {
+      "epoch": 60.63,
+      "learning_rate": 0.0001239807692307692,
+      "loss": 0.6792,
+      "step": 6609
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 0.00012395192307692308,
+      "loss": 0.665,
+      "step": 6610
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 0.00012392307692307692,
+      "loss": 0.6727,
+      "step": 6611
+    },
+    {
+      "epoch": 60.66,
+      "learning_rate": 0.00012389423076923075,
+      "loss": 0.7579,
+      "step": 6612
+    },
+    {
+      "epoch": 60.67,
+      "learning_rate": 0.00012386538461538462,
+      "loss": 0.668,
+      "step": 6613
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 0.00012383653846153846,
+      "loss": 0.7967,
+      "step": 6614
+    },
+    {
+      "epoch": 60.69,
+      "learning_rate": 0.0001238076923076923,
+      "loss": 0.7211,
+      "step": 6615
+    },
+    {
+      "epoch": 60.7,
+      "learning_rate": 0.00012377884615384613,
+      "loss": 0.9025,
+      "step": 6616
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 0.00012374999999999997,
+      "loss": 0.6367,
+      "step": 6617
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 0.00012372115384615384,
+      "loss": 0.6918,
+      "step": 6618
+    },
+    {
+      "epoch": 60.72,
+      "learning_rate": 0.00012369230769230768,
+      "loss": 0.668,
+      "step": 6619
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 0.00012366346153846154,
+      "loss": 0.5528,
+      "step": 6620
+    },
+    {
+      "epoch": 60.74,
+      "learning_rate": 0.00012363461538461538,
+      "loss": 0.6676,
+      "step": 6621
+    },
+    {
+      "epoch": 60.75,
+      "learning_rate": 0.00012360576923076922,
+      "loss": 0.6874,
+      "step": 6622
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 0.00012357692307692308,
+      "loss": 0.6694,
+      "step": 6623
+    },
+    {
+      "epoch": 60.77,
+      "learning_rate": 0.00012354807692307692,
+      "loss": 0.6546,
+      "step": 6624
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 0.00012351923076923076,
+      "loss": 0.631,
+      "step": 6625
+    },
+    {
+      "epoch": 60.79,
+      "learning_rate": 0.0001234903846153846,
+      "loss": 0.6704,
+      "step": 6626
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 0.00012346153846153844,
+      "loss": 0.8265,
+      "step": 6627
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 0.0001234326923076923,
+      "loss": 0.7124,
+      "step": 6628
+    },
+    {
+      "epoch": 60.82,
+      "learning_rate": 0.00012340384615384614,
+      "loss": 0.7578,
+      "step": 6629
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 0.000123375,
+      "loss": 0.7202,
+      "step": 6630
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 0.00012334615384615384,
+      "loss": 0.5866,
+      "step": 6631
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 0.00012331730769230768,
+      "loss": 0.7541,
+      "step": 6632
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 0.00012328846153846155,
+      "loss": 0.6705,
+      "step": 6633
+    },
+    {
+      "epoch": 60.86,
+      "learning_rate": 0.00012325961538461538,
+      "loss": 0.5647,
+      "step": 6634
+    },
+    {
+      "epoch": 60.87,
+      "learning_rate": 0.00012323076923076922,
+      "loss": 0.7535,
+      "step": 6635
+    },
+    {
+      "epoch": 60.88,
+      "learning_rate": 0.00012320192307692306,
+      "loss": 0.7074,
+      "step": 6636
+    },
+    {
+      "epoch": 60.89,
+      "learning_rate": 0.0001231730769230769,
+      "loss": 0.6826,
+      "step": 6637
+    },
+    {
+      "epoch": 60.9,
+      "learning_rate": 0.00012314423076923076,
+      "loss": 0.7569,
+      "step": 6638
+    },
+    {
+      "epoch": 60.91,
+      "learning_rate": 0.0001231153846153846,
+      "loss": 0.6793,
+      "step": 6639
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 0.00012308653846153847,
+      "loss": 0.7714,
+      "step": 6640
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 0.0001230576923076923,
+      "loss": 0.6367,
+      "step": 6641
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 0.00012302884615384614,
+      "loss": 0.6822,
+      "step": 6642
+    },
+    {
+      "epoch": 60.94,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 0.6026,
+      "step": 6643
+    },
+    {
+      "epoch": 60.95,
+      "learning_rate": 0.00012297115384615385,
+      "loss": 0.6496,
+      "step": 6644
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 0.00012294230769230768,
+      "loss": 0.8566,
+      "step": 6645
+    },
+    {
+      "epoch": 60.97,
+      "learning_rate": 0.00012291346153846152,
+      "loss": 0.7423,
+      "step": 6646
+    },
+    {
+      "epoch": 60.98,
+      "learning_rate": 0.00012288461538461536,
+      "loss": 0.8245,
+      "step": 6647
+    },
+    {
+      "epoch": 60.99,
+      "learning_rate": 0.00012285576923076923,
+      "loss": 0.68,
+      "step": 6648
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 0.00012282692307692306,
+      "loss": 0.5841,
+      "step": 6649
+    },
+    {
+      "epoch": 61.01,
+      "learning_rate": 0.0001227980769230769,
+      "loss": 0.7332,
+      "step": 6650
+    },
+    {
+      "epoch": 61.02,
+      "learning_rate": 0.00012276923076923077,
+      "loss": 0.6285,
+      "step": 6651
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 0.0001227403846153846,
+      "loss": 0.6632,
+      "step": 6652
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 0.00012271153846153844,
+      "loss": 0.6922,
+      "step": 6653
+    },
+    {
+      "epoch": 61.05,
+      "learning_rate": 0.0001226826923076923,
+      "loss": 0.6906,
+      "step": 6654
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 0.00012265384615384615,
+      "loss": 0.7113,
+      "step": 6655
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 0.00012262499999999999,
+      "loss": 0.6649,
+      "step": 6656
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 0.00012259615384615382,
+      "loss": 0.5988,
+      "step": 6657
+    },
+    {
+      "epoch": 61.08,
+      "learning_rate": 0.0001225673076923077,
+      "loss": 0.5787,
+      "step": 6658
+    },
+    {
+      "epoch": 61.09,
+      "learning_rate": 0.00012253846153846153,
+      "loss": 0.7379,
+      "step": 6659
+    },
+    {
+      "epoch": 61.1,
+      "learning_rate": 0.00012250961538461537,
+      "loss": 0.5613,
+      "step": 6660
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 0.00012248076923076923,
+      "loss": 0.6382,
+      "step": 6661
+    },
+    {
+      "epoch": 61.12,
+      "learning_rate": 0.00012245192307692307,
+      "loss": 0.5678,
+      "step": 6662
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 0.0001224230769230769,
+      "loss": 0.7061,
+      "step": 6663
+    },
+    {
+      "epoch": 61.14,
+      "learning_rate": 0.00012239423076923077,
+      "loss": 0.6511,
+      "step": 6664
+    },
+    {
+      "epoch": 61.15,
+      "learning_rate": 0.0001223653846153846,
+      "loss": 0.7658,
+      "step": 6665
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 0.00012233653846153845,
+      "loss": 0.7134,
+      "step": 6666
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 0.0001223076923076923,
+      "loss": 0.6321,
+      "step": 6667
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 0.00012227884615384615,
+      "loss": 0.6661,
+      "step": 6668
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 0.00012225,
+      "loss": 0.726,
+      "step": 6669
+    },
+    {
+      "epoch": 61.19,
+      "learning_rate": 0.00012222115384615383,
+      "loss": 0.5639,
+      "step": 6670
+    },
+    {
+      "epoch": 61.2,
+      "learning_rate": 0.0001221923076923077,
+      "loss": 0.6403,
+      "step": 6671
+    },
+    {
+      "epoch": 61.21,
+      "learning_rate": 0.00012216346153846153,
+      "loss": 0.5987,
+      "step": 6672
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 0.00012213461538461537,
+      "loss": 0.6426,
+      "step": 6673
+    },
+    {
+      "epoch": 61.23,
+      "learning_rate": 0.0001221057692307692,
+      "loss": 0.6088,
+      "step": 6674
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 0.00012207692307692307,
+      "loss": 0.6898,
+      "step": 6675
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 0.00012204807692307691,
+      "loss": 0.6198,
+      "step": 6676
+    },
+    {
+      "epoch": 61.26,
+      "learning_rate": 0.00012201923076923076,
+      "loss": 0.6993,
+      "step": 6677
+    },
+    {
+      "epoch": 61.27,
+      "learning_rate": 0.0001219903846153846,
+      "loss": 0.6717,
+      "step": 6678
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 0.00012196153846153844,
+      "loss": 0.7637,
+      "step": 6679
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 0.0001219326923076923,
+      "loss": 0.7044,
+      "step": 6680
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 0.00012190384615384614,
+      "loss": 0.5729,
+      "step": 6681
+    },
+    {
+      "epoch": 61.3,
+      "learning_rate": 0.000121875,
+      "loss": 0.6534,
+      "step": 6682
+    },
+    {
+      "epoch": 61.31,
+      "learning_rate": 0.00012184615384615383,
+      "loss": 0.6787,
+      "step": 6683
+    },
+    {
+      "epoch": 61.32,
+      "learning_rate": 0.00012181730769230767,
+      "loss": 0.7443,
+      "step": 6684
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 0.00012178846153846154,
+      "loss": 0.6696,
+      "step": 6685
+    },
+    {
+      "epoch": 61.34,
+      "learning_rate": 0.00012175961538461537,
+      "loss": 0.6731,
+      "step": 6686
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 0.00012173076923076923,
+      "loss": 0.6252,
+      "step": 6687
+    },
+    {
+      "epoch": 61.36,
+      "learning_rate": 0.00012170192307692306,
+      "loss": 0.8355,
+      "step": 6688
+    },
+    {
+      "epoch": 61.37,
+      "learning_rate": 0.0001216730769230769,
+      "loss": 0.6464,
+      "step": 6689
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 0.00012164423076923077,
+      "loss": 0.7568,
+      "step": 6690
+    },
+    {
+      "epoch": 61.39,
+      "learning_rate": 0.0001216153846153846,
+      "loss": 0.7232,
+      "step": 6691
+    },
+    {
+      "epoch": 61.39,
+      "learning_rate": 0.00012158653846153846,
+      "loss": 0.7618,
+      "step": 6692
+    },
+    {
+      "epoch": 61.4,
+      "learning_rate": 0.0001215576923076923,
+      "loss": 0.6488,
+      "step": 6693
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 0.00012152884615384613,
+      "loss": 0.6735,
+      "step": 6694
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 0.0001215,
+      "loss": 0.6894,
+      "step": 6695
+    },
+    {
+      "epoch": 61.43,
+      "learning_rate": 0.00012147115384615384,
+      "loss": 0.6613,
+      "step": 6696
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 0.00012144230769230769,
+      "loss": 0.7317,
+      "step": 6697
+    },
+    {
+      "epoch": 61.45,
+      "learning_rate": 0.00012141346153846153,
+      "loss": 0.7178,
+      "step": 6698
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 0.00012138461538461537,
+      "loss": 0.6222,
+      "step": 6699
+    },
+    {
+      "epoch": 61.47,
+      "learning_rate": 0.00012135576923076922,
+      "loss": 0.6942,
+      "step": 6700
+    },
+    {
+      "epoch": 61.48,
+      "learning_rate": 0.00012132692307692307,
+      "loss": 0.7034,
+      "step": 6701
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 0.00012129807692307692,
+      "loss": 0.6496,
+      "step": 6702
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 0.00012126923076923076,
+      "loss": 0.6936,
+      "step": 6703
+    },
+    {
+      "epoch": 61.5,
+      "learning_rate": 0.0001212403846153846,
+      "loss": 0.6718,
+      "step": 6704
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 0.00012121153846153845,
+      "loss": 0.6544,
+      "step": 6705
+    },
+    {
+      "epoch": 61.52,
+      "learning_rate": 0.0001211826923076923,
+      "loss": 0.7092,
+      "step": 6706
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 0.00012115384615384615,
+      "loss": 0.671,
+      "step": 6707
+    },
+    {
+      "epoch": 61.54,
+      "learning_rate": 0.00012112499999999999,
+      "loss": 0.6443,
+      "step": 6708
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 0.00012109615384615383,
+      "loss": 0.7084,
+      "step": 6709
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 0.00012106730769230768,
+      "loss": 0.7351,
+      "step": 6710
+    },
+    {
+      "epoch": 61.57,
+      "learning_rate": 0.00012103846153846153,
+      "loss": 0.7171,
+      "step": 6711
+    },
+    {
+      "epoch": 61.58,
+      "learning_rate": 0.00012100961538461538,
+      "loss": 0.5426,
+      "step": 6712
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 0.00012098076923076922,
+      "loss": 0.6294,
+      "step": 6713
+    },
+    {
+      "epoch": 61.6,
+      "learning_rate": 0.00012095192307692306,
+      "loss": 0.7307,
+      "step": 6714
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 0.00012092307692307691,
+      "loss": 0.7549,
+      "step": 6715
+    },
+    {
+      "epoch": 61.61,
+      "learning_rate": 0.00012089423076923076,
+      "loss": 0.7478,
+      "step": 6716
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 0.00012086538461538461,
+      "loss": 0.6302,
+      "step": 6717
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 0.00012083653846153845,
+      "loss": 0.6276,
+      "step": 6718
+    },
+    {
+      "epoch": 61.64,
+      "learning_rate": 0.00012080769230769229,
+      "loss": 0.728,
+      "step": 6719
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 0.00012077884615384614,
+      "loss": 0.7202,
+      "step": 6720
+    },
+    {
+      "epoch": 61.66,
+      "learning_rate": 0.00012075,
+      "loss": 0.8905,
+      "step": 6721
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 0.00012072115384615385,
+      "loss": 0.7244,
+      "step": 6722
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 0.00012069230769230768,
+      "loss": 0.7178,
+      "step": 6723
+    },
+    {
+      "epoch": 61.69,
+      "learning_rate": 0.00012066346153846152,
+      "loss": 0.5702,
+      "step": 6724
+    },
+    {
+      "epoch": 61.7,
+      "learning_rate": 0.00012063461538461537,
+      "loss": 0.7782,
+      "step": 6725
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 0.00012060576923076923,
+      "loss": 0.5377,
+      "step": 6726
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 0.00012057692307692308,
+      "loss": 0.6406,
+      "step": 6727
+    },
+    {
+      "epoch": 61.72,
+      "learning_rate": 0.00012054807692307692,
+      "loss": 0.6016,
+      "step": 6728
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 0.00012051923076923075,
+      "loss": 0.6813,
+      "step": 6729
+    },
+    {
+      "epoch": 61.74,
+      "learning_rate": 0.0001204903846153846,
+      "loss": 0.6766,
+      "step": 6730
+    },
+    {
+      "epoch": 61.75,
+      "learning_rate": 0.00012046153846153844,
+      "loss": 0.73,
+      "step": 6731
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 0.00012043269230769231,
+      "loss": 0.7049,
+      "step": 6732
+    },
+    {
+      "epoch": 61.77,
+      "learning_rate": 0.00012040384615384615,
+      "loss": 0.6814,
+      "step": 6733
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 0.00012037499999999999,
+      "loss": 0.7165,
+      "step": 6734
+    },
+    {
+      "epoch": 61.79,
+      "learning_rate": 0.00012034615384615384,
+      "loss": 0.7495,
+      "step": 6735
+    },
+    {
+      "epoch": 61.8,
+      "learning_rate": 0.00012031730769230767,
+      "loss": 0.6233,
+      "step": 6736
+    },
+    {
+      "epoch": 61.81,
+      "learning_rate": 0.00012028846153846154,
+      "loss": 0.5637,
+      "step": 6737
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 0.00012025961538461538,
+      "loss": 0.7039,
+      "step": 6738
+    },
+    {
+      "epoch": 61.83,
+      "learning_rate": 0.00012023076923076922,
+      "loss": 0.6085,
+      "step": 6739
+    },
+    {
+      "epoch": 61.83,
+      "learning_rate": 0.00012020192307692307,
+      "loss": 0.6571,
+      "step": 6740
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 0.0001201730769230769,
+      "loss": 0.7223,
+      "step": 6741
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 0.00012014423076923077,
+      "loss": 0.8108,
+      "step": 6742
+    },
+    {
+      "epoch": 61.86,
+      "learning_rate": 0.00012011538461538461,
+      "loss": 0.7606,
+      "step": 6743
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 0.00012008653846153845,
+      "loss": 0.6994,
+      "step": 6744
+    },
+    {
+      "epoch": 61.88,
+      "learning_rate": 0.0001200576923076923,
+      "loss": 0.5849,
+      "step": 6745
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 0.00012002884615384614,
+      "loss": 0.7705,
+      "step": 6746
+    },
+    {
+      "epoch": 61.9,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 0.6715,
+      "step": 6747
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 0.00011997115384615384,
+      "loss": 0.6671,
+      "step": 6748
+    },
+    {
+      "epoch": 61.92,
+      "learning_rate": 0.00011994230769230768,
+      "loss": 0.6305,
+      "step": 6749
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 0.00011991346153846153,
+      "loss": 0.7009,
+      "step": 6750
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 0.00011988461538461537,
+      "loss": 0.6833,
+      "step": 6751
+    },
+    {
+      "epoch": 61.94,
+      "learning_rate": 0.00011985576923076922,
+      "loss": 0.528,
+      "step": 6752
+    },
+    {
+      "epoch": 61.95,
+      "learning_rate": 0.00011982692307692307,
+      "loss": 0.6859,
+      "step": 6753
+    },
+    {
+      "epoch": 61.96,
+      "learning_rate": 0.00011979807692307691,
+      "loss": 0.5703,
+      "step": 6754
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 0.00011976923076923076,
+      "loss": 0.6001,
+      "step": 6755
+    },
+    {
+      "epoch": 61.98,
+      "learning_rate": 0.0001197403846153846,
+      "loss": 0.8063,
+      "step": 6756
+    },
+    {
+      "epoch": 61.99,
+      "learning_rate": 0.00011971153846153845,
+      "loss": 0.7473,
+      "step": 6757
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 0.0001196826923076923,
+      "loss": 0.7197,
+      "step": 6758
+    },
+    {
+      "epoch": 62.01,
+      "learning_rate": 0.00011965384615384614,
+      "loss": 0.7543,
+      "step": 6759
+    },
+    {
+      "epoch": 62.02,
+      "learning_rate": 0.000119625,
+      "loss": 0.6614,
+      "step": 6760
+    },
+    {
+      "epoch": 62.03,
+      "learning_rate": 0.00011959615384615383,
+      "loss": 0.6661,
+      "step": 6761
+    },
+    {
+      "epoch": 62.04,
+      "learning_rate": 0.00011956730769230768,
+      "loss": 0.61,
+      "step": 6762
+    },
+    {
+      "epoch": 62.05,
+      "learning_rate": 0.00011953846153846154,
+      "loss": 0.5783,
+      "step": 6763
+    },
+    {
+      "epoch": 62.06,
+      "learning_rate": 0.00011950961538461537,
+      "loss": 0.6303,
+      "step": 6764
+    },
+    {
+      "epoch": 62.06,
+      "learning_rate": 0.00011948076923076923,
+      "loss": 0.564,
+      "step": 6765
+    },
+    {
+      "epoch": 62.07,
+      "learning_rate": 0.00011945192307692306,
+      "loss": 0.7237,
+      "step": 6766
+    },
+    {
+      "epoch": 62.08,
+      "learning_rate": 0.0001194230769230769,
+      "loss": 0.6842,
+      "step": 6767
+    },
+    {
+      "epoch": 62.09,
+      "learning_rate": 0.00011939423076923077,
+      "loss": 0.7441,
+      "step": 6768
+    },
+    {
+      "epoch": 62.1,
+      "learning_rate": 0.0001193653846153846,
+      "loss": 0.593,
+      "step": 6769
+    },
+    {
+      "epoch": 62.11,
+      "learning_rate": 0.00011933653846153846,
+      "loss": 0.6711,
+      "step": 6770
+    },
+    {
+      "epoch": 62.12,
+      "learning_rate": 0.0001193076923076923,
+      "loss": 0.6162,
+      "step": 6771
+    },
+    {
+      "epoch": 62.13,
+      "learning_rate": 0.00011927884615384613,
+      "loss": 0.6566,
+      "step": 6772
+    },
+    {
+      "epoch": 62.14,
+      "learning_rate": 0.00011925,
+      "loss": 0.6914,
+      "step": 6773
+    },
+    {
+      "epoch": 62.15,
+      "learning_rate": 0.00011922115384615384,
+      "loss": 0.7576,
+      "step": 6774
+    },
+    {
+      "epoch": 62.16,
+      "learning_rate": 0.00011919230769230767,
+      "loss": 0.7717,
+      "step": 6775
+    },
+    {
+      "epoch": 62.17,
+      "learning_rate": 0.00011916346153846153,
+      "loss": 0.6388,
+      "step": 6776
+    },
+    {
+      "epoch": 62.17,
+      "learning_rate": 0.00011913461538461536,
+      "loss": 0.7204,
+      "step": 6777
+    },
+    {
+      "epoch": 62.18,
+      "learning_rate": 0.00011910576923076923,
+      "loss": 0.6748,
+      "step": 6778
+    },
+    {
+      "epoch": 62.19,
+      "learning_rate": 0.00011907692307692307,
+      "loss": 0.6678,
+      "step": 6779
+    },
+    {
+      "epoch": 62.2,
+      "learning_rate": 0.0001190480769230769,
+      "loss": 0.6182,
+      "step": 6780
+    },
+    {
+      "epoch": 62.21,
+      "learning_rate": 0.00011901923076923076,
+      "loss": 0.6087,
+      "step": 6781
+    },
+    {
+      "epoch": 62.22,
+      "learning_rate": 0.0001189903846153846,
+      "loss": 0.7321,
+      "step": 6782
+    },
+    {
+      "epoch": 62.23,
+      "learning_rate": 0.00011896153846153846,
+      "loss": 0.6331,
+      "step": 6783
+    },
+    {
+      "epoch": 62.24,
+      "learning_rate": 0.0001189326923076923,
+      "loss": 0.6237,
+      "step": 6784
+    },
+    {
+      "epoch": 62.25,
+      "learning_rate": 0.00011890384615384614,
+      "loss": 0.5599,
+      "step": 6785
+    },
+    {
+      "epoch": 62.26,
+      "learning_rate": 0.00011887499999999999,
+      "loss": 0.6404,
+      "step": 6786
+    },
+    {
+      "epoch": 62.27,
+      "learning_rate": 0.00011884615384615383,
+      "loss": 0.6507,
+      "step": 6787
+    },
+    {
+      "epoch": 62.28,
+      "learning_rate": 0.00011881730769230769,
+      "loss": 0.6777,
+      "step": 6788
+    },
+    {
+      "epoch": 62.28,
+      "learning_rate": 0.00011878846153846153,
+      "loss": 0.5925,
+      "step": 6789
+    },
+    {
+      "epoch": 62.29,
+      "learning_rate": 0.00011875961538461537,
+      "loss": 0.7019,
+      "step": 6790
+    },
+    {
+      "epoch": 62.3,
+      "learning_rate": 0.00011873076923076922,
+      "loss": 0.6713,
+      "step": 6791
+    },
+    {
+      "epoch": 62.31,
+      "learning_rate": 0.00011870192307692306,
+      "loss": 0.6884,
+      "step": 6792
+    },
+    {
+      "epoch": 62.32,
+      "learning_rate": 0.00011867307692307692,
+      "loss": 0.6142,
+      "step": 6793
+    },
+    {
+      "epoch": 62.33,
+      "learning_rate": 0.00011864423076923076,
+      "loss": 0.6645,
+      "step": 6794
+    },
+    {
+      "epoch": 62.34,
+      "learning_rate": 0.0001186153846153846,
+      "loss": 0.6743,
+      "step": 6795
+    },
+    {
+      "epoch": 62.35,
+      "learning_rate": 0.00011858653846153845,
+      "loss": 0.7713,
+      "step": 6796
+    },
+    {
+      "epoch": 62.36,
+      "learning_rate": 0.00011855769230769229,
+      "loss": 0.7394,
+      "step": 6797
+    },
+    {
+      "epoch": 62.37,
+      "learning_rate": 0.00011852884615384614,
+      "loss": 0.5682,
+      "step": 6798
+    },
+    {
+      "epoch": 62.38,
+      "learning_rate": 0.0001185,
+      "loss": 0.735,
+      "step": 6799
+    },
+    {
+      "epoch": 62.39,
+      "learning_rate": 0.00011847115384615383,
+      "loss": 0.7167,
+      "step": 6800
+    },
+    {
+      "epoch": 62.39,
+      "learning_rate": 0.00011844230769230768,
+      "loss": 0.7276,
+      "step": 6801
+    },
+    {
+      "epoch": 62.4,
+      "learning_rate": 0.00011841346153846152,
+      "loss": 0.6002,
+      "step": 6802
+    },
+    {
+      "epoch": 62.41,
+      "learning_rate": 0.00011838461538461537,
+      "loss": 0.8182,
+      "step": 6803
+    },
+    {
+      "epoch": 62.42,
+      "learning_rate": 0.00011835576923076923,
+      "loss": 0.5917,
+      "step": 6804
+    },
+    {
+      "epoch": 62.43,
+      "learning_rate": 0.00011832692307692306,
+      "loss": 0.5362,
+      "step": 6805
+    },
+    {
+      "epoch": 62.44,
+      "learning_rate": 0.00011829807692307691,
+      "loss": 0.6491,
+      "step": 6806
+    },
+    {
+      "epoch": 62.45,
+      "learning_rate": 0.00011826923076923075,
+      "loss": 0.691,
+      "step": 6807
+    },
+    {
+      "epoch": 62.46,
+      "learning_rate": 0.00011824038461538459,
+      "loss": 0.8123,
+      "step": 6808
+    },
+    {
+      "epoch": 62.47,
+      "learning_rate": 0.00011821153846153846,
+      "loss": 0.6243,
+      "step": 6809
+    },
+    {
+      "epoch": 62.48,
+      "learning_rate": 0.0001181826923076923,
+      "loss": 0.6729,
+      "step": 6810
+    },
+    {
+      "epoch": 62.49,
+      "learning_rate": 0.00011815384615384615,
+      "loss": 0.5788,
+      "step": 6811
+    },
+    {
+      "epoch": 62.5,
+      "learning_rate": 0.00011812499999999998,
+      "loss": 0.6619,
+      "step": 6812
+    },
+    {
+      "epoch": 62.5,
+      "learning_rate": 0.00011809615384615382,
+      "loss": 0.7402,
+      "step": 6813
+    },
+    {
+      "epoch": 62.51,
+      "learning_rate": 0.00011806730769230769,
+      "loss": 0.7842,
+      "step": 6814
+    },
+    {
+      "epoch": 62.52,
+      "learning_rate": 0.00011803846153846153,
+      "loss": 0.7637,
+      "step": 6815
+    },
+    {
+      "epoch": 62.53,
+      "learning_rate": 0.00011800961538461538,
+      "loss": 0.7414,
+      "step": 6816
+    },
+    {
+      "epoch": 62.54,
+      "learning_rate": 0.00011798076923076922,
+      "loss": 0.6475,
+      "step": 6817
+    },
+    {
+      "epoch": 62.55,
+      "learning_rate": 0.00011795192307692305,
+      "loss": 0.6788,
+      "step": 6818
+    },
+    {
+      "epoch": 62.56,
+      "learning_rate": 0.00011792307692307692,
+      "loss": 0.6763,
+      "step": 6819
+    },
+    {
+      "epoch": 62.57,
+      "learning_rate": 0.00011789423076923076,
+      "loss": 0.4721,
+      "step": 6820
+    },
+    {
+      "epoch": 62.58,
+      "learning_rate": 0.00011786538461538461,
+      "loss": 0.6642,
+      "step": 6821
+    },
+    {
+      "epoch": 62.59,
+      "learning_rate": 0.00011783653846153845,
+      "loss": 0.7094,
+      "step": 6822
+    },
+    {
+      "epoch": 62.6,
+      "learning_rate": 0.00011780769230769229,
+      "loss": 0.6889,
+      "step": 6823
+    },
+    {
+      "epoch": 62.61,
+      "learning_rate": 0.00011777884615384615,
+      "loss": 0.6837,
+      "step": 6824
+    },
+    {
+      "epoch": 62.61,
+      "learning_rate": 0.00011774999999999999,
+      "loss": 0.6908,
+      "step": 6825
+    },
+    {
+      "epoch": 62.62,
+      "learning_rate": 0.00011772115384615384,
+      "loss": 0.5608,
+      "step": 6826
+    },
+    {
+      "epoch": 62.63,
+      "learning_rate": 0.00011769230769230768,
+      "loss": 0.6326,
+      "step": 6827
+    },
+    {
+      "epoch": 62.64,
+      "learning_rate": 0.00011766346153846152,
+      "loss": 0.7435,
+      "step": 6828
+    },
+    {
+      "epoch": 62.65,
+      "learning_rate": 0.00011763461538461537,
+      "loss": 0.5991,
+      "step": 6829
+    },
+    {
+      "epoch": 62.66,
+      "learning_rate": 0.00011760576923076922,
+      "loss": 0.6401,
+      "step": 6830
+    },
+    {
+      "epoch": 62.67,
+      "learning_rate": 0.00011757692307692307,
+      "loss": 0.7182,
+      "step": 6831
+    },
+    {
+      "epoch": 62.68,
+      "learning_rate": 0.00011754807692307691,
+      "loss": 0.7921,
+      "step": 6832
+    },
+    {
+      "epoch": 62.69,
+      "learning_rate": 0.00011751923076923075,
+      "loss": 0.6916,
+      "step": 6833
+    },
+    {
+      "epoch": 62.7,
+      "learning_rate": 0.0001174903846153846,
+      "loss": 0.6879,
+      "step": 6834
+    },
+    {
+      "epoch": 62.71,
+      "learning_rate": 0.00011746153846153845,
+      "loss": 0.6597,
+      "step": 6835
+    },
+    {
+      "epoch": 62.72,
+      "learning_rate": 0.0001174326923076923,
+      "loss": 0.7561,
+      "step": 6836
+    },
+    {
+      "epoch": 62.72,
+      "learning_rate": 0.00011740384615384614,
+      "loss": 0.6694,
+      "step": 6837
+    },
+    {
+      "epoch": 62.73,
+      "learning_rate": 0.00011737499999999998,
+      "loss": 0.8913,
+      "step": 6838
+    },
+    {
+      "epoch": 62.74,
+      "learning_rate": 0.00011734615384615383,
+      "loss": 0.6592,
+      "step": 6839
+    },
+    {
+      "epoch": 62.75,
+      "learning_rate": 0.00011731730769230768,
+      "loss": 0.7559,
+      "step": 6840
+    },
+    {
+      "epoch": 62.76,
+      "learning_rate": 0.00011728846153846154,
+      "loss": 0.7192,
+      "step": 6841
+    },
+    {
+      "epoch": 62.77,
+      "learning_rate": 0.00011725961538461537,
+      "loss": 0.5776,
+      "step": 6842
+    },
+    {
+      "epoch": 62.78,
+      "learning_rate": 0.00011723076923076921,
+      "loss": 0.7895,
+      "step": 6843
+    },
+    {
+      "epoch": 62.79,
+      "learning_rate": 0.00011720192307692306,
+      "loss": 0.7444,
+      "step": 6844
+    },
+    {
+      "epoch": 62.8,
+      "learning_rate": 0.00011717307692307691,
+      "loss": 0.6179,
+      "step": 6845
+    },
+    {
+      "epoch": 62.81,
+      "learning_rate": 0.00011714423076923077,
+      "loss": 0.7,
+      "step": 6846
+    },
+    {
+      "epoch": 62.82,
+      "learning_rate": 0.0001171153846153846,
+      "loss": 0.6627,
+      "step": 6847
+    },
+    {
+      "epoch": 62.83,
+      "learning_rate": 0.00011708653846153844,
+      "loss": 0.688,
+      "step": 6848
+    },
+    {
+      "epoch": 62.83,
+      "learning_rate": 0.0001170576923076923,
+      "loss": 0.5913,
+      "step": 6849
+    },
+    {
+      "epoch": 62.84,
+      "learning_rate": 0.00011702884615384615,
+      "loss": 0.6936,
+      "step": 6850
+    },
+    {
+      "epoch": 62.85,
+      "learning_rate": 0.000117,
+      "loss": 0.9006,
+      "step": 6851
+    },
+    {
+      "epoch": 62.86,
+      "learning_rate": 0.00011697115384615384,
+      "loss": 0.6581,
+      "step": 6852
+    },
+    {
+      "epoch": 62.87,
+      "learning_rate": 0.00011694230769230767,
+      "loss": 0.6402,
+      "step": 6853
+    },
+    {
+      "epoch": 62.88,
+      "learning_rate": 0.00011691346153846153,
+      "loss": 0.6639,
+      "step": 6854
+    },
+    {
+      "epoch": 62.89,
+      "learning_rate": 0.00011688461538461538,
+      "loss": 0.6196,
+      "step": 6855
+    },
+    {
+      "epoch": 62.9,
+      "learning_rate": 0.00011685576923076923,
+      "loss": 0.6918,
+      "step": 6856
+    },
+    {
+      "epoch": 62.91,
+      "learning_rate": 0.00011682692307692307,
+      "loss": 0.7132,
+      "step": 6857
+    },
+    {
+      "epoch": 62.92,
+      "learning_rate": 0.0001167980769230769,
+      "loss": 0.7993,
+      "step": 6858
+    },
+    {
+      "epoch": 62.93,
+      "learning_rate": 0.00011676923076923076,
+      "loss": 0.7305,
+      "step": 6859
+    },
+    {
+      "epoch": 62.94,
+      "learning_rate": 0.00011674038461538461,
+      "loss": 0.6099,
+      "step": 6860
+    },
+    {
+      "epoch": 62.94,
+      "learning_rate": 0.00011671153846153846,
+      "loss": 0.8041,
+      "step": 6861
+    },
+    {
+      "epoch": 62.95,
+      "learning_rate": 0.0001166826923076923,
+      "loss": 0.61,
+      "step": 6862
+    },
+    {
+      "epoch": 62.96,
+      "learning_rate": 0.00011665384615384614,
+      "loss": 0.6791,
+      "step": 6863
+    },
+    {
+      "epoch": 62.97,
+      "learning_rate": 0.00011662499999999999,
+      "loss": 0.7344,
+      "step": 6864
+    },
+    {
+      "epoch": 62.98,
+      "learning_rate": 0.00011659615384615383,
+      "loss": 0.7001,
+      "step": 6865
+    },
+    {
+      "epoch": 62.99,
+      "learning_rate": 0.00011656730769230769,
+      "loss": 0.7625,
+      "step": 6866
+    },
+    {
+      "epoch": 63.0,
+      "learning_rate": 0.00011653846153846153,
+      "loss": 0.7495,
+      "step": 6867
+    },
+    {
+      "epoch": 63.01,
+      "learning_rate": 0.00011650961538461537,
+      "loss": 0.559,
+      "step": 6868
+    },
+    {
+      "epoch": 63.02,
+      "learning_rate": 0.00011648076923076922,
+      "loss": 0.7277,
+      "step": 6869
+    },
+    {
+      "epoch": 63.03,
+      "learning_rate": 0.00011645192307692306,
+      "loss": 0.6265,
+      "step": 6870
+    },
+    {
+      "epoch": 63.04,
+      "learning_rate": 0.00011642307692307692,
+      "loss": 0.6326,
+      "step": 6871
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 0.00011639423076923076,
+      "loss": 0.7292,
+      "step": 6872
+    },
+    {
+      "epoch": 63.06,
+      "learning_rate": 0.0001163653846153846,
+      "loss": 0.6308,
+      "step": 6873
+    },
+    {
+      "epoch": 63.06,
+      "learning_rate": 0.00011633653846153845,
+      "loss": 0.7646,
+      "step": 6874
+    },
+    {
+      "epoch": 63.07,
+      "learning_rate": 0.00011630769230769229,
+      "loss": 0.6504,
+      "step": 6875
+    },
+    {
+      "epoch": 63.08,
+      "learning_rate": 0.00011627884615384616,
+      "loss": 0.7445,
+      "step": 6876
+    },
+    {
+      "epoch": 63.09,
+      "learning_rate": 0.00011624999999999999,
+      "loss": 0.7077,
+      "step": 6877
+    },
+    {
+      "epoch": 63.1,
+      "learning_rate": 0.00011622115384615383,
+      "loss": 0.6515,
+      "step": 6878
+    },
+    {
+      "epoch": 63.11,
+      "learning_rate": 0.00011619230769230768,
+      "loss": 0.6705,
+      "step": 6879
+    },
+    {
+      "epoch": 63.12,
+      "learning_rate": 0.00011616346153846152,
+      "loss": 0.6515,
+      "step": 6880
+    },
+    {
+      "epoch": 63.13,
+      "learning_rate": 0.00011613461538461539,
+      "loss": 0.6167,
+      "step": 6881
+    },
+    {
+      "epoch": 63.14,
+      "learning_rate": 0.00011610576923076922,
+      "loss": 0.6808,
+      "step": 6882
+    },
+    {
+      "epoch": 63.15,
+      "learning_rate": 0.00011607692307692306,
+      "loss": 0.675,
+      "step": 6883
+    },
+    {
+      "epoch": 63.16,
+      "learning_rate": 0.00011604807692307691,
+      "loss": 0.7235,
+      "step": 6884
+    },
+    {
+      "epoch": 63.17,
+      "learning_rate": 0.00011601923076923075,
+      "loss": 0.604,
+      "step": 6885
+    },
+    {
+      "epoch": 63.17,
+      "learning_rate": 0.00011599038461538462,
+      "loss": 0.8084,
+      "step": 6886
+    },
+    {
+      "epoch": 63.18,
+      "learning_rate": 0.00011596153846153846,
+      "loss": 0.6571,
+      "step": 6887
+    },
+    {
+      "epoch": 63.19,
+      "learning_rate": 0.0001159326923076923,
+      "loss": 0.576,
+      "step": 6888
+    },
+    {
+      "epoch": 63.2,
+      "learning_rate": 0.00011590384615384615,
+      "loss": 0.6838,
+      "step": 6889
+    },
+    {
+      "epoch": 63.21,
+      "learning_rate": 0.00011587499999999998,
+      "loss": 0.6925,
+      "step": 6890
+    },
+    {
+      "epoch": 63.22,
+      "learning_rate": 0.00011584615384615385,
+      "loss": 0.5733,
+      "step": 6891
+    },
+    {
+      "epoch": 63.23,
+      "learning_rate": 0.00011581730769230769,
+      "loss": 0.6317,
+      "step": 6892
+    },
+    {
+      "epoch": 63.24,
+      "learning_rate": 0.00011578846153846153,
+      "loss": 0.7083,
+      "step": 6893
+    },
+    {
+      "epoch": 63.25,
+      "learning_rate": 0.00011575961538461538,
+      "loss": 0.6499,
+      "step": 6894
+    },
+    {
+      "epoch": 63.26,
+      "learning_rate": 0.00011573076923076922,
+      "loss": 0.7143,
+      "step": 6895
+    },
+    {
+      "epoch": 63.27,
+      "learning_rate": 0.00011570192307692305,
+      "loss": 0.7443,
+      "step": 6896
+    },
+    {
+      "epoch": 63.28,
+      "learning_rate": 0.00011567307692307692,
+      "loss": 0.6652,
+      "step": 6897
+    },
+    {
+      "epoch": 63.28,
+      "learning_rate": 0.00011564423076923076,
+      "loss": 0.6999,
+      "step": 6898
+    },
+    {
+      "epoch": 63.29,
+      "learning_rate": 0.00011561538461538461,
+      "loss": 0.6629,
+      "step": 6899
+    },
+    {
+      "epoch": 63.3,
+      "learning_rate": 0.00011558653846153845,
+      "loss": 0.6994,
+      "step": 6900
+    },
+    {
+      "epoch": 63.31,
+      "learning_rate": 0.00011555769230769229,
+      "loss": 0.6922,
+      "step": 6901
+    },
+    {
+      "epoch": 63.32,
+      "learning_rate": 0.00011552884615384615,
+      "loss": 0.7769,
+      "step": 6902
+    },
+    {
+      "epoch": 63.33,
+      "learning_rate": 0.00011549999999999999,
+      "loss": 0.6512,
+      "step": 6903
+    },
+    {
+      "epoch": 63.34,
+      "learning_rate": 0.00011547115384615384,
+      "loss": 0.5558,
+      "step": 6904
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 0.00011544230769230768,
+      "loss": 0.7751,
+      "step": 6905
+    },
+    {
+      "epoch": 63.36,
+      "learning_rate": 0.00011541346153846152,
+      "loss": 0.7144,
+      "step": 6906
+    },
+    {
+      "epoch": 63.37,
+      "learning_rate": 0.00011538461538461538,
+      "loss": 0.7127,
+      "step": 6907
+    },
+    {
+      "epoch": 63.38,
+      "learning_rate": 0.00011535576923076922,
+      "loss": 0.7608,
+      "step": 6908
+    },
+    {
+      "epoch": 63.39,
+      "learning_rate": 0.00011532692307692307,
+      "loss": 0.6849,
+      "step": 6909
+    },
+    {
+      "epoch": 63.39,
+      "learning_rate": 0.00011529807692307691,
+      "loss": 0.7598,
+      "step": 6910
+    },
+    {
+      "epoch": 63.4,
+      "learning_rate": 0.00011526923076923075,
+      "loss": 0.7145,
+      "step": 6911
+    },
+    {
+      "epoch": 63.41,
+      "learning_rate": 0.00011524038461538461,
+      "loss": 0.6582,
+      "step": 6912
+    },
+    {
+      "epoch": 63.42,
+      "learning_rate": 0.00011521153846153845,
+      "loss": 0.7216,
+      "step": 6913
+    },
+    {
+      "epoch": 63.43,
+      "learning_rate": 0.0001151826923076923,
+      "loss": 0.6593,
+      "step": 6914
+    },
+    {
+      "epoch": 63.44,
+      "learning_rate": 0.00011515384615384614,
+      "loss": 0.6581,
+      "step": 6915
+    },
+    {
+      "epoch": 63.45,
+      "learning_rate": 0.00011512499999999998,
+      "loss": 0.6405,
+      "step": 6916
+    },
+    {
+      "epoch": 63.46,
+      "learning_rate": 0.00011509615384615384,
+      "loss": 0.6105,
+      "step": 6917
+    },
+    {
+      "epoch": 63.47,
+      "learning_rate": 0.00011506730769230768,
+      "loss": 0.6518,
+      "step": 6918
+    },
+    {
+      "epoch": 63.48,
+      "learning_rate": 0.00011503846153846153,
+      "loss": 0.6716,
+      "step": 6919
+    },
+    {
+      "epoch": 63.49,
+      "learning_rate": 0.00011500961538461537,
+      "loss": 0.6137,
+      "step": 6920
+    },
+    {
+      "epoch": 63.5,
+      "learning_rate": 0.00011498076923076921,
+      "loss": 0.7448,
+      "step": 6921
+    },
+    {
+      "epoch": 63.5,
+      "learning_rate": 0.00011495192307692308,
+      "loss": 0.6541,
+      "step": 6922
+    },
+    {
+      "epoch": 63.51,
+      "learning_rate": 0.00011492307692307691,
+      "loss": 0.6204,
+      "step": 6923
+    },
+    {
+      "epoch": 63.52,
+      "learning_rate": 0.00011489423076923077,
+      "loss": 0.6647,
+      "step": 6924
+    },
+    {
+      "epoch": 63.53,
+      "learning_rate": 0.0001148653846153846,
+      "loss": 0.7122,
+      "step": 6925
+    },
+    {
+      "epoch": 63.54,
+      "learning_rate": 0.00011483653846153844,
+      "loss": 0.8503,
+      "step": 6926
+    },
+    {
+      "epoch": 63.55,
+      "learning_rate": 0.0001148076923076923,
+      "loss": 0.7735,
+      "step": 6927
+    },
+    {
+      "epoch": 63.56,
+      "learning_rate": 0.00011477884615384615,
+      "loss": 0.6172,
+      "step": 6928
+    },
+    {
+      "epoch": 63.57,
+      "learning_rate": 0.00011475,
+      "loss": 0.6748,
+      "step": 6929
+    },
+    {
+      "epoch": 63.58,
+      "learning_rate": 0.00011472115384615384,
+      "loss": 0.6771,
+      "step": 6930
+    },
+    {
+      "epoch": 63.59,
+      "learning_rate": 0.00011469230769230767,
+      "loss": 0.6343,
+      "step": 6931
+    },
+    {
+      "epoch": 63.6,
+      "learning_rate": 0.00011466346153846153,
+      "loss": 0.4696,
+      "step": 6932
+    },
+    {
+      "epoch": 63.61,
+      "learning_rate": 0.00011463461538461538,
+      "loss": 0.6522,
+      "step": 6933
+    },
+    {
+      "epoch": 63.61,
+      "learning_rate": 0.00011460576923076923,
+      "loss": 0.6229,
+      "step": 6934
+    },
+    {
+      "epoch": 63.62,
+      "learning_rate": 0.00011457692307692307,
+      "loss": 0.5424,
+      "step": 6935
+    },
+    {
+      "epoch": 63.63,
+      "learning_rate": 0.0001145480769230769,
+      "loss": 0.6657,
+      "step": 6936
+    },
+    {
+      "epoch": 63.64,
+      "learning_rate": 0.00011451923076923076,
+      "loss": 0.6365,
+      "step": 6937
+    },
+    {
+      "epoch": 63.65,
+      "learning_rate": 0.00011449038461538461,
+      "loss": 0.6538,
+      "step": 6938
+    },
+    {
+      "epoch": 63.66,
+      "learning_rate": 0.00011446153846153846,
+      "loss": 0.704,
+      "step": 6939
+    },
+    {
+      "epoch": 63.67,
+      "learning_rate": 0.0001144326923076923,
+      "loss": 0.6489,
+      "step": 6940
+    },
+    {
+      "epoch": 63.68,
+      "learning_rate": 0.00011440384615384614,
+      "loss": 0.8483,
+      "step": 6941
+    },
+    {
+      "epoch": 63.69,
+      "learning_rate": 0.00011437499999999999,
+      "loss": 0.6281,
+      "step": 6942
+    },
+    {
+      "epoch": 63.7,
+      "learning_rate": 0.00011434615384615384,
+      "loss": 0.7028,
+      "step": 6943
+    },
+    {
+      "epoch": 63.71,
+      "learning_rate": 0.00011431730769230769,
+      "loss": 0.6904,
+      "step": 6944
+    },
+    {
+      "epoch": 63.72,
+      "learning_rate": 0.00011428846153846153,
+      "loss": 0.6687,
+      "step": 6945
+    },
+    {
+      "epoch": 63.72,
+      "learning_rate": 0.00011425961538461537,
+      "loss": 0.6265,
+      "step": 6946
+    },
+    {
+      "epoch": 63.73,
+      "learning_rate": 0.00011423076923076922,
+      "loss": 0.6665,
+      "step": 6947
+    },
+    {
+      "epoch": 63.74,
+      "learning_rate": 0.00011420192307692307,
+      "loss": 0.7082,
+      "step": 6948
+    },
+    {
+      "epoch": 63.75,
+      "learning_rate": 0.00011417307692307692,
+      "loss": 0.5554,
+      "step": 6949
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 0.00011414423076923076,
+      "loss": 0.6367,
+      "step": 6950
+    },
+    {
+      "epoch": 63.77,
+      "learning_rate": 0.0001141153846153846,
+      "loss": 0.7057,
+      "step": 6951
+    },
+    {
+      "epoch": 63.78,
+      "learning_rate": 0.00011408653846153845,
+      "loss": 0.6942,
+      "step": 6952
+    },
+    {
+      "epoch": 63.79,
+      "learning_rate": 0.0001140576923076923,
+      "loss": 0.769,
+      "step": 6953
+    },
+    {
+      "epoch": 63.8,
+      "learning_rate": 0.00011402884615384615,
+      "loss": 0.7003,
+      "step": 6954
+    },
+    {
+      "epoch": 63.81,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 0.7229,
+      "step": 6955
+    },
+    {
+      "epoch": 63.82,
+      "learning_rate": 0.00011397115384615383,
+      "loss": 0.6165,
+      "step": 6956
+    },
+    {
+      "epoch": 63.83,
+      "learning_rate": 0.00011394230769230768,
+      "loss": 0.7013,
+      "step": 6957
+    },
+    {
+      "epoch": 63.83,
+      "learning_rate": 0.00011391346153846153,
+      "loss": 0.6987,
+      "step": 6958
+    },
+    {
+      "epoch": 63.84,
+      "learning_rate": 0.00011388461538461539,
+      "loss": 0.6016,
+      "step": 6959
+    },
+    {
+      "epoch": 63.85,
+      "learning_rate": 0.00011385576923076922,
+      "loss": 0.6188,
+      "step": 6960
+    },
+    {
+      "epoch": 63.86,
+      "learning_rate": 0.00011382692307692306,
+      "loss": 0.7362,
+      "step": 6961
+    },
+    {
+      "epoch": 63.87,
+      "learning_rate": 0.00011379807692307691,
+      "loss": 0.5504,
+      "step": 6962
+    },
+    {
+      "epoch": 63.88,
+      "learning_rate": 0.00011376923076923075,
+      "loss": 0.6242,
+      "step": 6963
+    },
+    {
+      "epoch": 63.89,
+      "learning_rate": 0.00011374038461538462,
+      "loss": 0.7043,
+      "step": 6964
+    },
+    {
+      "epoch": 63.9,
+      "learning_rate": 0.00011371153846153846,
+      "loss": 0.6604,
+      "step": 6965
+    },
+    {
+      "epoch": 63.91,
+      "learning_rate": 0.0001136826923076923,
+      "loss": 0.7152,
+      "step": 6966
+    },
+    {
+      "epoch": 63.92,
+      "learning_rate": 0.00011365384615384615,
+      "loss": 0.7554,
+      "step": 6967
+    },
+    {
+      "epoch": 63.93,
+      "learning_rate": 0.00011362499999999998,
+      "loss": 0.6706,
+      "step": 6968
+    },
+    {
+      "epoch": 63.94,
+      "learning_rate": 0.00011359615384615385,
+      "loss": 0.6681,
+      "step": 6969
+    },
+    {
+      "epoch": 63.94,
+      "learning_rate": 0.00011356730769230769,
+      "loss": 0.6899,
+      "step": 6970
+    },
+    {
+      "epoch": 63.95,
+      "learning_rate": 0.00011353846153846153,
+      "loss": 0.5772,
+      "step": 6971
+    },
+    {
+      "epoch": 63.96,
+      "learning_rate": 0.00011350961538461538,
+      "loss": 0.6089,
+      "step": 6972
+    },
+    {
+      "epoch": 63.97,
+      "learning_rate": 0.00011348076923076922,
+      "loss": 0.7885,
+      "step": 6973
+    },
+    {
+      "epoch": 63.98,
+      "learning_rate": 0.00011345192307692308,
+      "loss": 0.6564,
+      "step": 6974
+    },
+    {
+      "epoch": 63.99,
+      "learning_rate": 0.00011342307692307692,
+      "loss": 0.6613,
+      "step": 6975
+    },
+    {
+      "epoch": 64.0,
+      "learning_rate": 0.00011339423076923076,
+      "loss": 0.5829,
+      "step": 6976
+    },
+    {
+      "epoch": 64.01,
+      "learning_rate": 0.00011336538461538461,
+      "loss": 0.638,
+      "step": 6977
+    },
+    {
+      "epoch": 64.02,
+      "learning_rate": 0.00011333653846153845,
+      "loss": 0.5925,
+      "step": 6978
+    },
+    {
+      "epoch": 64.03,
+      "learning_rate": 0.0001133076923076923,
+      "loss": 0.6467,
+      "step": 6979
+    },
+    {
+      "epoch": 64.04,
+      "learning_rate": 0.00011327884615384615,
+      "loss": 0.6485,
+      "step": 6980
+    },
+    {
+      "epoch": 64.05,
+      "learning_rate": 0.00011324999999999999,
+      "loss": 0.7239,
+      "step": 6981
+    },
+    {
+      "epoch": 64.06,
+      "learning_rate": 0.00011322115384615384,
+      "loss": 0.6042,
+      "step": 6982
+    },
+    {
+      "epoch": 64.06,
+      "learning_rate": 0.00011319230769230768,
+      "loss": 0.739,
+      "step": 6983
+    },
+    {
+      "epoch": 64.07,
+      "learning_rate": 0.00011316346153846153,
+      "loss": 0.6699,
+      "step": 6984
+    },
+    {
+      "epoch": 64.08,
+      "learning_rate": 0.00011313461538461538,
+      "loss": 0.6158,
+      "step": 6985
+    },
+    {
+      "epoch": 64.09,
+      "learning_rate": 0.00011310576923076922,
+      "loss": 0.6568,
+      "step": 6986
+    },
+    {
+      "epoch": 64.1,
+      "learning_rate": 0.00011307692307692307,
+      "loss": 0.5236,
+      "step": 6987
+    },
+    {
+      "epoch": 64.11,
+      "learning_rate": 0.00011304807692307691,
+      "loss": 0.6911,
+      "step": 6988
+    },
+    {
+      "epoch": 64.12,
+      "learning_rate": 0.00011301923076923076,
+      "loss": 0.7146,
+      "step": 6989
+    },
+    {
+      "epoch": 64.13,
+      "learning_rate": 0.00011299038461538461,
+      "loss": 0.7012,
+      "step": 6990
+    },
+    {
+      "epoch": 64.14,
+      "learning_rate": 0.00011296153846153845,
+      "loss": 0.7096,
+      "step": 6991
+    },
+    {
+      "epoch": 64.15,
+      "learning_rate": 0.0001129326923076923,
+      "loss": 0.6465,
+      "step": 6992
+    },
+    {
+      "epoch": 64.16,
+      "learning_rate": 0.00011290384615384614,
+      "loss": 0.7351,
+      "step": 6993
+    },
+    {
+      "epoch": 64.17,
+      "learning_rate": 0.00011287499999999998,
+      "loss": 0.7307,
+      "step": 6994
+    },
+    {
+      "epoch": 64.17,
+      "learning_rate": 0.00011284615384615384,
+      "loss": 0.7751,
+      "step": 6995
+    },
+    {
+      "epoch": 64.18,
+      "learning_rate": 0.00011281730769230768,
+      "loss": 0.6113,
+      "step": 6996
+    },
+    {
+      "epoch": 64.19,
+      "learning_rate": 0.00011278846153846153,
+      "loss": 0.6921,
+      "step": 6997
+    },
+    {
+      "epoch": 64.2,
+      "learning_rate": 0.00011275961538461537,
+      "loss": 0.6412,
+      "step": 6998
+    },
+    {
+      "epoch": 64.21,
+      "learning_rate": 0.00011273076923076921,
+      "loss": 0.6631,
+      "step": 6999
+    },
+    {
+      "epoch": 64.22,
+      "learning_rate": 0.00011270192307692308,
+      "loss": 0.778,
+      "step": 7000
+    },
+    {
+      "epoch": 64.22,
+      "eval_cer": 0.09284996776563367,
+      "eval_loss": 0.373200923204422,
+      "eval_runtime": 14.0009,
+      "eval_samples_per_second": 117.635,
+      "eval_steps_per_second": 1.857,
+      "eval_wer": 0.3790215504034317,
+      "step": 7000
+    },
+    {
+      "epoch": 64.23,
+      "learning_rate": 0.00011267307692307691,
+      "loss": 0.5305,
+      "step": 7001
+    },
+    {
+      "epoch": 64.24,
+      "learning_rate": 0.00011264423076923077,
+      "loss": 0.6806,
+      "step": 7002
+    },
+    {
+      "epoch": 64.25,
+      "learning_rate": 0.0001126153846153846,
+      "loss": 0.6452,
+      "step": 7003
+    },
+    {
+      "epoch": 64.26,
+      "learning_rate": 0.00011258653846153844,
+      "loss": 0.6326,
+      "step": 7004
+    },
+    {
+      "epoch": 64.27,
+      "learning_rate": 0.00011255769230769231,
+      "loss": 0.6738,
+      "step": 7005
+    },
+    {
+      "epoch": 64.28,
+      "learning_rate": 0.00011252884615384615,
+      "loss": 0.6564,
+      "step": 7006
+    },
+    {
+      "epoch": 64.28,
+      "learning_rate": 0.0001125,
+      "loss": 0.7645,
+      "step": 7007
+    },
+    {
+      "epoch": 64.29,
+      "learning_rate": 0.00011247115384615384,
+      "loss": 0.6462,
+      "step": 7008
+    },
+    {
+      "epoch": 64.3,
+      "learning_rate": 0.00011244230769230767,
+      "loss": 0.6737,
+      "step": 7009
+    },
+    {
+      "epoch": 64.31,
+      "learning_rate": 0.00011241346153846154,
+      "loss": 0.5754,
+      "step": 7010
+    },
+    {
+      "epoch": 64.32,
+      "learning_rate": 0.00011238461538461538,
+      "loss": 0.6318,
+      "step": 7011
+    },
+    {
+      "epoch": 64.33,
+      "learning_rate": 0.00011235576923076921,
+      "loss": 0.654,
+      "step": 7012
+    },
+    {
+      "epoch": 64.34,
+      "learning_rate": 0.00011232692307692307,
+      "loss": 0.8006,
+      "step": 7013
+    },
+    {
+      "epoch": 64.35,
+      "learning_rate": 0.0001122980769230769,
+      "loss": 0.7029,
+      "step": 7014
+    },
+    {
+      "epoch": 64.36,
+      "learning_rate": 0.00011226923076923077,
+      "loss": 0.6975,
+      "step": 7015
+    },
+    {
+      "epoch": 64.37,
+      "learning_rate": 0.00011224038461538461,
+      "loss": 0.6125,
+      "step": 7016
+    },
+    {
+      "epoch": 64.38,
+      "learning_rate": 0.00011221153846153845,
+      "loss": 0.7693,
+      "step": 7017
+    },
+    {
+      "epoch": 64.39,
+      "learning_rate": 0.0001121826923076923,
+      "loss": 0.6227,
+      "step": 7018
+    },
+    {
+      "epoch": 64.39,
+      "learning_rate": 0.00011215384615384614,
+      "loss": 0.7099,
+      "step": 7019
+    },
+    {
+      "epoch": 64.4,
+      "learning_rate": 0.000112125,
+      "loss": 0.6037,
+      "step": 7020
+    },
+    {
+      "epoch": 64.41,
+      "learning_rate": 0.00011209615384615384,
+      "loss": 0.7203,
+      "step": 7021
+    },
+    {
+      "epoch": 64.42,
+      "learning_rate": 0.00011206730769230768,
+      "loss": 0.6997,
+      "step": 7022
+    },
+    {
+      "epoch": 64.43,
+      "learning_rate": 0.00011203846153846153,
+      "loss": 0.7007,
+      "step": 7023
+    },
+    {
+      "epoch": 64.44,
+      "learning_rate": 0.00011200961538461537,
+      "loss": 0.7313,
+      "step": 7024
+    },
+    {
+      "epoch": 64.45,
+      "learning_rate": 0.00011198076923076922,
+      "loss": 0.5181,
+      "step": 7025
+    },
+    {
+      "epoch": 64.46,
+      "learning_rate": 0.00011195192307692307,
+      "loss": 0.604,
+      "step": 7026
+    },
+    {
+      "epoch": 64.47,
+      "learning_rate": 0.00011192307692307691,
+      "loss": 0.7023,
+      "step": 7027
+    },
+    {
+      "epoch": 64.48,
+      "learning_rate": 0.00011189423076923076,
+      "loss": 0.7097,
+      "step": 7028
+    },
+    {
+      "epoch": 64.49,
+      "learning_rate": 0.0001118653846153846,
+      "loss": 0.7329,
+      "step": 7029
+    },
+    {
+      "epoch": 64.5,
+      "learning_rate": 0.00011183653846153845,
+      "loss": 0.7273,
+      "step": 7030
+    },
+    {
+      "epoch": 64.5,
+      "learning_rate": 0.0001118076923076923,
+      "loss": 0.7325,
+      "step": 7031
+    },
+    {
+      "epoch": 64.51,
+      "learning_rate": 0.00011177884615384614,
+      "loss": 0.7084,
+      "step": 7032
+    },
+    {
+      "epoch": 64.52,
+      "learning_rate": 0.00011174999999999999,
+      "loss": 0.6282,
+      "step": 7033
+    },
+    {
+      "epoch": 64.53,
+      "learning_rate": 0.00011172115384615383,
+      "loss": 0.7405,
+      "step": 7034
+    },
+    {
+      "epoch": 64.54,
+      "learning_rate": 0.00011169230769230768,
+      "loss": 0.6462,
+      "step": 7035
+    },
+    {
+      "epoch": 64.55,
+      "learning_rate": 0.00011166346153846153,
+      "loss": 0.593,
+      "step": 7036
+    },
+    {
+      "epoch": 64.56,
+      "learning_rate": 0.00011163461538461537,
+      "loss": 0.4742,
+      "step": 7037
+    },
+    {
+      "epoch": 64.57,
+      "learning_rate": 0.00011160576923076922,
+      "loss": 0.6954,
+      "step": 7038
+    },
+    {
+      "epoch": 64.58,
+      "learning_rate": 0.00011157692307692306,
+      "loss": 0.5971,
+      "step": 7039
+    },
+    {
+      "epoch": 64.59,
+      "learning_rate": 0.0001115480769230769,
+      "loss": 0.7649,
+      "step": 7040
+    },
+    {
+      "epoch": 64.6,
+      "learning_rate": 0.00011151923076923077,
+      "loss": 0.7373,
+      "step": 7041
+    },
+    {
+      "epoch": 64.61,
+      "learning_rate": 0.0001114903846153846,
+      "loss": 0.6257,
+      "step": 7042
+    },
+    {
+      "epoch": 64.61,
+      "learning_rate": 0.00011146153846153846,
+      "loss": 0.587,
+      "step": 7043
+    },
+    {
+      "epoch": 64.62,
+      "learning_rate": 0.0001114326923076923,
+      "loss": 0.6716,
+      "step": 7044
+    },
+    {
+      "epoch": 64.63,
+      "learning_rate": 0.00011140384615384613,
+      "loss": 0.6431,
+      "step": 7045
+    },
+    {
+      "epoch": 64.64,
+      "learning_rate": 0.000111375,
+      "loss": 0.6412,
+      "step": 7046
+    },
+    {
+      "epoch": 64.65,
+      "learning_rate": 0.00011134615384615383,
+      "loss": 0.6188,
+      "step": 7047
+    },
+    {
+      "epoch": 64.66,
+      "learning_rate": 0.00011131730769230769,
+      "loss": 0.7282,
+      "step": 7048
+    },
+    {
+      "epoch": 64.67,
+      "learning_rate": 0.00011128846153846152,
+      "loss": 0.6188,
+      "step": 7049
+    },
+    {
+      "epoch": 64.68,
+      "learning_rate": 0.00011125961538461536,
+      "loss": 0.6658,
+      "step": 7050
+    },
+    {
+      "epoch": 64.69,
+      "learning_rate": 0.00011123076923076923,
+      "loss": 0.6233,
+      "step": 7051
+    },
+    {
+      "epoch": 64.7,
+      "learning_rate": 0.00011120192307692307,
+      "loss": 0.7207,
+      "step": 7052
+    },
+    {
+      "epoch": 64.71,
+      "learning_rate": 0.00011117307692307692,
+      "loss": 0.6991,
+      "step": 7053
+    },
+    {
+      "epoch": 64.72,
+      "learning_rate": 0.00011114423076923076,
+      "loss": 0.7585,
+      "step": 7054
+    },
+    {
+      "epoch": 64.72,
+      "learning_rate": 0.0001111153846153846,
+      "loss": 0.7065,
+      "step": 7055
+    },
+    {
+      "epoch": 64.73,
+      "learning_rate": 0.00011108653846153845,
+      "loss": 0.7316,
+      "step": 7056
+    },
+    {
+      "epoch": 64.74,
+      "learning_rate": 0.0001110576923076923,
+      "loss": 0.6411,
+      "step": 7057
+    },
+    {
+      "epoch": 64.75,
+      "learning_rate": 0.00011102884615384615,
+      "loss": 0.6033,
+      "step": 7058
+    },
+    {
+      "epoch": 64.76,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 0.604,
+      "step": 7059
+    },
+    {
+      "epoch": 64.77,
+      "learning_rate": 0.00011097115384615383,
+      "loss": 0.6478,
+      "step": 7060
+    },
+    {
+      "epoch": 64.78,
+      "learning_rate": 0.00011094230769230768,
+      "loss": 0.7382,
+      "step": 7061
+    },
+    {
+      "epoch": 64.79,
+      "learning_rate": 0.00011091346153846153,
+      "loss": 0.6264,
+      "step": 7062
+    },
+    {
+      "epoch": 64.8,
+      "learning_rate": 0.00011088461538461538,
+      "loss": 0.7113,
+      "step": 7063
+    },
+    {
+      "epoch": 64.81,
+      "learning_rate": 0.00011085576923076922,
+      "loss": 0.6033,
+      "step": 7064
+    },
+    {
+      "epoch": 64.82,
+      "learning_rate": 0.00011082692307692306,
+      "loss": 0.5814,
+      "step": 7065
+    },
+    {
+      "epoch": 64.83,
+      "learning_rate": 0.00011079807692307691,
+      "loss": 0.6253,
+      "step": 7066
+    },
+    {
+      "epoch": 64.83,
+      "learning_rate": 0.00011076923076923076,
+      "loss": 0.631,
+      "step": 7067
+    },
+    {
+      "epoch": 64.84,
+      "learning_rate": 0.00011074038461538461,
+      "loss": 0.6703,
+      "step": 7068
+    },
+    {
+      "epoch": 64.85,
+      "learning_rate": 0.00011071153846153845,
+      "loss": 0.6842,
+      "step": 7069
+    },
+    {
+      "epoch": 64.86,
+      "learning_rate": 0.00011068269230769229,
+      "loss": 0.5999,
+      "step": 7070
+    },
+    {
+      "epoch": 64.87,
+      "learning_rate": 0.00011065384615384614,
+      "loss": 0.7066,
+      "step": 7071
+    },
+    {
+      "epoch": 64.88,
+      "learning_rate": 0.00011062499999999999,
+      "loss": 0.7059,
+      "step": 7072
+    },
+    {
+      "epoch": 64.89,
+      "learning_rate": 0.00011059615384615384,
+      "loss": 0.6437,
+      "step": 7073
+    },
+    {
+      "epoch": 64.9,
+      "learning_rate": 0.00011056730769230768,
+      "loss": 0.6567,
+      "step": 7074
+    },
+    {
+      "epoch": 64.91,
+      "learning_rate": 0.00011053846153846152,
+      "loss": 0.7263,
+      "step": 7075
+    },
+    {
+      "epoch": 64.92,
+      "learning_rate": 0.00011050961538461537,
+      "loss": 0.6885,
+      "step": 7076
+    },
+    {
+      "epoch": 64.93,
+      "learning_rate": 0.00011048076923076922,
+      "loss": 0.7178,
+      "step": 7077
+    },
+    {
+      "epoch": 64.94,
+      "learning_rate": 0.00011045192307692308,
+      "loss": 0.7017,
+      "step": 7078
+    },
+    {
+      "epoch": 64.94,
+      "learning_rate": 0.00011042307692307691,
+      "loss": 0.6094,
+      "step": 7079
+    },
+    {
+      "epoch": 64.95,
+      "learning_rate": 0.00011039423076923075,
+      "loss": 0.7126,
+      "step": 7080
+    },
+    {
+      "epoch": 64.96,
+      "learning_rate": 0.0001103653846153846,
+      "loss": 0.633,
+      "step": 7081
+    },
+    {
+      "epoch": 64.97,
+      "learning_rate": 0.00011033653846153845,
+      "loss": 0.5448,
+      "step": 7082
+    },
+    {
+      "epoch": 64.98,
+      "learning_rate": 0.0001103076923076923,
+      "loss": 0.6279,
+      "step": 7083
+    },
+    {
+      "epoch": 64.99,
+      "learning_rate": 0.00011027884615384614,
+      "loss": 0.7304,
+      "step": 7084
+    },
+    {
+      "epoch": 65.0,
+      "learning_rate": 0.00011024999999999998,
+      "loss": 0.7805,
+      "step": 7085
+    },
+    {
+      "epoch": 65.01,
+      "learning_rate": 0.00011022115384615383,
+      "loss": 0.6355,
+      "step": 7086
+    },
+    {
+      "epoch": 65.02,
+      "learning_rate": 0.00011019230769230769,
+      "loss": 0.6901,
+      "step": 7087
+    },
+    {
+      "epoch": 65.03,
+      "learning_rate": 0.00011016346153846154,
+      "loss": 0.5692,
+      "step": 7088
+    },
+    {
+      "epoch": 65.04,
+      "learning_rate": 0.00011013461538461538,
+      "loss": 0.7217,
+      "step": 7089
+    },
+    {
+      "epoch": 65.05,
+      "learning_rate": 0.00011010576923076921,
+      "loss": 0.6332,
+      "step": 7090
+    },
+    {
+      "epoch": 65.06,
+      "learning_rate": 0.00011007692307692307,
+      "loss": 0.5914,
+      "step": 7091
+    },
+    {
+      "epoch": 65.06,
+      "learning_rate": 0.0001100480769230769,
+      "loss": 0.6077,
+      "step": 7092
+    },
+    {
+      "epoch": 65.07,
+      "learning_rate": 0.00011001923076923077,
+      "loss": 0.6151,
+      "step": 7093
+    },
+    {
+      "epoch": 65.08,
+      "learning_rate": 0.00010999038461538461,
+      "loss": 0.6955,
+      "step": 7094
+    },
+    {
+      "epoch": 65.09,
+      "learning_rate": 0.00010996153846153845,
+      "loss": 0.7084,
+      "step": 7095
+    },
+    {
+      "epoch": 65.1,
+      "learning_rate": 0.0001099326923076923,
+      "loss": 0.7122,
+      "step": 7096
+    },
+    {
+      "epoch": 65.11,
+      "learning_rate": 0.00010990384615384614,
+      "loss": 0.6969,
+      "step": 7097
+    },
+    {
+      "epoch": 65.12,
+      "learning_rate": 0.000109875,
+      "loss": 0.5632,
+      "step": 7098
+    },
+    {
+      "epoch": 65.13,
+      "learning_rate": 0.00010984615384615384,
+      "loss": 0.6372,
+      "step": 7099
+    },
+    {
+      "epoch": 65.14,
+      "learning_rate": 0.00010981730769230768,
+      "loss": 0.644,
+      "step": 7100
+    },
+    {
+      "epoch": 65.15,
+      "learning_rate": 0.00010978846153846153,
+      "loss": 0.6423,
+      "step": 7101
+    },
+    {
+      "epoch": 65.16,
+      "learning_rate": 0.00010975961538461537,
+      "loss": 0.7441,
+      "step": 7102
+    },
+    {
+      "epoch": 65.17,
+      "learning_rate": 0.00010973076923076923,
+      "loss": 0.7109,
+      "step": 7103
+    },
+    {
+      "epoch": 65.17,
+      "learning_rate": 0.00010970192307692307,
+      "loss": 0.7404,
+      "step": 7104
+    },
+    {
+      "epoch": 65.18,
+      "learning_rate": 0.00010967307692307691,
+      "loss": 0.6619,
+      "step": 7105
+    },
+    {
+      "epoch": 65.19,
+      "learning_rate": 0.00010964423076923076,
+      "loss": 0.5989,
+      "step": 7106
+    },
+    {
+      "epoch": 65.2,
+      "learning_rate": 0.0001096153846153846,
+      "loss": 0.6941,
+      "step": 7107
+    },
+    {
+      "epoch": 65.21,
+      "learning_rate": 0.00010958653846153846,
+      "loss": 0.6646,
+      "step": 7108
+    },
+    {
+      "epoch": 65.22,
+      "learning_rate": 0.0001095576923076923,
+      "loss": 0.6496,
+      "step": 7109
+    },
+    {
+      "epoch": 65.23,
+      "learning_rate": 0.00010952884615384614,
+      "loss": 0.554,
+      "step": 7110
+    },
+    {
+      "epoch": 65.24,
+      "learning_rate": 0.00010949999999999999,
+      "loss": 0.7399,
+      "step": 7111
+    },
+    {
+      "epoch": 65.25,
+      "learning_rate": 0.00010947115384615383,
+      "loss": 0.7028,
+      "step": 7112
+    },
+    {
+      "epoch": 65.26,
+      "learning_rate": 0.0001094423076923077,
+      "loss": 0.6336,
+      "step": 7113
+    },
+    {
+      "epoch": 65.27,
+      "learning_rate": 0.00010941346153846153,
+      "loss": 0.6407,
+      "step": 7114
+    },
+    {
+      "epoch": 65.28,
+      "learning_rate": 0.00010938461538461537,
+      "loss": 0.611,
+      "step": 7115
+    },
+    {
+      "epoch": 65.28,
+      "learning_rate": 0.00010935576923076922,
+      "loss": 0.5053,
+      "step": 7116
+    },
+    {
+      "epoch": 65.29,
+      "learning_rate": 0.00010932692307692306,
+      "loss": 0.7596,
+      "step": 7117
+    },
+    {
+      "epoch": 65.3,
+      "learning_rate": 0.00010929807692307693,
+      "loss": 0.6229,
+      "step": 7118
+    },
+    {
+      "epoch": 65.31,
+      "learning_rate": 0.00010926923076923076,
+      "loss": 0.639,
+      "step": 7119
+    },
+    {
+      "epoch": 65.32,
+      "learning_rate": 0.0001092403846153846,
+      "loss": 0.6657,
+      "step": 7120
+    },
+    {
+      "epoch": 65.33,
+      "learning_rate": 0.00010921153846153845,
+      "loss": 0.6509,
+      "step": 7121
+    },
+    {
+      "epoch": 65.34,
+      "learning_rate": 0.00010918269230769229,
+      "loss": 0.7671,
+      "step": 7122
+    },
+    {
+      "epoch": 65.35,
+      "learning_rate": 0.00010915384615384613,
+      "loss": 0.5531,
+      "step": 7123
+    },
+    {
+      "epoch": 65.36,
+      "learning_rate": 0.000109125,
+      "loss": 0.6067,
+      "step": 7124
+    },
+    {
+      "epoch": 65.37,
+      "learning_rate": 0.00010909615384615383,
+      "loss": 0.7328,
+      "step": 7125
+    },
+    {
+      "epoch": 65.38,
+      "learning_rate": 0.00010906730769230769,
+      "loss": 0.6603,
+      "step": 7126
+    },
+    {
+      "epoch": 65.39,
+      "learning_rate": 0.00010903846153846152,
+      "loss": 0.7739,
+      "step": 7127
+    },
+    {
+      "epoch": 65.39,
+      "learning_rate": 0.00010900961538461536,
+      "loss": 0.6833,
+      "step": 7128
+    },
+    {
+      "epoch": 65.4,
+      "learning_rate": 0.00010898076923076923,
+      "loss": 0.6973,
+      "step": 7129
+    },
+    {
+      "epoch": 65.41,
+      "learning_rate": 0.00010895192307692307,
+      "loss": 0.5409,
+      "step": 7130
+    },
+    {
+      "epoch": 65.42,
+      "learning_rate": 0.00010892307692307692,
+      "loss": 0.6776,
+      "step": 7131
+    },
+    {
+      "epoch": 65.43,
+      "learning_rate": 0.00010889423076923076,
+      "loss": 0.5767,
+      "step": 7132
+    },
+    {
+      "epoch": 65.44,
+      "learning_rate": 0.0001088653846153846,
+      "loss": 0.652,
+      "step": 7133
+    },
+    {
+      "epoch": 65.45,
+      "learning_rate": 0.00010883653846153846,
+      "loss": 0.682,
+      "step": 7134
+    },
+    {
+      "epoch": 65.46,
+      "learning_rate": 0.0001088076923076923,
+      "loss": 0.7373,
+      "step": 7135
+    },
+    {
+      "epoch": 65.47,
+      "learning_rate": 0.00010877884615384615,
+      "loss": 0.6708,
+      "step": 7136
+    },
+    {
+      "epoch": 65.48,
+      "learning_rate": 0.00010874999999999999,
+      "loss": 0.6869,
+      "step": 7137
+    },
+    {
+      "epoch": 65.49,
+      "learning_rate": 0.00010872115384615383,
+      "loss": 0.6526,
+      "step": 7138
+    },
+    {
+      "epoch": 65.5,
+      "learning_rate": 0.00010869230769230769,
+      "loss": 0.5452,
+      "step": 7139
+    },
+    {
+      "epoch": 65.5,
+      "learning_rate": 0.00010866346153846153,
+      "loss": 0.6146,
+      "step": 7140
+    },
+    {
+      "epoch": 65.51,
+      "learning_rate": 0.00010863461538461538,
+      "loss": 0.7538,
+      "step": 7141
+    },
+    {
+      "epoch": 65.52,
+      "learning_rate": 0.00010860576923076922,
+      "loss": 0.5004,
+      "step": 7142
+    },
+    {
+      "epoch": 65.53,
+      "learning_rate": 0.00010857692307692306,
+      "loss": 0.6381,
+      "step": 7143
+    },
+    {
+      "epoch": 65.54,
+      "learning_rate": 0.00010854807692307692,
+      "loss": 0.655,
+      "step": 7144
+    },
+    {
+      "epoch": 65.55,
+      "learning_rate": 0.00010851923076923076,
+      "loss": 0.6885,
+      "step": 7145
+    },
+    {
+      "epoch": 65.56,
+      "learning_rate": 0.00010849038461538461,
+      "loss": 0.602,
+      "step": 7146
+    },
+    {
+      "epoch": 65.57,
+      "learning_rate": 0.00010846153846153845,
+      "loss": 0.5563,
+      "step": 7147
+    },
+    {
+      "epoch": 65.58,
+      "learning_rate": 0.00010843269230769229,
+      "loss": 0.6627,
+      "step": 7148
+    },
+    {
+      "epoch": 65.59,
+      "learning_rate": 0.00010840384615384615,
+      "loss": 0.5624,
+      "step": 7149
+    },
+    {
+      "epoch": 65.6,
+      "learning_rate": 0.00010837499999999999,
+      "loss": 0.6282,
+      "step": 7150
+    },
+    {
+      "epoch": 65.61,
+      "learning_rate": 0.00010834615384615384,
+      "loss": 0.6314,
+      "step": 7151
+    },
+    {
+      "epoch": 65.61,
+      "learning_rate": 0.00010831730769230768,
+      "loss": 0.7569,
+      "step": 7152
+    },
+    {
+      "epoch": 65.62,
+      "learning_rate": 0.00010828846153846152,
+      "loss": 0.5688,
+      "step": 7153
+    },
+    {
+      "epoch": 65.63,
+      "learning_rate": 0.00010825961538461537,
+      "loss": 0.5663,
+      "step": 7154
+    },
+    {
+      "epoch": 65.64,
+      "learning_rate": 0.00010823076923076922,
+      "loss": 0.6143,
+      "step": 7155
+    },
+    {
+      "epoch": 65.65,
+      "learning_rate": 0.00010820192307692307,
+      "loss": 0.6456,
+      "step": 7156
+    },
+    {
+      "epoch": 65.66,
+      "learning_rate": 0.00010817307692307691,
+      "loss": 0.5997,
+      "step": 7157
+    },
+    {
+      "epoch": 65.67,
+      "learning_rate": 0.00010814423076923075,
+      "loss": 0.6479,
+      "step": 7158
+    },
+    {
+      "epoch": 65.68,
+      "learning_rate": 0.0001081153846153846,
+      "loss": 0.6803,
+      "step": 7159
+    },
+    {
+      "epoch": 65.69,
+      "learning_rate": 0.00010808653846153845,
+      "loss": 0.5419,
+      "step": 7160
+    },
+    {
+      "epoch": 65.7,
+      "learning_rate": 0.0001080576923076923,
+      "loss": 0.6584,
+      "step": 7161
+    },
+    {
+      "epoch": 65.71,
+      "learning_rate": 0.00010802884615384614,
+      "loss": 0.6182,
+      "step": 7162
+    },
+    {
+      "epoch": 65.72,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 0.6,
+      "step": 7163
+    },
+    {
+      "epoch": 65.72,
+      "learning_rate": 0.00010797115384615383,
+      "loss": 0.6775,
+      "step": 7164
+    },
+    {
+      "epoch": 65.73,
+      "learning_rate": 0.00010794230769230769,
+      "loss": 0.7766,
+      "step": 7165
+    },
+    {
+      "epoch": 65.74,
+      "learning_rate": 0.00010791346153846154,
+      "loss": 0.5887,
+      "step": 7166
+    },
+    {
+      "epoch": 65.75,
+      "learning_rate": 0.00010788461538461538,
+      "loss": 0.6658,
+      "step": 7167
+    },
+    {
+      "epoch": 65.76,
+      "learning_rate": 0.00010785576923076921,
+      "loss": 0.6755,
+      "step": 7168
+    },
+    {
+      "epoch": 65.77,
+      "learning_rate": 0.00010782692307692307,
+      "loss": 0.6365,
+      "step": 7169
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 0.00010779807692307692,
+      "loss": 0.659,
+      "step": 7170
+    },
+    {
+      "epoch": 65.79,
+      "learning_rate": 0.00010776923076923077,
+      "loss": 0.811,
+      "step": 7171
+    },
+    {
+      "epoch": 65.8,
+      "learning_rate": 0.00010774038461538461,
+      "loss": 0.7588,
+      "step": 7172
+    },
+    {
+      "epoch": 65.81,
+      "learning_rate": 0.00010771153846153845,
+      "loss": 0.6674,
+      "step": 7173
+    },
+    {
+      "epoch": 65.82,
+      "learning_rate": 0.0001076826923076923,
+      "loss": 0.5815,
+      "step": 7174
+    },
+    {
+      "epoch": 65.83,
+      "learning_rate": 0.00010765384615384615,
+      "loss": 0.7868,
+      "step": 7175
+    },
+    {
+      "epoch": 65.83,
+      "learning_rate": 0.000107625,
+      "loss": 0.5662,
+      "step": 7176
+    },
+    {
+      "epoch": 65.84,
+      "learning_rate": 0.00010759615384615384,
+      "loss": 0.6692,
+      "step": 7177
+    },
+    {
+      "epoch": 65.85,
+      "learning_rate": 0.00010756730769230768,
+      "loss": 0.7763,
+      "step": 7178
+    },
+    {
+      "epoch": 65.86,
+      "learning_rate": 0.00010753846153846153,
+      "loss": 0.525,
+      "step": 7179
+    },
+    {
+      "epoch": 65.87,
+      "learning_rate": 0.00010750961538461538,
+      "loss": 0.5742,
+      "step": 7180
+    },
+    {
+      "epoch": 65.88,
+      "learning_rate": 0.00010748076923076923,
+      "loss": 0.6854,
+      "step": 7181
+    },
+    {
+      "epoch": 65.89,
+      "learning_rate": 0.00010745192307692307,
+      "loss": 0.7206,
+      "step": 7182
+    },
+    {
+      "epoch": 65.9,
+      "learning_rate": 0.00010742307692307691,
+      "loss": 0.5283,
+      "step": 7183
+    },
+    {
+      "epoch": 65.91,
+      "learning_rate": 0.00010739423076923076,
+      "loss": 0.587,
+      "step": 7184
+    },
+    {
+      "epoch": 65.92,
+      "learning_rate": 0.00010736538461538461,
+      "loss": 0.6347,
+      "step": 7185
+    },
+    {
+      "epoch": 65.93,
+      "learning_rate": 0.00010733653846153846,
+      "loss": 0.613,
+      "step": 7186
+    },
+    {
+      "epoch": 65.94,
+      "learning_rate": 0.0001073076923076923,
+      "loss": 0.7206,
+      "step": 7187
+    },
+    {
+      "epoch": 65.94,
+      "learning_rate": 0.00010727884615384614,
+      "loss": 0.6193,
+      "step": 7188
+    },
+    {
+      "epoch": 65.95,
+      "learning_rate": 0.00010724999999999999,
+      "loss": 0.6443,
+      "step": 7189
+    },
+    {
+      "epoch": 65.96,
+      "learning_rate": 0.00010722115384615383,
+      "loss": 0.7052,
+      "step": 7190
+    },
+    {
+      "epoch": 65.97,
+      "learning_rate": 0.0001071923076923077,
+      "loss": 0.5531,
+      "step": 7191
+    },
+    {
+      "epoch": 65.98,
+      "learning_rate": 0.00010716346153846153,
+      "loss": 0.6403,
+      "step": 7192
+    },
+    {
+      "epoch": 65.99,
+      "learning_rate": 0.00010713461538461537,
+      "loss": 0.616,
+      "step": 7193
+    },
+    {
+      "epoch": 66.0,
+      "learning_rate": 0.00010710576923076922,
+      "loss": 0.6374,
+      "step": 7194
+    },
+    {
+      "epoch": 66.01,
+      "learning_rate": 0.00010707692307692306,
+      "loss": 0.6972,
+      "step": 7195
+    },
+    {
+      "epoch": 66.02,
+      "learning_rate": 0.00010704807692307693,
+      "loss": 0.6783,
+      "step": 7196
+    },
+    {
+      "epoch": 66.03,
+      "learning_rate": 0.00010701923076923076,
+      "loss": 0.6528,
+      "step": 7197
+    },
+    {
+      "epoch": 66.04,
+      "learning_rate": 0.0001069903846153846,
+      "loss": 0.7049,
+      "step": 7198
+    },
+    {
+      "epoch": 66.05,
+      "learning_rate": 0.00010696153846153845,
+      "loss": 0.5949,
+      "step": 7199
+    },
+    {
+      "epoch": 66.06,
+      "learning_rate": 0.00010693269230769229,
+      "loss": 0.5901,
+      "step": 7200
+    },
+    {
+      "epoch": 66.06,
+      "learning_rate": 0.00010690384615384616,
+      "loss": 0.6482,
+      "step": 7201
+    },
+    {
+      "epoch": 66.07,
+      "learning_rate": 0.000106875,
+      "loss": 0.7775,
+      "step": 7202
+    },
+    {
+      "epoch": 66.08,
+      "learning_rate": 0.00010684615384615383,
+      "loss": 0.827,
+      "step": 7203
+    },
+    {
+      "epoch": 66.09,
+      "learning_rate": 0.00010681730769230769,
+      "loss": 0.7117,
+      "step": 7204
+    },
+    {
+      "epoch": 66.1,
+      "learning_rate": 0.00010678846153846152,
+      "loss": 0.7202,
+      "step": 7205
+    },
+    {
+      "epoch": 66.11,
+      "learning_rate": 0.00010675961538461539,
+      "loss": 0.6951,
+      "step": 7206
+    },
+    {
+      "epoch": 66.12,
+      "learning_rate": 0.00010673076923076923,
+      "loss": 0.563,
+      "step": 7207
+    },
+    {
+      "epoch": 66.13,
+      "learning_rate": 0.00010670192307692307,
+      "loss": 0.5889,
+      "step": 7208
+    },
+    {
+      "epoch": 66.14,
+      "learning_rate": 0.00010667307692307692,
+      "loss": 0.6331,
+      "step": 7209
+    },
+    {
+      "epoch": 66.15,
+      "learning_rate": 0.00010664423076923076,
+      "loss": 0.6179,
+      "step": 7210
+    },
+    {
+      "epoch": 66.16,
+      "learning_rate": 0.00010661538461538461,
+      "loss": 0.6931,
+      "step": 7211
+    },
+    {
+      "epoch": 66.17,
+      "learning_rate": 0.00010658653846153846,
+      "loss": 0.7091,
+      "step": 7212
+    },
+    {
+      "epoch": 66.17,
+      "learning_rate": 0.0001065576923076923,
+      "loss": 0.6593,
+      "step": 7213
+    },
+    {
+      "epoch": 66.18,
+      "learning_rate": 0.00010652884615384615,
+      "loss": 0.6674,
+      "step": 7214
+    },
+    {
+      "epoch": 66.19,
+      "learning_rate": 0.00010649999999999999,
+      "loss": 0.6746,
+      "step": 7215
+    },
+    {
+      "epoch": 66.2,
+      "learning_rate": 0.00010647115384615384,
+      "loss": 0.7079,
+      "step": 7216
+    },
+    {
+      "epoch": 66.21,
+      "learning_rate": 0.00010644230769230769,
+      "loss": 0.7182,
+      "step": 7217
+    },
+    {
+      "epoch": 66.22,
+      "learning_rate": 0.00010641346153846153,
+      "loss": 0.5636,
+      "step": 7218
+    },
+    {
+      "epoch": 66.23,
+      "learning_rate": 0.00010638461538461538,
+      "loss": 0.5755,
+      "step": 7219
+    },
+    {
+      "epoch": 66.24,
+      "learning_rate": 0.00010635576923076922,
+      "loss": 0.7933,
+      "step": 7220
+    },
+    {
+      "epoch": 66.25,
+      "learning_rate": 0.00010632692307692306,
+      "loss": 0.6865,
+      "step": 7221
+    },
+    {
+      "epoch": 66.26,
+      "learning_rate": 0.00010629807692307692,
+      "loss": 0.5679,
+      "step": 7222
+    },
+    {
+      "epoch": 66.27,
+      "learning_rate": 0.00010626923076923076,
+      "loss": 0.6844,
+      "step": 7223
+    },
+    {
+      "epoch": 66.28,
+      "learning_rate": 0.00010624038461538461,
+      "loss": 0.5895,
+      "step": 7224
+    },
+    {
+      "epoch": 66.28,
+      "learning_rate": 0.00010621153846153845,
+      "loss": 0.7517,
+      "step": 7225
+    },
+    {
+      "epoch": 66.29,
+      "learning_rate": 0.00010618269230769229,
+      "loss": 0.5476,
+      "step": 7226
+    },
+    {
+      "epoch": 66.3,
+      "learning_rate": 0.00010615384615384615,
+      "loss": 0.6494,
+      "step": 7227
+    },
+    {
+      "epoch": 66.31,
+      "learning_rate": 0.00010612499999999999,
+      "loss": 0.6319,
+      "step": 7228
+    },
+    {
+      "epoch": 66.32,
+      "learning_rate": 0.00010609615384615384,
+      "loss": 0.6994,
+      "step": 7229
+    },
+    {
+      "epoch": 66.33,
+      "learning_rate": 0.00010606730769230768,
+      "loss": 0.659,
+      "step": 7230
+    },
+    {
+      "epoch": 66.34,
+      "learning_rate": 0.00010603846153846152,
+      "loss": 0.6559,
+      "step": 7231
+    },
+    {
+      "epoch": 66.35,
+      "learning_rate": 0.00010600961538461538,
+      "loss": 0.6714,
+      "step": 7232
+    },
+    {
+      "epoch": 66.36,
+      "learning_rate": 0.00010598076923076922,
+      "loss": 0.701,
+      "step": 7233
+    },
+    {
+      "epoch": 66.37,
+      "learning_rate": 0.00010595192307692307,
+      "loss": 0.5752,
+      "step": 7234
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 0.00010592307692307691,
+      "loss": 0.6683,
+      "step": 7235
+    },
+    {
+      "epoch": 66.39,
+      "learning_rate": 0.00010589423076923075,
+      "loss": 0.6433,
+      "step": 7236
+    },
+    {
+      "epoch": 66.39,
+      "learning_rate": 0.00010586538461538462,
+      "loss": 0.6427,
+      "step": 7237
+    },
+    {
+      "epoch": 66.4,
+      "learning_rate": 0.00010583653846153845,
+      "loss": 0.6484,
+      "step": 7238
+    },
+    {
+      "epoch": 66.41,
+      "learning_rate": 0.0001058076923076923,
+      "loss": 0.6337,
+      "step": 7239
+    },
+    {
+      "epoch": 66.42,
+      "learning_rate": 0.00010577884615384614,
+      "loss": 0.7172,
+      "step": 7240
+    },
+    {
+      "epoch": 66.43,
+      "learning_rate": 0.00010574999999999998,
+      "loss": 0.5687,
+      "step": 7241
+    },
+    {
+      "epoch": 66.44,
+      "learning_rate": 0.00010572115384615385,
+      "loss": 0.7648,
+      "step": 7242
+    },
+    {
+      "epoch": 66.45,
+      "learning_rate": 0.00010569230769230769,
+      "loss": 0.7063,
+      "step": 7243
+    },
+    {
+      "epoch": 66.46,
+      "learning_rate": 0.00010566346153846152,
+      "loss": 0.6433,
+      "step": 7244
+    },
+    {
+      "epoch": 66.47,
+      "learning_rate": 0.00010563461538461538,
+      "loss": 0.6429,
+      "step": 7245
+    },
+    {
+      "epoch": 66.48,
+      "learning_rate": 0.00010560576923076921,
+      "loss": 0.626,
+      "step": 7246
+    },
+    {
+      "epoch": 66.49,
+      "learning_rate": 0.00010557692307692308,
+      "loss": 0.6926,
+      "step": 7247
+    },
+    {
+      "epoch": 66.5,
+      "learning_rate": 0.00010554807692307692,
+      "loss": 0.6524,
+      "step": 7248
+    },
+    {
+      "epoch": 66.5,
+      "learning_rate": 0.00010551923076923075,
+      "loss": 0.7357,
+      "step": 7249
+    },
+    {
+      "epoch": 66.51,
+      "learning_rate": 0.0001054903846153846,
+      "loss": 0.6364,
+      "step": 7250
+    },
+    {
+      "epoch": 66.52,
+      "learning_rate": 0.00010546153846153844,
+      "loss": 0.6271,
+      "step": 7251
+    },
+    {
+      "epoch": 66.53,
+      "learning_rate": 0.0001054326923076923,
+      "loss": 0.8,
+      "step": 7252
+    },
+    {
+      "epoch": 66.54,
+      "learning_rate": 0.00010540384615384615,
+      "loss": 0.5559,
+      "step": 7253
+    },
+    {
+      "epoch": 66.55,
+      "learning_rate": 0.00010537499999999999,
+      "loss": 0.7186,
+      "step": 7254
+    },
+    {
+      "epoch": 66.56,
+      "learning_rate": 0.00010534615384615384,
+      "loss": 0.667,
+      "step": 7255
+    },
+    {
+      "epoch": 66.57,
+      "learning_rate": 0.00010531730769230768,
+      "loss": 0.7542,
+      "step": 7256
+    },
+    {
+      "epoch": 66.58,
+      "learning_rate": 0.00010528846153846153,
+      "loss": 0.5835,
+      "step": 7257
+    },
+    {
+      "epoch": 66.59,
+      "learning_rate": 0.00010525961538461538,
+      "loss": 0.8017,
+      "step": 7258
+    },
+    {
+      "epoch": 66.6,
+      "learning_rate": 0.00010523076923076922,
+      "loss": 0.6755,
+      "step": 7259
+    },
+    {
+      "epoch": 66.61,
+      "learning_rate": 0.00010520192307692307,
+      "loss": 0.6724,
+      "step": 7260
+    },
+    {
+      "epoch": 66.61,
+      "learning_rate": 0.00010517307692307691,
+      "loss": 0.646,
+      "step": 7261
+    },
+    {
+      "epoch": 66.62,
+      "learning_rate": 0.00010514423076923076,
+      "loss": 0.6417,
+      "step": 7262
+    },
+    {
+      "epoch": 66.63,
+      "learning_rate": 0.00010511538461538461,
+      "loss": 0.7221,
+      "step": 7263
+    },
+    {
+      "epoch": 66.64,
+      "learning_rate": 0.00010508653846153845,
+      "loss": 0.6705,
+      "step": 7264
+    },
+    {
+      "epoch": 66.65,
+      "learning_rate": 0.0001050576923076923,
+      "loss": 0.6016,
+      "step": 7265
+    },
+    {
+      "epoch": 66.66,
+      "learning_rate": 0.00010502884615384614,
+      "loss": 0.7025,
+      "step": 7266
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 0.5987,
+      "step": 7267
+    },
+    {
+      "epoch": 66.68,
+      "learning_rate": 0.00010497115384615384,
+      "loss": 0.7714,
+      "step": 7268
+    },
+    {
+      "epoch": 66.69,
+      "learning_rate": 0.00010494230769230768,
+      "loss": 0.5329,
+      "step": 7269
+    },
+    {
+      "epoch": 66.7,
+      "learning_rate": 0.00010491346153846153,
+      "loss": 0.7008,
+      "step": 7270
+    },
+    {
+      "epoch": 66.71,
+      "learning_rate": 0.00010488461538461537,
+      "loss": 0.702,
+      "step": 7271
+    },
+    {
+      "epoch": 66.72,
+      "learning_rate": 0.00010485576923076921,
+      "loss": 0.7305,
+      "step": 7272
+    },
+    {
+      "epoch": 66.72,
+      "learning_rate": 0.00010482692307692307,
+      "loss": 0.6549,
+      "step": 7273
+    },
+    {
+      "epoch": 66.73,
+      "learning_rate": 0.00010479807692307691,
+      "loss": 0.6732,
+      "step": 7274
+    },
+    {
+      "epoch": 66.74,
+      "learning_rate": 0.00010476923076923076,
+      "loss": 0.6557,
+      "step": 7275
+    },
+    {
+      "epoch": 66.75,
+      "learning_rate": 0.0001047403846153846,
+      "loss": 0.7696,
+      "step": 7276
+    },
+    {
+      "epoch": 66.76,
+      "learning_rate": 0.00010471153846153844,
+      "loss": 0.6716,
+      "step": 7277
+    },
+    {
+      "epoch": 66.77,
+      "learning_rate": 0.0001046826923076923,
+      "loss": 0.653,
+      "step": 7278
+    },
+    {
+      "epoch": 66.78,
+      "learning_rate": 0.00010465384615384614,
+      "loss": 0.6195,
+      "step": 7279
+    },
+    {
+      "epoch": 66.79,
+      "learning_rate": 0.000104625,
+      "loss": 0.689,
+      "step": 7280
+    },
+    {
+      "epoch": 66.8,
+      "learning_rate": 0.00010459615384615383,
+      "loss": 0.6049,
+      "step": 7281
+    },
+    {
+      "epoch": 66.81,
+      "learning_rate": 0.00010456730769230767,
+      "loss": 0.5863,
+      "step": 7282
+    },
+    {
+      "epoch": 66.82,
+      "learning_rate": 0.00010453846153846154,
+      "loss": 0.8819,
+      "step": 7283
+    },
+    {
+      "epoch": 66.83,
+      "learning_rate": 0.00010450961538461537,
+      "loss": 0.6683,
+      "step": 7284
+    },
+    {
+      "epoch": 66.83,
+      "learning_rate": 0.00010448076923076923,
+      "loss": 0.6926,
+      "step": 7285
+    },
+    {
+      "epoch": 66.84,
+      "learning_rate": 0.00010445192307692306,
+      "loss": 0.6095,
+      "step": 7286
+    },
+    {
+      "epoch": 66.85,
+      "learning_rate": 0.0001044230769230769,
+      "loss": 0.6939,
+      "step": 7287
+    },
+    {
+      "epoch": 66.86,
+      "learning_rate": 0.00010439423076923075,
+      "loss": 0.612,
+      "step": 7288
+    },
+    {
+      "epoch": 66.87,
+      "learning_rate": 0.0001043653846153846,
+      "loss": 0.7546,
+      "step": 7289
+    },
+    {
+      "epoch": 66.88,
+      "learning_rate": 0.00010433653846153846,
+      "loss": 0.6222,
+      "step": 7290
+    },
+    {
+      "epoch": 66.89,
+      "learning_rate": 0.0001043076923076923,
+      "loss": 0.7062,
+      "step": 7291
+    },
+    {
+      "epoch": 66.9,
+      "learning_rate": 0.00010427884615384613,
+      "loss": 0.7147,
+      "step": 7292
+    },
+    {
+      "epoch": 66.91,
+      "learning_rate": 0.00010424999999999999,
+      "loss": 0.6888,
+      "step": 7293
+    },
+    {
+      "epoch": 66.92,
+      "learning_rate": 0.00010422115384615384,
+      "loss": 0.6111,
+      "step": 7294
+    },
+    {
+      "epoch": 66.93,
+      "learning_rate": 0.00010419230769230769,
+      "loss": 0.6173,
+      "step": 7295
+    },
+    {
+      "epoch": 66.94,
+      "learning_rate": 0.00010416346153846153,
+      "loss": 0.5304,
+      "step": 7296
+    },
+    {
+      "epoch": 66.94,
+      "learning_rate": 0.00010413461538461537,
+      "loss": 0.7187,
+      "step": 7297
+    },
+    {
+      "epoch": 66.95,
+      "learning_rate": 0.00010410576923076922,
+      "loss": 0.5295,
+      "step": 7298
+    },
+    {
+      "epoch": 66.96,
+      "learning_rate": 0.00010407692307692307,
+      "loss": 0.5659,
+      "step": 7299
+    },
+    {
+      "epoch": 66.97,
+      "learning_rate": 0.00010404807692307692,
+      "loss": 0.7045,
+      "step": 7300
+    },
+    {
+      "epoch": 66.98,
+      "learning_rate": 0.00010401923076923076,
+      "loss": 0.7484,
+      "step": 7301
+    },
+    {
+      "epoch": 66.99,
+      "learning_rate": 0.0001039903846153846,
+      "loss": 0.6169,
+      "step": 7302
+    },
+    {
+      "epoch": 67.0,
+      "learning_rate": 0.00010396153846153845,
+      "loss": 0.6749,
+      "step": 7303
+    },
+    {
+      "epoch": 67.01,
+      "learning_rate": 0.0001039326923076923,
+      "loss": 0.7286,
+      "step": 7304
+    },
+    {
+      "epoch": 67.02,
+      "learning_rate": 0.00010390384615384615,
+      "loss": 0.4816,
+      "step": 7305
+    },
+    {
+      "epoch": 67.03,
+      "learning_rate": 0.00010387499999999999,
+      "loss": 0.6752,
+      "step": 7306
+    },
+    {
+      "epoch": 67.04,
+      "learning_rate": 0.00010384615384615383,
+      "loss": 0.582,
+      "step": 7307
+    },
+    {
+      "epoch": 67.05,
+      "learning_rate": 0.00010381730769230768,
+      "loss": 0.637,
+      "step": 7308
+    },
+    {
+      "epoch": 67.06,
+      "learning_rate": 0.00010378846153846153,
+      "loss": 0.6222,
+      "step": 7309
+    },
+    {
+      "epoch": 67.06,
+      "learning_rate": 0.00010375961538461538,
+      "loss": 0.7421,
+      "step": 7310
+    },
+    {
+      "epoch": 67.07,
+      "learning_rate": 0.00010373076923076922,
+      "loss": 0.6314,
+      "step": 7311
+    },
+    {
+      "epoch": 67.08,
+      "learning_rate": 0.00010370192307692306,
+      "loss": 0.5986,
+      "step": 7312
+    },
+    {
+      "epoch": 67.09,
+      "learning_rate": 0.00010367307692307691,
+      "loss": 0.5838,
+      "step": 7313
+    },
+    {
+      "epoch": 67.1,
+      "learning_rate": 0.00010364423076923076,
+      "loss": 0.7139,
+      "step": 7314
+    },
+    {
+      "epoch": 67.11,
+      "learning_rate": 0.00010361538461538462,
+      "loss": 0.6326,
+      "step": 7315
+    },
+    {
+      "epoch": 67.12,
+      "learning_rate": 0.00010358653846153845,
+      "loss": 0.6411,
+      "step": 7316
+    },
+    {
+      "epoch": 67.13,
+      "learning_rate": 0.00010355769230769229,
+      "loss": 0.6095,
+      "step": 7317
+    },
+    {
+      "epoch": 67.14,
+      "learning_rate": 0.00010352884615384614,
+      "loss": 0.6336,
+      "step": 7318
+    },
+    {
+      "epoch": 67.15,
+      "learning_rate": 0.00010349999999999998,
+      "loss": 0.5781,
+      "step": 7319
+    },
+    {
+      "epoch": 67.16,
+      "learning_rate": 0.00010347115384615385,
+      "loss": 0.5656,
+      "step": 7320
+    },
+    {
+      "epoch": 67.17,
+      "learning_rate": 0.00010344230769230768,
+      "loss": 0.6729,
+      "step": 7321
+    },
+    {
+      "epoch": 67.17,
+      "learning_rate": 0.00010341346153846152,
+      "loss": 0.6335,
+      "step": 7322
+    },
+    {
+      "epoch": 67.18,
+      "learning_rate": 0.00010338461538461537,
+      "loss": 0.6901,
+      "step": 7323
+    },
+    {
+      "epoch": 67.19,
+      "learning_rate": 0.00010335576923076921,
+      "loss": 0.6444,
+      "step": 7324
+    },
+    {
+      "epoch": 67.2,
+      "learning_rate": 0.00010332692307692308,
+      "loss": 0.7628,
+      "step": 7325
+    },
+    {
+      "epoch": 67.21,
+      "learning_rate": 0.00010329807692307692,
+      "loss": 0.6638,
+      "step": 7326
+    },
+    {
+      "epoch": 67.22,
+      "learning_rate": 0.00010326923076923075,
+      "loss": 0.6342,
+      "step": 7327
+    },
+    {
+      "epoch": 67.23,
+      "learning_rate": 0.0001032403846153846,
+      "loss": 0.5664,
+      "step": 7328
+    },
+    {
+      "epoch": 67.24,
+      "learning_rate": 0.00010321153846153844,
+      "loss": 0.5812,
+      "step": 7329
+    },
+    {
+      "epoch": 67.25,
+      "learning_rate": 0.00010318269230769231,
+      "loss": 0.6977,
+      "step": 7330
+    },
+    {
+      "epoch": 67.26,
+      "learning_rate": 0.00010315384615384615,
+      "loss": 0.616,
+      "step": 7331
+    },
+    {
+      "epoch": 67.27,
+      "learning_rate": 0.00010312499999999999,
+      "loss": 0.5691,
+      "step": 7332
+    },
+    {
+      "epoch": 67.28,
+      "learning_rate": 0.00010309615384615384,
+      "loss": 0.6549,
+      "step": 7333
+    },
+    {
+      "epoch": 67.28,
+      "learning_rate": 0.00010306730769230768,
+      "loss": 0.6901,
+      "step": 7334
+    },
+    {
+      "epoch": 67.29,
+      "learning_rate": 0.00010303846153846154,
+      "loss": 0.6316,
+      "step": 7335
+    },
+    {
+      "epoch": 67.3,
+      "learning_rate": 0.00010300961538461538,
+      "loss": 0.5483,
+      "step": 7336
+    },
+    {
+      "epoch": 67.31,
+      "learning_rate": 0.00010298076923076922,
+      "loss": 0.5973,
+      "step": 7337
+    },
+    {
+      "epoch": 67.32,
+      "learning_rate": 0.00010295192307692307,
+      "loss": 0.6124,
+      "step": 7338
+    },
+    {
+      "epoch": 67.33,
+      "learning_rate": 0.00010292307692307691,
+      "loss": 0.747,
+      "step": 7339
+    },
+    {
+      "epoch": 67.34,
+      "learning_rate": 0.00010289423076923077,
+      "loss": 0.6114,
+      "step": 7340
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 0.00010286538461538461,
+      "loss": 0.6841,
+      "step": 7341
+    },
+    {
+      "epoch": 67.36,
+      "learning_rate": 0.00010283653846153845,
+      "loss": 0.6639,
+      "step": 7342
+    },
+    {
+      "epoch": 67.37,
+      "learning_rate": 0.0001028076923076923,
+      "loss": 0.6519,
+      "step": 7343
+    },
+    {
+      "epoch": 67.38,
+      "learning_rate": 0.00010277884615384614,
+      "loss": 0.6088,
+      "step": 7344
+    },
+    {
+      "epoch": 67.39,
+      "learning_rate": 0.00010275,
+      "loss": 0.6934,
+      "step": 7345
+    },
+    {
+      "epoch": 67.39,
+      "learning_rate": 0.00010272115384615384,
+      "loss": 0.5294,
+      "step": 7346
+    },
+    {
+      "epoch": 67.4,
+      "learning_rate": 0.00010269230769230768,
+      "loss": 0.6256,
+      "step": 7347
+    },
+    {
+      "epoch": 67.41,
+      "learning_rate": 0.00010266346153846153,
+      "loss": 0.6314,
+      "step": 7348
+    },
+    {
+      "epoch": 67.42,
+      "learning_rate": 0.00010263461538461537,
+      "loss": 0.6846,
+      "step": 7349
+    },
+    {
+      "epoch": 67.43,
+      "learning_rate": 0.00010260576923076921,
+      "loss": 0.5119,
+      "step": 7350
+    },
+    {
+      "epoch": 67.44,
+      "learning_rate": 0.00010257692307692307,
+      "loss": 0.6549,
+      "step": 7351
+    },
+    {
+      "epoch": 67.45,
+      "learning_rate": 0.00010254807692307691,
+      "loss": 0.6998,
+      "step": 7352
+    },
+    {
+      "epoch": 67.46,
+      "learning_rate": 0.00010251923076923076,
+      "loss": 0.5698,
+      "step": 7353
+    },
+    {
+      "epoch": 67.47,
+      "learning_rate": 0.0001024903846153846,
+      "loss": 0.6298,
+      "step": 7354
+    },
+    {
+      "epoch": 67.48,
+      "learning_rate": 0.00010246153846153844,
+      "loss": 0.6624,
+      "step": 7355
+    },
+    {
+      "epoch": 67.49,
+      "learning_rate": 0.0001024326923076923,
+      "loss": 0.7514,
+      "step": 7356
+    },
+    {
+      "epoch": 67.5,
+      "learning_rate": 0.00010240384615384614,
+      "loss": 0.6118,
+      "step": 7357
+    },
+    {
+      "epoch": 67.5,
+      "learning_rate": 0.000102375,
+      "loss": 0.7116,
+      "step": 7358
+    },
+    {
+      "epoch": 67.51,
+      "learning_rate": 0.00010234615384615383,
+      "loss": 0.6372,
+      "step": 7359
+    },
+    {
+      "epoch": 67.52,
+      "learning_rate": 0.00010231730769230767,
+      "loss": 0.6533,
+      "step": 7360
+    },
+    {
+      "epoch": 67.53,
+      "learning_rate": 0.00010228846153846154,
+      "loss": 0.6948,
+      "step": 7361
+    },
+    {
+      "epoch": 67.54,
+      "learning_rate": 0.00010225961538461537,
+      "loss": 0.6112,
+      "step": 7362
+    },
+    {
+      "epoch": 67.55,
+      "learning_rate": 0.00010223076923076923,
+      "loss": 0.6761,
+      "step": 7363
+    },
+    {
+      "epoch": 67.56,
+      "learning_rate": 0.00010220192307692306,
+      "loss": 0.5838,
+      "step": 7364
+    },
+    {
+      "epoch": 67.57,
+      "learning_rate": 0.0001021730769230769,
+      "loss": 0.6477,
+      "step": 7365
+    },
+    {
+      "epoch": 67.58,
+      "learning_rate": 0.00010214423076923077,
+      "loss": 0.7231,
+      "step": 7366
+    },
+    {
+      "epoch": 67.59,
+      "learning_rate": 0.0001021153846153846,
+      "loss": 0.5619,
+      "step": 7367
+    },
+    {
+      "epoch": 67.6,
+      "learning_rate": 0.00010208653846153846,
+      "loss": 0.6792,
+      "step": 7368
+    },
+    {
+      "epoch": 67.61,
+      "learning_rate": 0.0001020576923076923,
+      "loss": 0.6597,
+      "step": 7369
+    },
+    {
+      "epoch": 67.61,
+      "learning_rate": 0.00010202884615384613,
+      "loss": 0.6905,
+      "step": 7370
+    },
+    {
+      "epoch": 67.62,
+      "learning_rate": 0.000102,
+      "loss": 0.5467,
+      "step": 7371
+    },
+    {
+      "epoch": 67.63,
+      "learning_rate": 0.00010197115384615384,
+      "loss": 0.4766,
+      "step": 7372
+    },
+    {
+      "epoch": 67.64,
+      "learning_rate": 0.00010194230769230769,
+      "loss": 0.762,
+      "step": 7373
+    },
+    {
+      "epoch": 67.65,
+      "learning_rate": 0.00010191346153846153,
+      "loss": 0.4779,
+      "step": 7374
+    },
+    {
+      "epoch": 67.66,
+      "learning_rate": 0.00010188461538461537,
+      "loss": 0.6539,
+      "step": 7375
+    },
+    {
+      "epoch": 67.67,
+      "learning_rate": 0.00010185576923076923,
+      "loss": 0.653,
+      "step": 7376
+    },
+    {
+      "epoch": 67.68,
+      "learning_rate": 0.00010182692307692307,
+      "loss": 0.7096,
+      "step": 7377
+    },
+    {
+      "epoch": 67.69,
+      "learning_rate": 0.00010179807692307692,
+      "loss": 0.548,
+      "step": 7378
+    },
+    {
+      "epoch": 67.7,
+      "learning_rate": 0.00010176923076923076,
+      "loss": 0.7392,
+      "step": 7379
+    },
+    {
+      "epoch": 67.71,
+      "learning_rate": 0.0001017403846153846,
+      "loss": 0.5345,
+      "step": 7380
+    },
+    {
+      "epoch": 67.72,
+      "learning_rate": 0.00010171153846153845,
+      "loss": 0.5753,
+      "step": 7381
+    },
+    {
+      "epoch": 67.72,
+      "learning_rate": 0.0001016826923076923,
+      "loss": 0.5715,
+      "step": 7382
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 0.00010165384615384615,
+      "loss": 0.7617,
+      "step": 7383
+    },
+    {
+      "epoch": 67.74,
+      "learning_rate": 0.00010162499999999999,
+      "loss": 0.6093,
+      "step": 7384
+    },
+    {
+      "epoch": 67.75,
+      "learning_rate": 0.00010159615384615383,
+      "loss": 0.582,
+      "step": 7385
+    },
+    {
+      "epoch": 67.76,
+      "learning_rate": 0.00010156730769230768,
+      "loss": 0.6739,
+      "step": 7386
+    },
+    {
+      "epoch": 67.77,
+      "learning_rate": 0.00010153846153846153,
+      "loss": 0.7083,
+      "step": 7387
+    },
+    {
+      "epoch": 67.78,
+      "learning_rate": 0.00010150961538461538,
+      "loss": 0.6576,
+      "step": 7388
+    },
+    {
+      "epoch": 67.79,
+      "learning_rate": 0.00010148076923076922,
+      "loss": 0.6119,
+      "step": 7389
+    },
+    {
+      "epoch": 67.8,
+      "learning_rate": 0.00010145192307692306,
+      "loss": 0.6865,
+      "step": 7390
+    },
+    {
+      "epoch": 67.81,
+      "learning_rate": 0.00010142307692307691,
+      "loss": 0.6464,
+      "step": 7391
+    },
+    {
+      "epoch": 67.82,
+      "learning_rate": 0.00010139423076923076,
+      "loss": 0.6587,
+      "step": 7392
+    },
+    {
+      "epoch": 67.83,
+      "learning_rate": 0.00010136538461538461,
+      "loss": 0.7307,
+      "step": 7393
+    },
+    {
+      "epoch": 67.83,
+      "learning_rate": 0.00010133653846153845,
+      "loss": 0.5697,
+      "step": 7394
+    },
+    {
+      "epoch": 67.84,
+      "learning_rate": 0.00010130769230769229,
+      "loss": 0.6716,
+      "step": 7395
+    },
+    {
+      "epoch": 67.85,
+      "learning_rate": 0.00010127884615384614,
+      "loss": 0.6896,
+      "step": 7396
+    },
+    {
+      "epoch": 67.86,
+      "learning_rate": 0.00010125,
+      "loss": 0.6454,
+      "step": 7397
+    },
+    {
+      "epoch": 67.87,
+      "learning_rate": 0.00010122115384615385,
+      "loss": 0.6668,
+      "step": 7398
+    },
+    {
+      "epoch": 67.88,
+      "learning_rate": 0.00010119230769230768,
+      "loss": 0.687,
+      "step": 7399
+    },
+    {
+      "epoch": 67.89,
+      "learning_rate": 0.00010116346153846152,
+      "loss": 0.6501,
+      "step": 7400
+    },
+    {
+      "epoch": 67.9,
+      "learning_rate": 0.00010113461538461537,
+      "loss": 0.5902,
+      "step": 7401
+    },
+    {
+      "epoch": 67.91,
+      "learning_rate": 0.00010110576923076923,
+      "loss": 0.643,
+      "step": 7402
+    },
+    {
+      "epoch": 67.92,
+      "learning_rate": 0.00010107692307692308,
+      "loss": 0.6032,
+      "step": 7403
+    },
+    {
+      "epoch": 67.93,
+      "learning_rate": 0.00010104807692307692,
+      "loss": 0.7405,
+      "step": 7404
+    },
+    {
+      "epoch": 67.94,
+      "learning_rate": 0.00010101923076923075,
+      "loss": 0.6708,
+      "step": 7405
+    },
+    {
+      "epoch": 67.94,
+      "learning_rate": 0.0001009903846153846,
+      "loss": 0.6984,
+      "step": 7406
+    },
+    {
+      "epoch": 67.95,
+      "learning_rate": 0.00010096153846153846,
+      "loss": 0.6125,
+      "step": 7407
+    },
+    {
+      "epoch": 67.96,
+      "learning_rate": 0.00010093269230769231,
+      "loss": 0.6436,
+      "step": 7408
+    },
+    {
+      "epoch": 67.97,
+      "learning_rate": 0.00010090384615384615,
+      "loss": 0.5501,
+      "step": 7409
+    },
+    {
+      "epoch": 67.98,
+      "learning_rate": 0.00010087499999999999,
+      "loss": 0.7173,
+      "step": 7410
+    },
+    {
+      "epoch": 67.99,
+      "learning_rate": 0.00010084615384615384,
+      "loss": 0.5794,
+      "step": 7411
+    },
+    {
+      "epoch": 68.0,
+      "learning_rate": 0.00010081730769230769,
+      "loss": 0.7277,
+      "step": 7412
+    },
+    {
+      "epoch": 68.01,
+      "learning_rate": 0.00010078846153846154,
+      "loss": 0.5918,
+      "step": 7413
+    },
+    {
+      "epoch": 68.02,
+      "learning_rate": 0.00010075961538461538,
+      "loss": 0.7393,
+      "step": 7414
+    },
+    {
+      "epoch": 68.03,
+      "learning_rate": 0.00010073076923076922,
+      "loss": 0.6568,
+      "step": 7415
+    },
+    {
+      "epoch": 68.04,
+      "learning_rate": 0.00010070192307692307,
+      "loss": 0.6465,
+      "step": 7416
+    },
+    {
+      "epoch": 68.05,
+      "learning_rate": 0.0001006730769230769,
+      "loss": 0.7367,
+      "step": 7417
+    },
+    {
+      "epoch": 68.06,
+      "learning_rate": 0.00010064423076923077,
+      "loss": 0.7324,
+      "step": 7418
+    },
+    {
+      "epoch": 68.06,
+      "learning_rate": 0.00010061538461538461,
+      "loss": 0.818,
+      "step": 7419
+    },
+    {
+      "epoch": 68.07,
+      "learning_rate": 0.00010058653846153845,
+      "loss": 0.6709,
+      "step": 7420
+    },
+    {
+      "epoch": 68.08,
+      "learning_rate": 0.0001005576923076923,
+      "loss": 0.5623,
+      "step": 7421
+    },
+    {
+      "epoch": 68.09,
+      "learning_rate": 0.00010052884615384614,
+      "loss": 0.7319,
+      "step": 7422
+    },
+    {
+      "epoch": 68.1,
+      "learning_rate": 0.0001005,
+      "loss": 0.5919,
+      "step": 7423
+    },
+    {
+      "epoch": 68.11,
+      "learning_rate": 0.00010047115384615384,
+      "loss": 0.6679,
+      "step": 7424
+    },
+    {
+      "epoch": 68.12,
+      "learning_rate": 0.00010044230769230768,
+      "loss": 0.6805,
+      "step": 7425
+    },
+    {
+      "epoch": 68.13,
+      "learning_rate": 0.00010041346153846153,
+      "loss": 0.6625,
+      "step": 7426
+    },
+    {
+      "epoch": 68.14,
+      "learning_rate": 0.00010038461538461537,
+      "loss": 0.4648,
+      "step": 7427
+    },
+    {
+      "epoch": 68.15,
+      "learning_rate": 0.00010035576923076923,
+      "loss": 0.5752,
+      "step": 7428
+    },
+    {
+      "epoch": 68.16,
+      "learning_rate": 0.00010032692307692307,
+      "loss": 0.6694,
+      "step": 7429
+    },
+    {
+      "epoch": 68.17,
+      "learning_rate": 0.00010029807692307691,
+      "loss": 0.542,
+      "step": 7430
+    },
+    {
+      "epoch": 68.17,
+      "learning_rate": 0.00010026923076923076,
+      "loss": 0.5915,
+      "step": 7431
+    },
+    {
+      "epoch": 68.18,
+      "learning_rate": 0.0001002403846153846,
+      "loss": 0.6199,
+      "step": 7432
+    },
+    {
+      "epoch": 68.19,
+      "learning_rate": 0.00010021153846153847,
+      "loss": 0.5483,
+      "step": 7433
+    },
+    {
+      "epoch": 68.2,
+      "learning_rate": 0.0001001826923076923,
+      "loss": 0.6997,
+      "step": 7434
+    },
+    {
+      "epoch": 68.21,
+      "learning_rate": 0.00010015384615384614,
+      "loss": 0.5957,
+      "step": 7435
+    },
+    {
+      "epoch": 68.22,
+      "learning_rate": 0.000100125,
+      "loss": 0.6955,
+      "step": 7436
+    },
+    {
+      "epoch": 68.23,
+      "learning_rate": 0.00010009615384615383,
+      "loss": 0.5879,
+      "step": 7437
+    },
+    {
+      "epoch": 68.24,
+      "learning_rate": 0.0001000673076923077,
+      "loss": 0.6721,
+      "step": 7438
+    },
+    {
+      "epoch": 68.25,
+      "learning_rate": 0.00010003846153846154,
+      "loss": 0.5809,
+      "step": 7439
+    },
+    {
+      "epoch": 68.26,
+      "learning_rate": 0.00010000961538461537,
+      "loss": 0.6918,
+      "step": 7440
+    },
+    {
+      "epoch": 68.27,
+      "learning_rate": 9.998076923076923e-05,
+      "loss": 0.6239,
+      "step": 7441
+    },
+    {
+      "epoch": 68.28,
+      "learning_rate": 9.995192307692306e-05,
+      "loss": 0.7155,
+      "step": 7442
+    },
+    {
+      "epoch": 68.28,
+      "learning_rate": 9.992307692307693e-05,
+      "loss": 0.7899,
+      "step": 7443
+    },
+    {
+      "epoch": 68.29,
+      "learning_rate": 9.989423076923077e-05,
+      "loss": 0.6795,
+      "step": 7444
+    },
+    {
+      "epoch": 68.3,
+      "learning_rate": 9.98653846153846e-05,
+      "loss": 0.6585,
+      "step": 7445
+    },
+    {
+      "epoch": 68.31,
+      "learning_rate": 9.983653846153846e-05,
+      "loss": 0.665,
+      "step": 7446
+    },
+    {
+      "epoch": 68.32,
+      "learning_rate": 9.98076923076923e-05,
+      "loss": 0.6763,
+      "step": 7447
+    },
+    {
+      "epoch": 68.33,
+      "learning_rate": 9.977884615384613e-05,
+      "loss": 0.7642,
+      "step": 7448
+    },
+    {
+      "epoch": 68.34,
+      "learning_rate": 9.975e-05,
+      "loss": 0.6485,
+      "step": 7449
+    },
+    {
+      "epoch": 68.35,
+      "learning_rate": 9.972115384615384e-05,
+      "loss": 0.5091,
+      "step": 7450
+    },
+    {
+      "epoch": 68.36,
+      "learning_rate": 9.969230769230769e-05,
+      "loss": 0.6145,
+      "step": 7451
+    },
+    {
+      "epoch": 68.37,
+      "learning_rate": 9.966346153846153e-05,
+      "loss": 0.6066,
+      "step": 7452
+    },
+    {
+      "epoch": 68.38,
+      "learning_rate": 9.963461538461536e-05,
+      "loss": 0.5712,
+      "step": 7453
+    },
+    {
+      "epoch": 68.39,
+      "learning_rate": 9.960576923076923e-05,
+      "loss": 0.627,
+      "step": 7454
+    },
+    {
+      "epoch": 68.39,
+      "learning_rate": 9.957692307692307e-05,
+      "loss": 0.7209,
+      "step": 7455
+    },
+    {
+      "epoch": 68.4,
+      "learning_rate": 9.954807692307692e-05,
+      "loss": 0.5427,
+      "step": 7456
+    },
+    {
+      "epoch": 68.41,
+      "learning_rate": 9.951923076923076e-05,
+      "loss": 0.7161,
+      "step": 7457
+    },
+    {
+      "epoch": 68.42,
+      "learning_rate": 9.94903846153846e-05,
+      "loss": 0.606,
+      "step": 7458
+    },
+    {
+      "epoch": 68.43,
+      "learning_rate": 9.946153846153846e-05,
+      "loss": 0.6255,
+      "step": 7459
+    },
+    {
+      "epoch": 68.44,
+      "learning_rate": 9.94326923076923e-05,
+      "loss": 0.6121,
+      "step": 7460
+    },
+    {
+      "epoch": 68.45,
+      "learning_rate": 9.940384615384615e-05,
+      "loss": 0.6489,
+      "step": 7461
+    },
+    {
+      "epoch": 68.46,
+      "learning_rate": 9.937499999999999e-05,
+      "loss": 0.5252,
+      "step": 7462
+    },
+    {
+      "epoch": 68.47,
+      "learning_rate": 9.934615384615383e-05,
+      "loss": 0.6285,
+      "step": 7463
+    },
+    {
+      "epoch": 68.48,
+      "learning_rate": 9.931730769230769e-05,
+      "loss": 0.6262,
+      "step": 7464
+    },
+    {
+      "epoch": 68.49,
+      "learning_rate": 9.928846153846153e-05,
+      "loss": 0.6105,
+      "step": 7465
+    },
+    {
+      "epoch": 68.5,
+      "learning_rate": 9.925961538461538e-05,
+      "loss": 0.6912,
+      "step": 7466
+    },
+    {
+      "epoch": 68.5,
+      "learning_rate": 9.923076923076922e-05,
+      "loss": 0.6013,
+      "step": 7467
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 9.920192307692306e-05,
+      "loss": 0.7252,
+      "step": 7468
+    },
+    {
+      "epoch": 68.52,
+      "learning_rate": 9.917307692307692e-05,
+      "loss": 0.637,
+      "step": 7469
+    },
+    {
+      "epoch": 68.53,
+      "learning_rate": 9.914423076923076e-05,
+      "loss": 0.6669,
+      "step": 7470
+    },
+    {
+      "epoch": 68.54,
+      "learning_rate": 9.911538461538461e-05,
+      "loss": 0.5937,
+      "step": 7471
+    },
+    {
+      "epoch": 68.55,
+      "learning_rate": 9.908653846153845e-05,
+      "loss": 0.6838,
+      "step": 7472
+    },
+    {
+      "epoch": 68.56,
+      "learning_rate": 9.905769230769229e-05,
+      "loss": 0.5885,
+      "step": 7473
+    },
+    {
+      "epoch": 68.57,
+      "learning_rate": 9.902884615384616e-05,
+      "loss": 0.5786,
+      "step": 7474
+    },
+    {
+      "epoch": 68.58,
+      "learning_rate": 9.9e-05,
+      "loss": 0.6523,
+      "step": 7475
+    },
+    {
+      "epoch": 68.59,
+      "learning_rate": 9.897115384615383e-05,
+      "loss": 0.7215,
+      "step": 7476
+    },
+    {
+      "epoch": 68.6,
+      "learning_rate": 9.894230769230768e-05,
+      "loss": 0.6293,
+      "step": 7477
+    },
+    {
+      "epoch": 68.61,
+      "learning_rate": 9.891346153846152e-05,
+      "loss": 0.5963,
+      "step": 7478
+    },
+    {
+      "epoch": 68.61,
+      "learning_rate": 9.888461538461537e-05,
+      "loss": 0.6133,
+      "step": 7479
+    },
+    {
+      "epoch": 68.62,
+      "learning_rate": 9.885576923076923e-05,
+      "loss": 0.6326,
+      "step": 7480
+    },
+    {
+      "epoch": 68.63,
+      "learning_rate": 9.882692307692306e-05,
+      "loss": 0.6013,
+      "step": 7481
+    },
+    {
+      "epoch": 68.64,
+      "learning_rate": 9.879807692307692e-05,
+      "loss": 0.6055,
+      "step": 7482
+    },
+    {
+      "epoch": 68.65,
+      "learning_rate": 9.876923076923075e-05,
+      "loss": 0.5433,
+      "step": 7483
+    },
+    {
+      "epoch": 68.66,
+      "learning_rate": 9.87403846153846e-05,
+      "loss": 0.7149,
+      "step": 7484
+    },
+    {
+      "epoch": 68.67,
+      "learning_rate": 9.871153846153846e-05,
+      "loss": 0.6858,
+      "step": 7485
+    },
+    {
+      "epoch": 68.68,
+      "learning_rate": 9.86826923076923e-05,
+      "loss": 0.6515,
+      "step": 7486
+    },
+    {
+      "epoch": 68.69,
+      "learning_rate": 9.865384615384615e-05,
+      "loss": 0.6915,
+      "step": 7487
+    },
+    {
+      "epoch": 68.7,
+      "learning_rate": 9.862499999999998e-05,
+      "loss": 0.6459,
+      "step": 7488
+    },
+    {
+      "epoch": 68.71,
+      "learning_rate": 9.859615384615384e-05,
+      "loss": 0.7043,
+      "step": 7489
+    },
+    {
+      "epoch": 68.72,
+      "learning_rate": 9.856730769230769e-05,
+      "loss": 0.5538,
+      "step": 7490
+    },
+    {
+      "epoch": 68.72,
+      "learning_rate": 9.853846153846153e-05,
+      "loss": 0.6913,
+      "step": 7491
+    },
+    {
+      "epoch": 68.73,
+      "learning_rate": 9.850961538461538e-05,
+      "loss": 0.6582,
+      "step": 7492
+    },
+    {
+      "epoch": 68.74,
+      "learning_rate": 9.848076923076922e-05,
+      "loss": 0.5738,
+      "step": 7493
+    },
+    {
+      "epoch": 68.75,
+      "learning_rate": 9.845192307692307e-05,
+      "loss": 0.6607,
+      "step": 7494
+    },
+    {
+      "epoch": 68.76,
+      "learning_rate": 9.842307692307692e-05,
+      "loss": 0.688,
+      "step": 7495
+    },
+    {
+      "epoch": 68.77,
+      "learning_rate": 9.839423076923076e-05,
+      "loss": 0.5479,
+      "step": 7496
+    },
+    {
+      "epoch": 68.78,
+      "learning_rate": 9.836538461538461e-05,
+      "loss": 0.6016,
+      "step": 7497
+    },
+    {
+      "epoch": 68.79,
+      "learning_rate": 9.833653846153845e-05,
+      "loss": 0.63,
+      "step": 7498
+    },
+    {
+      "epoch": 68.8,
+      "learning_rate": 9.83076923076923e-05,
+      "loss": 0.59,
+      "step": 7499
+    },
+    {
+      "epoch": 68.81,
+      "learning_rate": 9.827884615384615e-05,
+      "loss": 0.7238,
+      "step": 7500
+    },
+    {
+      "epoch": 68.81,
+      "eval_cer": 0.09037616755873401,
+      "eval_loss": 0.3673691153526306,
+      "eval_runtime": 13.9272,
+      "eval_samples_per_second": 118.258,
+      "eval_steps_per_second": 1.867,
+      "eval_wer": 0.37340414666530486,
+      "step": 7500
+    },
+    {
+      "epoch": 68.82,
+      "learning_rate": 9.824999999999999e-05,
+      "loss": 0.6506,
+      "step": 7501
+    },
+    {
+      "epoch": 68.83,
+      "learning_rate": 9.822115384615384e-05,
+      "loss": 0.5781,
+      "step": 7502
+    },
+    {
+      "epoch": 68.83,
+      "learning_rate": 9.819230769230768e-05,
+      "loss": 0.5983,
+      "step": 7503
+    },
+    {
+      "epoch": 68.84,
+      "learning_rate": 9.816346153846153e-05,
+      "loss": 0.6839,
+      "step": 7504
+    },
+    {
+      "epoch": 68.85,
+      "learning_rate": 9.813461538461538e-05,
+      "loss": 0.5945,
+      "step": 7505
+    },
+    {
+      "epoch": 68.86,
+      "learning_rate": 9.810576923076922e-05,
+      "loss": 0.6176,
+      "step": 7506
+    },
+    {
+      "epoch": 68.87,
+      "learning_rate": 9.807692307692307e-05,
+      "loss": 0.7009,
+      "step": 7507
+    },
+    {
+      "epoch": 68.88,
+      "learning_rate": 9.804807692307691e-05,
+      "loss": 0.5981,
+      "step": 7508
+    },
+    {
+      "epoch": 68.89,
+      "learning_rate": 9.801923076923075e-05,
+      "loss": 0.5432,
+      "step": 7509
+    },
+    {
+      "epoch": 68.9,
+      "learning_rate": 9.799038461538461e-05,
+      "loss": 0.5179,
+      "step": 7510
+    },
+    {
+      "epoch": 68.91,
+      "learning_rate": 9.796153846153845e-05,
+      "loss": 0.6406,
+      "step": 7511
+    },
+    {
+      "epoch": 68.92,
+      "learning_rate": 9.79326923076923e-05,
+      "loss": 0.7383,
+      "step": 7512
+    },
+    {
+      "epoch": 68.93,
+      "learning_rate": 9.790384615384614e-05,
+      "loss": 0.6268,
+      "step": 7513
+    },
+    {
+      "epoch": 68.94,
+      "learning_rate": 9.787499999999998e-05,
+      "loss": 0.6773,
+      "step": 7514
+    },
+    {
+      "epoch": 68.94,
+      "learning_rate": 9.784615384615383e-05,
+      "loss": 0.6847,
+      "step": 7515
+    },
+    {
+      "epoch": 68.95,
+      "learning_rate": 9.781730769230768e-05,
+      "loss": 0.5947,
+      "step": 7516
+    },
+    {
+      "epoch": 68.96,
+      "learning_rate": 9.778846153846154e-05,
+      "loss": 0.6112,
+      "step": 7517
+    },
+    {
+      "epoch": 68.97,
+      "learning_rate": 9.775961538461537e-05,
+      "loss": 0.677,
+      "step": 7518
+    },
+    {
+      "epoch": 68.98,
+      "learning_rate": 9.773076923076921e-05,
+      "loss": 0.5296,
+      "step": 7519
+    },
+    {
+      "epoch": 68.99,
+      "learning_rate": 9.770192307692306e-05,
+      "loss": 0.657,
+      "step": 7520
+    },
+    {
+      "epoch": 69.0,
+      "learning_rate": 9.767307692307692e-05,
+      "loss": 0.7363,
+      "step": 7521
+    },
+    {
+      "epoch": 69.01,
+      "learning_rate": 9.764423076923077e-05,
+      "loss": 0.5122,
+      "step": 7522
+    },
+    {
+      "epoch": 69.02,
+      "learning_rate": 9.76153846153846e-05,
+      "loss": 0.6032,
+      "step": 7523
+    },
+    {
+      "epoch": 69.03,
+      "learning_rate": 9.758653846153844e-05,
+      "loss": 0.6262,
+      "step": 7524
+    },
+    {
+      "epoch": 69.04,
+      "learning_rate": 9.75576923076923e-05,
+      "loss": 0.7111,
+      "step": 7525
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 9.752884615384615e-05,
+      "loss": 0.7332,
+      "step": 7526
+    },
+    {
+      "epoch": 69.06,
+      "learning_rate": 9.75e-05,
+      "loss": 0.5051,
+      "step": 7527
+    },
+    {
+      "epoch": 69.06,
+      "learning_rate": 9.747115384615384e-05,
+      "loss": 0.5438,
+      "step": 7528
+    },
+    {
+      "epoch": 69.07,
+      "learning_rate": 9.744230769230767e-05,
+      "loss": 0.5682,
+      "step": 7529
+    },
+    {
+      "epoch": 69.08,
+      "learning_rate": 9.741346153846153e-05,
+      "loss": 0.8023,
+      "step": 7530
+    },
+    {
+      "epoch": 69.09,
+      "learning_rate": 9.738461538461538e-05,
+      "loss": 0.7189,
+      "step": 7531
+    },
+    {
+      "epoch": 69.1,
+      "learning_rate": 9.735576923076923e-05,
+      "loss": 0.559,
+      "step": 7532
+    },
+    {
+      "epoch": 69.11,
+      "learning_rate": 9.732692307692307e-05,
+      "loss": 0.6414,
+      "step": 7533
+    },
+    {
+      "epoch": 69.12,
+      "learning_rate": 9.72980769230769e-05,
+      "loss": 0.6324,
+      "step": 7534
+    },
+    {
+      "epoch": 69.13,
+      "learning_rate": 9.726923076923076e-05,
+      "loss": 0.7163,
+      "step": 7535
+    },
+    {
+      "epoch": 69.14,
+      "learning_rate": 9.724038461538461e-05,
+      "loss": 0.6207,
+      "step": 7536
+    },
+    {
+      "epoch": 69.15,
+      "learning_rate": 9.721153846153846e-05,
+      "loss": 0.543,
+      "step": 7537
+    },
+    {
+      "epoch": 69.16,
+      "learning_rate": 9.71826923076923e-05,
+      "loss": 0.5094,
+      "step": 7538
+    },
+    {
+      "epoch": 69.17,
+      "learning_rate": 9.715384615384614e-05,
+      "loss": 0.4965,
+      "step": 7539
+    },
+    {
+      "epoch": 69.17,
+      "learning_rate": 9.712499999999999e-05,
+      "loss": 0.5914,
+      "step": 7540
+    },
+    {
+      "epoch": 69.18,
+      "learning_rate": 9.709615384615384e-05,
+      "loss": 0.6072,
+      "step": 7541
+    },
+    {
+      "epoch": 69.19,
+      "learning_rate": 9.706730769230769e-05,
+      "loss": 0.6962,
+      "step": 7542
+    },
+    {
+      "epoch": 69.2,
+      "learning_rate": 9.703846153846153e-05,
+      "loss": 0.5649,
+      "step": 7543
+    },
+    {
+      "epoch": 69.21,
+      "learning_rate": 9.700961538461537e-05,
+      "loss": 0.6517,
+      "step": 7544
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 9.698076923076922e-05,
+      "loss": 0.6724,
+      "step": 7545
+    },
+    {
+      "epoch": 69.23,
+      "learning_rate": 9.695192307692306e-05,
+      "loss": 0.5906,
+      "step": 7546
+    },
+    {
+      "epoch": 69.24,
+      "learning_rate": 9.692307692307692e-05,
+      "loss": 0.6372,
+      "step": 7547
+    },
+    {
+      "epoch": 69.25,
+      "learning_rate": 9.689423076923076e-05,
+      "loss": 0.6019,
+      "step": 7548
+    },
+    {
+      "epoch": 69.26,
+      "learning_rate": 9.68653846153846e-05,
+      "loss": 0.6765,
+      "step": 7549
+    },
+    {
+      "epoch": 69.27,
+      "learning_rate": 9.683653846153845e-05,
+      "loss": 0.5459,
+      "step": 7550
+    },
+    {
+      "epoch": 69.28,
+      "learning_rate": 9.680769230769229e-05,
+      "loss": 0.7305,
+      "step": 7551
+    },
+    {
+      "epoch": 69.28,
+      "learning_rate": 9.677884615384616e-05,
+      "loss": 0.5867,
+      "step": 7552
+    },
+    {
+      "epoch": 69.29,
+      "learning_rate": 9.675e-05,
+      "loss": 0.6588,
+      "step": 7553
+    },
+    {
+      "epoch": 69.3,
+      "learning_rate": 9.672115384615383e-05,
+      "loss": 0.5595,
+      "step": 7554
+    },
+    {
+      "epoch": 69.31,
+      "learning_rate": 9.669230769230768e-05,
+      "loss": 0.6285,
+      "step": 7555
+    },
+    {
+      "epoch": 69.32,
+      "learning_rate": 9.666346153846152e-05,
+      "loss": 0.6173,
+      "step": 7556
+    },
+    {
+      "epoch": 69.33,
+      "learning_rate": 9.663461538461539e-05,
+      "loss": 0.6805,
+      "step": 7557
+    },
+    {
+      "epoch": 69.34,
+      "learning_rate": 9.660576923076922e-05,
+      "loss": 0.5642,
+      "step": 7558
+    },
+    {
+      "epoch": 69.35,
+      "learning_rate": 9.657692307692306e-05,
+      "loss": 0.5326,
+      "step": 7559
+    },
+    {
+      "epoch": 69.36,
+      "learning_rate": 9.654807692307691e-05,
+      "loss": 0.5847,
+      "step": 7560
+    },
+    {
+      "epoch": 69.37,
+      "learning_rate": 9.651923076923075e-05,
+      "loss": 0.6155,
+      "step": 7561
+    },
+    {
+      "epoch": 69.38,
+      "learning_rate": 9.649038461538462e-05,
+      "loss": 0.5894,
+      "step": 7562
+    },
+    {
+      "epoch": 69.39,
+      "learning_rate": 9.646153846153846e-05,
+      "loss": 0.696,
+      "step": 7563
+    },
+    {
+      "epoch": 69.39,
+      "learning_rate": 9.64326923076923e-05,
+      "loss": 0.6778,
+      "step": 7564
+    },
+    {
+      "epoch": 69.4,
+      "learning_rate": 9.640384615384615e-05,
+      "loss": 0.5763,
+      "step": 7565
+    },
+    {
+      "epoch": 69.41,
+      "learning_rate": 9.637499999999998e-05,
+      "loss": 0.6197,
+      "step": 7566
+    },
+    {
+      "epoch": 69.42,
+      "learning_rate": 9.634615384615385e-05,
+      "loss": 0.6019,
+      "step": 7567
+    },
+    {
+      "epoch": 69.43,
+      "learning_rate": 9.631730769230769e-05,
+      "loss": 0.6825,
+      "step": 7568
+    },
+    {
+      "epoch": 69.44,
+      "learning_rate": 9.628846153846153e-05,
+      "loss": 0.6003,
+      "step": 7569
+    },
+    {
+      "epoch": 69.45,
+      "learning_rate": 9.625961538461538e-05,
+      "loss": 0.8221,
+      "step": 7570
+    },
+    {
+      "epoch": 69.46,
+      "learning_rate": 9.623076923076922e-05,
+      "loss": 0.4115,
+      "step": 7571
+    },
+    {
+      "epoch": 69.47,
+      "learning_rate": 9.620192307692308e-05,
+      "loss": 0.6136,
+      "step": 7572
+    },
+    {
+      "epoch": 69.48,
+      "learning_rate": 9.617307692307692e-05,
+      "loss": 0.5746,
+      "step": 7573
+    },
+    {
+      "epoch": 69.49,
+      "learning_rate": 9.614423076923076e-05,
+      "loss": 0.6477,
+      "step": 7574
+    },
+    {
+      "epoch": 69.5,
+      "learning_rate": 9.611538461538461e-05,
+      "loss": 0.7553,
+      "step": 7575
+    },
+    {
+      "epoch": 69.5,
+      "learning_rate": 9.608653846153845e-05,
+      "loss": 0.7386,
+      "step": 7576
+    },
+    {
+      "epoch": 69.51,
+      "learning_rate": 9.605769230769229e-05,
+      "loss": 0.6614,
+      "step": 7577
+    },
+    {
+      "epoch": 69.52,
+      "learning_rate": 9.602884615384615e-05,
+      "loss": 0.6324,
+      "step": 7578
+    },
+    {
+      "epoch": 69.53,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 0.6353,
+      "step": 7579
+    },
+    {
+      "epoch": 69.54,
+      "learning_rate": 9.597115384615384e-05,
+      "loss": 0.783,
+      "step": 7580
+    },
+    {
+      "epoch": 69.55,
+      "learning_rate": 9.594230769230768e-05,
+      "loss": 0.6968,
+      "step": 7581
+    },
+    {
+      "epoch": 69.56,
+      "learning_rate": 9.591346153846152e-05,
+      "loss": 0.6233,
+      "step": 7582
+    },
+    {
+      "epoch": 69.57,
+      "learning_rate": 9.588461538461538e-05,
+      "loss": 0.5865,
+      "step": 7583
+    },
+    {
+      "epoch": 69.58,
+      "learning_rate": 9.585576923076922e-05,
+      "loss": 0.5405,
+      "step": 7584
+    },
+    {
+      "epoch": 69.59,
+      "learning_rate": 9.582692307692307e-05,
+      "loss": 0.6919,
+      "step": 7585
+    },
+    {
+      "epoch": 69.6,
+      "learning_rate": 9.579807692307691e-05,
+      "loss": 0.5752,
+      "step": 7586
+    },
+    {
+      "epoch": 69.61,
+      "learning_rate": 9.576923076923075e-05,
+      "loss": 0.6843,
+      "step": 7587
+    },
+    {
+      "epoch": 69.61,
+      "learning_rate": 9.574038461538461e-05,
+      "loss": 0.6308,
+      "step": 7588
+    },
+    {
+      "epoch": 69.62,
+      "learning_rate": 9.571153846153845e-05,
+      "loss": 0.6301,
+      "step": 7589
+    },
+    {
+      "epoch": 69.63,
+      "learning_rate": 9.56826923076923e-05,
+      "loss": 0.6207,
+      "step": 7590
+    },
+    {
+      "epoch": 69.64,
+      "learning_rate": 9.565384615384614e-05,
+      "loss": 0.4837,
+      "step": 7591
+    },
+    {
+      "epoch": 69.65,
+      "learning_rate": 9.562499999999998e-05,
+      "loss": 0.5824,
+      "step": 7592
+    },
+    {
+      "epoch": 69.66,
+      "learning_rate": 9.559615384615385e-05,
+      "loss": 0.774,
+      "step": 7593
+    },
+    {
+      "epoch": 69.67,
+      "learning_rate": 9.556730769230768e-05,
+      "loss": 0.5199,
+      "step": 7594
+    },
+    {
+      "epoch": 69.68,
+      "learning_rate": 9.553846153846153e-05,
+      "loss": 0.6645,
+      "step": 7595
+    },
+    {
+      "epoch": 69.69,
+      "learning_rate": 9.550961538461537e-05,
+      "loss": 0.5338,
+      "step": 7596
+    },
+    {
+      "epoch": 69.7,
+      "learning_rate": 9.548076923076921e-05,
+      "loss": 0.7306,
+      "step": 7597
+    },
+    {
+      "epoch": 69.71,
+      "learning_rate": 9.545192307692308e-05,
+      "loss": 0.5505,
+      "step": 7598
+    },
+    {
+      "epoch": 69.72,
+      "learning_rate": 9.542307692307691e-05,
+      "loss": 0.6583,
+      "step": 7599
+    },
+    {
+      "epoch": 69.72,
+      "learning_rate": 9.539423076923077e-05,
+      "loss": 0.5847,
+      "step": 7600
+    },
+    {
+      "epoch": 69.73,
+      "learning_rate": 9.53653846153846e-05,
+      "loss": 0.5585,
+      "step": 7601
+    },
+    {
+      "epoch": 69.74,
+      "learning_rate": 9.533653846153844e-05,
+      "loss": 0.6728,
+      "step": 7602
+    },
+    {
+      "epoch": 69.75,
+      "learning_rate": 9.530769230769231e-05,
+      "loss": 0.6341,
+      "step": 7603
+    },
+    {
+      "epoch": 69.76,
+      "learning_rate": 9.527884615384615e-05,
+      "loss": 0.5655,
+      "step": 7604
+    },
+    {
+      "epoch": 69.77,
+      "learning_rate": 9.525e-05,
+      "loss": 0.6218,
+      "step": 7605
+    },
+    {
+      "epoch": 69.78,
+      "learning_rate": 9.522115384615384e-05,
+      "loss": 0.7182,
+      "step": 7606
+    },
+    {
+      "epoch": 69.79,
+      "learning_rate": 9.519230769230767e-05,
+      "loss": 0.7092,
+      "step": 7607
+    },
+    {
+      "epoch": 69.8,
+      "learning_rate": 9.516346153846154e-05,
+      "loss": 0.6725,
+      "step": 7608
+    },
+    {
+      "epoch": 69.81,
+      "learning_rate": 9.513461538461538e-05,
+      "loss": 0.5803,
+      "step": 7609
+    },
+    {
+      "epoch": 69.82,
+      "learning_rate": 9.510576923076923e-05,
+      "loss": 0.6269,
+      "step": 7610
+    },
+    {
+      "epoch": 69.83,
+      "learning_rate": 9.507692307692307e-05,
+      "loss": 0.536,
+      "step": 7611
+    },
+    {
+      "epoch": 69.83,
+      "learning_rate": 9.50480769230769e-05,
+      "loss": 0.6097,
+      "step": 7612
+    },
+    {
+      "epoch": 69.84,
+      "learning_rate": 9.501923076923076e-05,
+      "loss": 0.513,
+      "step": 7613
+    },
+    {
+      "epoch": 69.85,
+      "learning_rate": 9.499038461538461e-05,
+      "loss": 0.6486,
+      "step": 7614
+    },
+    {
+      "epoch": 69.86,
+      "learning_rate": 9.496153846153846e-05,
+      "loss": 0.6584,
+      "step": 7615
+    },
+    {
+      "epoch": 69.87,
+      "learning_rate": 9.49326923076923e-05,
+      "loss": 0.6709,
+      "step": 7616
+    },
+    {
+      "epoch": 69.88,
+      "learning_rate": 9.490384615384614e-05,
+      "loss": 0.7296,
+      "step": 7617
+    },
+    {
+      "epoch": 69.89,
+      "learning_rate": 9.487499999999999e-05,
+      "loss": 0.5885,
+      "step": 7618
+    },
+    {
+      "epoch": 69.9,
+      "learning_rate": 9.484615384615384e-05,
+      "loss": 0.6241,
+      "step": 7619
+    },
+    {
+      "epoch": 69.91,
+      "learning_rate": 9.481730769230769e-05,
+      "loss": 0.7358,
+      "step": 7620
+    },
+    {
+      "epoch": 69.92,
+      "learning_rate": 9.478846153846153e-05,
+      "loss": 0.5824,
+      "step": 7621
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 9.475961538461537e-05,
+      "loss": 0.7567,
+      "step": 7622
+    },
+    {
+      "epoch": 69.94,
+      "learning_rate": 9.473076923076922e-05,
+      "loss": 0.656,
+      "step": 7623
+    },
+    {
+      "epoch": 69.94,
+      "learning_rate": 9.470192307692307e-05,
+      "loss": 0.7133,
+      "step": 7624
+    },
+    {
+      "epoch": 69.95,
+      "learning_rate": 9.467307692307692e-05,
+      "loss": 0.6478,
+      "step": 7625
+    },
+    {
+      "epoch": 69.96,
+      "learning_rate": 9.464423076923076e-05,
+      "loss": 0.6809,
+      "step": 7626
+    },
+    {
+      "epoch": 69.97,
+      "learning_rate": 9.46153846153846e-05,
+      "loss": 0.5564,
+      "step": 7627
+    },
+    {
+      "epoch": 69.98,
+      "learning_rate": 9.458653846153845e-05,
+      "loss": 0.6732,
+      "step": 7628
+    },
+    {
+      "epoch": 69.99,
+      "learning_rate": 9.45576923076923e-05,
+      "loss": 0.5942,
+      "step": 7629
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 9.452884615384615e-05,
+      "loss": 0.6542,
+      "step": 7630
+    },
+    {
+      "epoch": 70.01,
+      "learning_rate": 9.449999999999999e-05,
+      "loss": 0.6066,
+      "step": 7631
+    },
+    {
+      "epoch": 70.02,
+      "learning_rate": 9.447115384615383e-05,
+      "loss": 0.6316,
+      "step": 7632
+    },
+    {
+      "epoch": 70.03,
+      "learning_rate": 9.444230769230768e-05,
+      "loss": 0.7088,
+      "step": 7633
+    },
+    {
+      "epoch": 70.04,
+      "learning_rate": 9.441346153846153e-05,
+      "loss": 0.7593,
+      "step": 7634
+    },
+    {
+      "epoch": 70.05,
+      "learning_rate": 9.438461538461539e-05,
+      "loss": 0.7096,
+      "step": 7635
+    },
+    {
+      "epoch": 70.06,
+      "learning_rate": 9.435576923076922e-05,
+      "loss": 0.5576,
+      "step": 7636
+    },
+    {
+      "epoch": 70.06,
+      "learning_rate": 9.432692307692306e-05,
+      "loss": 0.6991,
+      "step": 7637
+    },
+    {
+      "epoch": 70.07,
+      "learning_rate": 9.429807692307691e-05,
+      "loss": 0.533,
+      "step": 7638
+    },
+    {
+      "epoch": 70.08,
+      "learning_rate": 9.426923076923077e-05,
+      "loss": 0.6038,
+      "step": 7639
+    },
+    {
+      "epoch": 70.09,
+      "learning_rate": 9.424038461538462e-05,
+      "loss": 0.6138,
+      "step": 7640
+    },
+    {
+      "epoch": 70.1,
+      "learning_rate": 9.421153846153846e-05,
+      "loss": 0.7137,
+      "step": 7641
+    },
+    {
+      "epoch": 70.11,
+      "learning_rate": 9.41826923076923e-05,
+      "loss": 0.7638,
+      "step": 7642
+    },
+    {
+      "epoch": 70.12,
+      "learning_rate": 9.415384615384615e-05,
+      "loss": 0.5201,
+      "step": 7643
+    },
+    {
+      "epoch": 70.13,
+      "learning_rate": 9.412499999999998e-05,
+      "loss": 0.6814,
+      "step": 7644
+    },
+    {
+      "epoch": 70.14,
+      "learning_rate": 9.409615384615385e-05,
+      "loss": 0.5522,
+      "step": 7645
+    },
+    {
+      "epoch": 70.15,
+      "learning_rate": 9.406730769230769e-05,
+      "loss": 0.6068,
+      "step": 7646
+    },
+    {
+      "epoch": 70.16,
+      "learning_rate": 9.403846153846153e-05,
+      "loss": 0.6993,
+      "step": 7647
+    },
+    {
+      "epoch": 70.17,
+      "learning_rate": 9.400961538461538e-05,
+      "loss": 0.5882,
+      "step": 7648
+    },
+    {
+      "epoch": 70.17,
+      "learning_rate": 9.398076923076922e-05,
+      "loss": 0.5233,
+      "step": 7649
+    },
+    {
+      "epoch": 70.18,
+      "learning_rate": 9.395192307692308e-05,
+      "loss": 0.5272,
+      "step": 7650
+    },
+    {
+      "epoch": 70.19,
+      "learning_rate": 9.392307692307692e-05,
+      "loss": 0.5506,
+      "step": 7651
+    },
+    {
+      "epoch": 70.2,
+      "learning_rate": 9.389423076923076e-05,
+      "loss": 0.63,
+      "step": 7652
+    },
+    {
+      "epoch": 70.21,
+      "learning_rate": 9.386538461538461e-05,
+      "loss": 0.645,
+      "step": 7653
+    },
+    {
+      "epoch": 70.22,
+      "learning_rate": 9.383653846153845e-05,
+      "loss": 0.5959,
+      "step": 7654
+    },
+    {
+      "epoch": 70.23,
+      "learning_rate": 9.380769230769231e-05,
+      "loss": 0.6028,
+      "step": 7655
+    },
+    {
+      "epoch": 70.24,
+      "learning_rate": 9.377884615384615e-05,
+      "loss": 0.6701,
+      "step": 7656
+    },
+    {
+      "epoch": 70.25,
+      "learning_rate": 9.374999999999999e-05,
+      "loss": 0.7639,
+      "step": 7657
+    },
+    {
+      "epoch": 70.26,
+      "learning_rate": 9.372115384615384e-05,
+      "loss": 0.5631,
+      "step": 7658
+    },
+    {
+      "epoch": 70.27,
+      "learning_rate": 9.369230769230768e-05,
+      "loss": 0.7543,
+      "step": 7659
+    },
+    {
+      "epoch": 70.28,
+      "learning_rate": 9.366346153846154e-05,
+      "loss": 0.4809,
+      "step": 7660
+    },
+    {
+      "epoch": 70.28,
+      "learning_rate": 9.363461538461538e-05,
+      "loss": 0.6536,
+      "step": 7661
+    },
+    {
+      "epoch": 70.29,
+      "learning_rate": 9.360576923076922e-05,
+      "loss": 0.6118,
+      "step": 7662
+    },
+    {
+      "epoch": 70.3,
+      "learning_rate": 9.357692307692307e-05,
+      "loss": 0.6106,
+      "step": 7663
+    },
+    {
+      "epoch": 70.31,
+      "learning_rate": 9.354807692307691e-05,
+      "loss": 0.5898,
+      "step": 7664
+    },
+    {
+      "epoch": 70.32,
+      "learning_rate": 9.351923076923077e-05,
+      "loss": 0.518,
+      "step": 7665
+    },
+    {
+      "epoch": 70.33,
+      "learning_rate": 9.349038461538461e-05,
+      "loss": 0.7205,
+      "step": 7666
+    },
+    {
+      "epoch": 70.34,
+      "learning_rate": 9.346153846153845e-05,
+      "loss": 0.716,
+      "step": 7667
+    },
+    {
+      "epoch": 70.35,
+      "learning_rate": 9.34326923076923e-05,
+      "loss": 0.6129,
+      "step": 7668
+    },
+    {
+      "epoch": 70.36,
+      "learning_rate": 9.340384615384614e-05,
+      "loss": 0.6781,
+      "step": 7669
+    },
+    {
+      "epoch": 70.37,
+      "learning_rate": 9.3375e-05,
+      "loss": 0.5682,
+      "step": 7670
+    },
+    {
+      "epoch": 70.38,
+      "learning_rate": 9.334615384615384e-05,
+      "loss": 0.6285,
+      "step": 7671
+    },
+    {
+      "epoch": 70.39,
+      "learning_rate": 9.331730769230768e-05,
+      "loss": 0.7751,
+      "step": 7672
+    },
+    {
+      "epoch": 70.39,
+      "learning_rate": 9.328846153846153e-05,
+      "loss": 0.6294,
+      "step": 7673
+    },
+    {
+      "epoch": 70.4,
+      "learning_rate": 9.325961538461537e-05,
+      "loss": 0.6542,
+      "step": 7674
+    },
+    {
+      "epoch": 70.41,
+      "learning_rate": 9.323076923076921e-05,
+      "loss": 0.5668,
+      "step": 7675
+    },
+    {
+      "epoch": 70.42,
+      "learning_rate": 9.320192307692308e-05,
+      "loss": 0.7248,
+      "step": 7676
+    },
+    {
+      "epoch": 70.43,
+      "learning_rate": 9.317307692307691e-05,
+      "loss": 0.6337,
+      "step": 7677
+    },
+    {
+      "epoch": 70.44,
+      "learning_rate": 9.314423076923077e-05,
+      "loss": 0.6459,
+      "step": 7678
+    },
+    {
+      "epoch": 70.45,
+      "learning_rate": 9.31153846153846e-05,
+      "loss": 0.5628,
+      "step": 7679
+    },
+    {
+      "epoch": 70.46,
+      "learning_rate": 9.308653846153844e-05,
+      "loss": 0.6468,
+      "step": 7680
+    },
+    {
+      "epoch": 70.47,
+      "learning_rate": 9.305769230769231e-05,
+      "loss": 0.5525,
+      "step": 7681
+    },
+    {
+      "epoch": 70.48,
+      "learning_rate": 9.302884615384615e-05,
+      "loss": 0.6262,
+      "step": 7682
+    },
+    {
+      "epoch": 70.49,
+      "learning_rate": 9.3e-05,
+      "loss": 0.6197,
+      "step": 7683
+    },
+    {
+      "epoch": 70.5,
+      "learning_rate": 9.297115384615384e-05,
+      "loss": 0.5868,
+      "step": 7684
+    },
+    {
+      "epoch": 70.5,
+      "learning_rate": 9.294230769230767e-05,
+      "loss": 0.6518,
+      "step": 7685
+    },
+    {
+      "epoch": 70.51,
+      "learning_rate": 9.291346153846154e-05,
+      "loss": 0.6074,
+      "step": 7686
+    },
+    {
+      "epoch": 70.52,
+      "learning_rate": 9.288461538461538e-05,
+      "loss": 0.6441,
+      "step": 7687
+    },
+    {
+      "epoch": 70.53,
+      "learning_rate": 9.285576923076923e-05,
+      "loss": 0.6234,
+      "step": 7688
+    },
+    {
+      "epoch": 70.54,
+      "learning_rate": 9.282692307692307e-05,
+      "loss": 0.587,
+      "step": 7689
+    },
+    {
+      "epoch": 70.55,
+      "learning_rate": 9.27980769230769e-05,
+      "loss": 0.6033,
+      "step": 7690
+    },
+    {
+      "epoch": 70.56,
+      "learning_rate": 9.276923076923077e-05,
+      "loss": 0.6608,
+      "step": 7691
+    },
+    {
+      "epoch": 70.57,
+      "learning_rate": 9.274038461538461e-05,
+      "loss": 0.5288,
+      "step": 7692
+    },
+    {
+      "epoch": 70.58,
+      "learning_rate": 9.271153846153846e-05,
+      "loss": 0.6053,
+      "step": 7693
+    },
+    {
+      "epoch": 70.59,
+      "learning_rate": 9.26826923076923e-05,
+      "loss": 0.5548,
+      "step": 7694
+    },
+    {
+      "epoch": 70.6,
+      "learning_rate": 9.265384615384614e-05,
+      "loss": 0.5324,
+      "step": 7695
+    },
+    {
+      "epoch": 70.61,
+      "learning_rate": 9.2625e-05,
+      "loss": 0.5663,
+      "step": 7696
+    },
+    {
+      "epoch": 70.61,
+      "learning_rate": 9.259615384615384e-05,
+      "loss": 0.5765,
+      "step": 7697
+    },
+    {
+      "epoch": 70.62,
+      "learning_rate": 9.256730769230769e-05,
+      "loss": 0.6719,
+      "step": 7698
+    },
+    {
+      "epoch": 70.63,
+      "learning_rate": 9.253846153846153e-05,
+      "loss": 0.6021,
+      "step": 7699
+    },
+    {
+      "epoch": 70.64,
+      "learning_rate": 9.250961538461537e-05,
+      "loss": 0.7301,
+      "step": 7700
+    },
+    {
+      "epoch": 70.65,
+      "learning_rate": 9.248076923076923e-05,
+      "loss": 0.5486,
+      "step": 7701
+    },
+    {
+      "epoch": 70.66,
+      "learning_rate": 9.245192307692307e-05,
+      "loss": 0.5917,
+      "step": 7702
+    },
+    {
+      "epoch": 70.67,
+      "learning_rate": 9.242307692307692e-05,
+      "loss": 0.6742,
+      "step": 7703
+    },
+    {
+      "epoch": 70.68,
+      "learning_rate": 9.239423076923076e-05,
+      "loss": 0.6047,
+      "step": 7704
+    },
+    {
+      "epoch": 70.69,
+      "learning_rate": 9.23653846153846e-05,
+      "loss": 0.6859,
+      "step": 7705
+    },
+    {
+      "epoch": 70.7,
+      "learning_rate": 9.233653846153845e-05,
+      "loss": 0.6059,
+      "step": 7706
+    },
+    {
+      "epoch": 70.71,
+      "learning_rate": 9.23076923076923e-05,
+      "loss": 0.6119,
+      "step": 7707
+    },
+    {
+      "epoch": 70.72,
+      "learning_rate": 9.227884615384614e-05,
+      "loss": 0.4763,
+      "step": 7708
+    },
+    {
+      "epoch": 70.72,
+      "learning_rate": 9.224999999999999e-05,
+      "loss": 0.7331,
+      "step": 7709
+    },
+    {
+      "epoch": 70.73,
+      "learning_rate": 9.222115384615383e-05,
+      "loss": 0.5413,
+      "step": 7710
+    },
+    {
+      "epoch": 70.74,
+      "learning_rate": 9.219230769230768e-05,
+      "loss": 0.5843,
+      "step": 7711
+    },
+    {
+      "epoch": 70.75,
+      "learning_rate": 9.216346153846153e-05,
+      "loss": 0.5838,
+      "step": 7712
+    },
+    {
+      "epoch": 70.76,
+      "learning_rate": 9.213461538461537e-05,
+      "loss": 0.693,
+      "step": 7713
+    },
+    {
+      "epoch": 70.77,
+      "learning_rate": 9.210576923076922e-05,
+      "loss": 0.5783,
+      "step": 7714
+    },
+    {
+      "epoch": 70.78,
+      "learning_rate": 9.207692307692306e-05,
+      "loss": 0.6215,
+      "step": 7715
+    },
+    {
+      "epoch": 70.79,
+      "learning_rate": 9.204807692307691e-05,
+      "loss": 0.6304,
+      "step": 7716
+    },
+    {
+      "epoch": 70.8,
+      "learning_rate": 9.201923076923077e-05,
+      "loss": 0.6842,
+      "step": 7717
+    },
+    {
+      "epoch": 70.81,
+      "learning_rate": 9.19903846153846e-05,
+      "loss": 0.544,
+      "step": 7718
+    },
+    {
+      "epoch": 70.82,
+      "learning_rate": 9.196153846153846e-05,
+      "loss": 0.7283,
+      "step": 7719
+    },
+    {
+      "epoch": 70.83,
+      "learning_rate": 9.19326923076923e-05,
+      "loss": 0.6478,
+      "step": 7720
+    },
+    {
+      "epoch": 70.83,
+      "learning_rate": 9.190384615384615e-05,
+      "loss": 0.6312,
+      "step": 7721
+    },
+    {
+      "epoch": 70.84,
+      "learning_rate": 9.1875e-05,
+      "loss": 0.7105,
+      "step": 7722
+    },
+    {
+      "epoch": 70.85,
+      "learning_rate": 9.184615384615384e-05,
+      "loss": 0.7524,
+      "step": 7723
+    },
+    {
+      "epoch": 70.86,
+      "learning_rate": 9.181730769230769e-05,
+      "loss": 0.729,
+      "step": 7724
+    },
+    {
+      "epoch": 70.87,
+      "learning_rate": 9.178846153846153e-05,
+      "loss": 0.5817,
+      "step": 7725
+    },
+    {
+      "epoch": 70.88,
+      "learning_rate": 9.175961538461538e-05,
+      "loss": 0.5782,
+      "step": 7726
+    },
+    {
+      "epoch": 70.89,
+      "learning_rate": 9.173076923076923e-05,
+      "loss": 0.5923,
+      "step": 7727
+    },
+    {
+      "epoch": 70.9,
+      "learning_rate": 9.170192307692307e-05,
+      "loss": 0.6072,
+      "step": 7728
+    },
+    {
+      "epoch": 70.91,
+      "learning_rate": 9.167307692307692e-05,
+      "loss": 0.6121,
+      "step": 7729
+    },
+    {
+      "epoch": 70.92,
+      "learning_rate": 9.164423076923076e-05,
+      "loss": 0.7277,
+      "step": 7730
+    },
+    {
+      "epoch": 70.93,
+      "learning_rate": 9.161538461538461e-05,
+      "loss": 0.5178,
+      "step": 7731
+    },
+    {
+      "epoch": 70.94,
+      "learning_rate": 9.158653846153846e-05,
+      "loss": 0.6093,
+      "step": 7732
+    },
+    {
+      "epoch": 70.94,
+      "learning_rate": 9.15576923076923e-05,
+      "loss": 0.6247,
+      "step": 7733
+    },
+    {
+      "epoch": 70.95,
+      "learning_rate": 9.152884615384615e-05,
+      "loss": 0.767,
+      "step": 7734
+    },
+    {
+      "epoch": 70.96,
+      "learning_rate": 9.149999999999999e-05,
+      "loss": 0.6775,
+      "step": 7735
+    },
+    {
+      "epoch": 70.97,
+      "learning_rate": 9.147115384615384e-05,
+      "loss": 0.6129,
+      "step": 7736
+    },
+    {
+      "epoch": 70.98,
+      "learning_rate": 9.144230769230769e-05,
+      "loss": 0.6058,
+      "step": 7737
+    },
+    {
+      "epoch": 70.99,
+      "learning_rate": 9.141346153846153e-05,
+      "loss": 0.7698,
+      "step": 7738
+    },
+    {
+      "epoch": 71.0,
+      "learning_rate": 9.138461538461538e-05,
+      "loss": 0.5209,
+      "step": 7739
+    },
+    {
+      "epoch": 71.01,
+      "learning_rate": 9.135576923076922e-05,
+      "loss": 0.5969,
+      "step": 7740
+    },
+    {
+      "epoch": 71.02,
+      "learning_rate": 9.132692307692306e-05,
+      "loss": 0.6802,
+      "step": 7741
+    },
+    {
+      "epoch": 71.03,
+      "learning_rate": 9.129807692307691e-05,
+      "loss": 0.6366,
+      "step": 7742
+    },
+    {
+      "epoch": 71.04,
+      "learning_rate": 9.126923076923076e-05,
+      "loss": 0.6223,
+      "step": 7743
+    },
+    {
+      "epoch": 71.05,
+      "learning_rate": 9.124038461538461e-05,
+      "loss": 0.6426,
+      "step": 7744
+    },
+    {
+      "epoch": 71.06,
+      "learning_rate": 9.121153846153845e-05,
+      "loss": 0.6043,
+      "step": 7745
+    },
+    {
+      "epoch": 71.06,
+      "learning_rate": 9.118269230769229e-05,
+      "loss": 0.5774,
+      "step": 7746
+    },
+    {
+      "epoch": 71.07,
+      "learning_rate": 9.115384615384614e-05,
+      "loss": 0.6878,
+      "step": 7747
+    },
+    {
+      "epoch": 71.08,
+      "learning_rate": 9.112499999999999e-05,
+      "loss": 0.5226,
+      "step": 7748
+    },
+    {
+      "epoch": 71.09,
+      "learning_rate": 9.109615384615384e-05,
+      "loss": 0.6739,
+      "step": 7749
+    },
+    {
+      "epoch": 71.1,
+      "learning_rate": 9.106730769230768e-05,
+      "loss": 0.6602,
+      "step": 7750
+    },
+    {
+      "epoch": 71.11,
+      "learning_rate": 9.103846153846152e-05,
+      "loss": 0.5827,
+      "step": 7751
+    },
+    {
+      "epoch": 71.12,
+      "learning_rate": 9.100961538461537e-05,
+      "loss": 0.6623,
+      "step": 7752
+    },
+    {
+      "epoch": 71.13,
+      "learning_rate": 9.098076923076922e-05,
+      "loss": 0.6344,
+      "step": 7753
+    },
+    {
+      "epoch": 71.14,
+      "learning_rate": 9.095192307692308e-05,
+      "loss": 0.6637,
+      "step": 7754
+    },
+    {
+      "epoch": 71.15,
+      "learning_rate": 9.092307692307691e-05,
+      "loss": 0.7726,
+      "step": 7755
+    },
+    {
+      "epoch": 71.16,
+      "learning_rate": 9.089423076923075e-05,
+      "loss": 0.6256,
+      "step": 7756
+    },
+    {
+      "epoch": 71.17,
+      "learning_rate": 9.08653846153846e-05,
+      "loss": 0.6117,
+      "step": 7757
+    },
+    {
+      "epoch": 71.17,
+      "learning_rate": 9.083653846153846e-05,
+      "loss": 0.7843,
+      "step": 7758
+    },
+    {
+      "epoch": 71.18,
+      "learning_rate": 9.080769230769231e-05,
+      "loss": 0.6168,
+      "step": 7759
+    },
+    {
+      "epoch": 71.19,
+      "learning_rate": 9.077884615384615e-05,
+      "loss": 0.5828,
+      "step": 7760
+    },
+    {
+      "epoch": 71.2,
+      "learning_rate": 9.074999999999998e-05,
+      "loss": 0.6165,
+      "step": 7761
+    },
+    {
+      "epoch": 71.21,
+      "learning_rate": 9.072115384615383e-05,
+      "loss": 0.6285,
+      "step": 7762
+    },
+    {
+      "epoch": 71.22,
+      "learning_rate": 9.069230769230769e-05,
+      "loss": 0.5305,
+      "step": 7763
+    },
+    {
+      "epoch": 71.23,
+      "learning_rate": 9.066346153846154e-05,
+      "loss": 0.5932,
+      "step": 7764
+    },
+    {
+      "epoch": 71.24,
+      "learning_rate": 9.063461538461538e-05,
+      "loss": 0.5534,
+      "step": 7765
+    },
+    {
+      "epoch": 71.25,
+      "learning_rate": 9.060576923076921e-05,
+      "loss": 0.6026,
+      "step": 7766
+    },
+    {
+      "epoch": 71.26,
+      "learning_rate": 9.057692307692307e-05,
+      "loss": 0.6282,
+      "step": 7767
+    },
+    {
+      "epoch": 71.27,
+      "learning_rate": 9.054807692307692e-05,
+      "loss": 0.6389,
+      "step": 7768
+    },
+    {
+      "epoch": 71.28,
+      "learning_rate": 9.051923076923077e-05,
+      "loss": 0.6495,
+      "step": 7769
+    },
+    {
+      "epoch": 71.28,
+      "learning_rate": 9.049038461538461e-05,
+      "loss": 0.4664,
+      "step": 7770
+    },
+    {
+      "epoch": 71.29,
+      "learning_rate": 9.046153846153845e-05,
+      "loss": 0.6803,
+      "step": 7771
+    },
+    {
+      "epoch": 71.3,
+      "learning_rate": 9.04326923076923e-05,
+      "loss": 0.5535,
+      "step": 7772
+    },
+    {
+      "epoch": 71.31,
+      "learning_rate": 9.040384615384614e-05,
+      "loss": 0.5436,
+      "step": 7773
+    },
+    {
+      "epoch": 71.32,
+      "learning_rate": 9.0375e-05,
+      "loss": 0.5745,
+      "step": 7774
+    },
+    {
+      "epoch": 71.33,
+      "learning_rate": 9.034615384615384e-05,
+      "loss": 0.7458,
+      "step": 7775
+    },
+    {
+      "epoch": 71.34,
+      "learning_rate": 9.031730769230768e-05,
+      "loss": 0.6659,
+      "step": 7776
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 9.028846153846153e-05,
+      "loss": 0.5473,
+      "step": 7777
+    },
+    {
+      "epoch": 71.36,
+      "learning_rate": 9.025961538461537e-05,
+      "loss": 0.6867,
+      "step": 7778
+    },
+    {
+      "epoch": 71.37,
+      "learning_rate": 9.023076923076923e-05,
+      "loss": 0.6172,
+      "step": 7779
+    },
+    {
+      "epoch": 71.38,
+      "learning_rate": 9.020192307692307e-05,
+      "loss": 0.5911,
+      "step": 7780
+    },
+    {
+      "epoch": 71.39,
+      "learning_rate": 9.017307692307691e-05,
+      "loss": 0.5626,
+      "step": 7781
+    },
+    {
+      "epoch": 71.39,
+      "learning_rate": 9.014423076923076e-05,
+      "loss": 0.572,
+      "step": 7782
+    },
+    {
+      "epoch": 71.4,
+      "learning_rate": 9.01153846153846e-05,
+      "loss": 0.6621,
+      "step": 7783
+    },
+    {
+      "epoch": 71.41,
+      "learning_rate": 9.008653846153846e-05,
+      "loss": 0.5388,
+      "step": 7784
+    },
+    {
+      "epoch": 71.42,
+      "learning_rate": 9.00576923076923e-05,
+      "loss": 0.6528,
+      "step": 7785
+    },
+    {
+      "epoch": 71.43,
+      "learning_rate": 9.002884615384614e-05,
+      "loss": 0.5854,
+      "step": 7786
+    },
+    {
+      "epoch": 71.44,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 0.5339,
+      "step": 7787
+    },
+    {
+      "epoch": 71.45,
+      "learning_rate": 8.997115384615383e-05,
+      "loss": 0.6297,
+      "step": 7788
+    },
+    {
+      "epoch": 71.46,
+      "learning_rate": 8.99423076923077e-05,
+      "loss": 0.4648,
+      "step": 7789
+    },
+    {
+      "epoch": 71.47,
+      "learning_rate": 8.991346153846153e-05,
+      "loss": 0.5842,
+      "step": 7790
+    },
+    {
+      "epoch": 71.48,
+      "learning_rate": 8.988461538461537e-05,
+      "loss": 0.5701,
+      "step": 7791
+    },
+    {
+      "epoch": 71.49,
+      "learning_rate": 8.985576923076922e-05,
+      "loss": 0.6425,
+      "step": 7792
+    },
+    {
+      "epoch": 71.5,
+      "learning_rate": 8.982692307692306e-05,
+      "loss": 0.6136,
+      "step": 7793
+    },
+    {
+      "epoch": 71.5,
+      "learning_rate": 8.979807692307693e-05,
+      "loss": 0.6506,
+      "step": 7794
+    },
+    {
+      "epoch": 71.51,
+      "learning_rate": 8.976923076923077e-05,
+      "loss": 0.526,
+      "step": 7795
+    },
+    {
+      "epoch": 71.52,
+      "learning_rate": 8.97403846153846e-05,
+      "loss": 0.622,
+      "step": 7796
+    },
+    {
+      "epoch": 71.53,
+      "learning_rate": 8.971153846153845e-05,
+      "loss": 0.5911,
+      "step": 7797
+    },
+    {
+      "epoch": 71.54,
+      "learning_rate": 8.968269230769229e-05,
+      "loss": 0.5979,
+      "step": 7798
+    },
+    {
+      "epoch": 71.55,
+      "learning_rate": 8.965384615384616e-05,
+      "loss": 0.56,
+      "step": 7799
+    },
+    {
+      "epoch": 71.56,
+      "learning_rate": 8.9625e-05,
+      "loss": 0.6148,
+      "step": 7800
+    },
+    {
+      "epoch": 71.57,
+      "learning_rate": 8.959615384615383e-05,
+      "loss": 0.7311,
+      "step": 7801
+    },
+    {
+      "epoch": 71.58,
+      "learning_rate": 8.956730769230769e-05,
+      "loss": 0.6233,
+      "step": 7802
+    },
+    {
+      "epoch": 71.59,
+      "learning_rate": 8.953846153846152e-05,
+      "loss": 0.5767,
+      "step": 7803
+    },
+    {
+      "epoch": 71.6,
+      "learning_rate": 8.950961538461536e-05,
+      "loss": 0.6122,
+      "step": 7804
+    },
+    {
+      "epoch": 71.61,
+      "learning_rate": 8.948076923076923e-05,
+      "loss": 0.6686,
+      "step": 7805
+    },
+    {
+      "epoch": 71.61,
+      "learning_rate": 8.945192307692307e-05,
+      "loss": 0.6979,
+      "step": 7806
+    },
+    {
+      "epoch": 71.62,
+      "learning_rate": 8.942307692307692e-05,
+      "loss": 0.6315,
+      "step": 7807
+    },
+    {
+      "epoch": 71.63,
+      "learning_rate": 8.939423076923076e-05,
+      "loss": 0.514,
+      "step": 7808
+    },
+    {
+      "epoch": 71.64,
+      "learning_rate": 8.93653846153846e-05,
+      "loss": 0.5392,
+      "step": 7809
+    },
+    {
+      "epoch": 71.65,
+      "learning_rate": 8.933653846153846e-05,
+      "loss": 0.4891,
+      "step": 7810
+    },
+    {
+      "epoch": 71.66,
+      "learning_rate": 8.93076923076923e-05,
+      "loss": 0.7019,
+      "step": 7811
+    },
+    {
+      "epoch": 71.67,
+      "learning_rate": 8.927884615384615e-05,
+      "loss": 0.6399,
+      "step": 7812
+    },
+    {
+      "epoch": 71.68,
+      "learning_rate": 8.924999999999999e-05,
+      "loss": 0.5384,
+      "step": 7813
+    },
+    {
+      "epoch": 71.69,
+      "learning_rate": 8.922115384615383e-05,
+      "loss": 0.7063,
+      "step": 7814
+    },
+    {
+      "epoch": 71.7,
+      "learning_rate": 8.919230769230769e-05,
+      "loss": 0.6848,
+      "step": 7815
+    },
+    {
+      "epoch": 71.71,
+      "learning_rate": 8.916346153846153e-05,
+      "loss": 0.7356,
+      "step": 7816
+    },
+    {
+      "epoch": 71.72,
+      "learning_rate": 8.913461538461538e-05,
+      "loss": 0.6528,
+      "step": 7817
+    },
+    {
+      "epoch": 71.72,
+      "learning_rate": 8.910576923076922e-05,
+      "loss": 0.6119,
+      "step": 7818
+    },
+    {
+      "epoch": 71.73,
+      "learning_rate": 8.907692307692306e-05,
+      "loss": 0.6276,
+      "step": 7819
+    },
+    {
+      "epoch": 71.74,
+      "learning_rate": 8.904807692307692e-05,
+      "loss": 0.6166,
+      "step": 7820
+    },
+    {
+      "epoch": 71.75,
+      "learning_rate": 8.901923076923076e-05,
+      "loss": 0.6395,
+      "step": 7821
+    },
+    {
+      "epoch": 71.76,
+      "learning_rate": 8.899038461538461e-05,
+      "loss": 0.5721,
+      "step": 7822
+    },
+    {
+      "epoch": 71.77,
+      "learning_rate": 8.896153846153845e-05,
+      "loss": 0.4862,
+      "step": 7823
+    },
+    {
+      "epoch": 71.78,
+      "learning_rate": 8.893269230769229e-05,
+      "loss": 0.6034,
+      "step": 7824
+    },
+    {
+      "epoch": 71.79,
+      "learning_rate": 8.890384615384615e-05,
+      "loss": 0.6254,
+      "step": 7825
+    },
+    {
+      "epoch": 71.8,
+      "learning_rate": 8.887499999999999e-05,
+      "loss": 0.6034,
+      "step": 7826
+    },
+    {
+      "epoch": 71.81,
+      "learning_rate": 8.884615384615384e-05,
+      "loss": 0.5842,
+      "step": 7827
+    },
+    {
+      "epoch": 71.82,
+      "learning_rate": 8.881730769230768e-05,
+      "loss": 0.6475,
+      "step": 7828
+    },
+    {
+      "epoch": 71.83,
+      "learning_rate": 8.878846153846152e-05,
+      "loss": 0.687,
+      "step": 7829
+    },
+    {
+      "epoch": 71.83,
+      "learning_rate": 8.875961538461539e-05,
+      "loss": 0.7091,
+      "step": 7830
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 8.873076923076922e-05,
+      "loss": 0.7007,
+      "step": 7831
+    },
+    {
+      "epoch": 71.85,
+      "learning_rate": 8.870192307692308e-05,
+      "loss": 0.6106,
+      "step": 7832
+    },
+    {
+      "epoch": 71.86,
+      "learning_rate": 8.867307692307691e-05,
+      "loss": 0.6571,
+      "step": 7833
+    },
+    {
+      "epoch": 71.87,
+      "learning_rate": 8.864423076923075e-05,
+      "loss": 0.7357,
+      "step": 7834
+    },
+    {
+      "epoch": 71.88,
+      "learning_rate": 8.861538461538462e-05,
+      "loss": 0.5527,
+      "step": 7835
+    },
+    {
+      "epoch": 71.89,
+      "learning_rate": 8.858653846153845e-05,
+      "loss": 0.6203,
+      "step": 7836
+    },
+    {
+      "epoch": 71.9,
+      "learning_rate": 8.85576923076923e-05,
+      "loss": 0.5633,
+      "step": 7837
+    },
+    {
+      "epoch": 71.91,
+      "learning_rate": 8.852884615384614e-05,
+      "loss": 0.5999,
+      "step": 7838
+    },
+    {
+      "epoch": 71.92,
+      "learning_rate": 8.849999999999998e-05,
+      "loss": 0.6235,
+      "step": 7839
+    },
+    {
+      "epoch": 71.93,
+      "learning_rate": 8.847115384615383e-05,
+      "loss": 0.6755,
+      "step": 7840
+    },
+    {
+      "epoch": 71.94,
+      "learning_rate": 8.844230769230769e-05,
+      "loss": 0.6634,
+      "step": 7841
+    },
+    {
+      "epoch": 71.94,
+      "learning_rate": 8.841346153846154e-05,
+      "loss": 0.559,
+      "step": 7842
+    },
+    {
+      "epoch": 71.95,
+      "learning_rate": 8.838461538461538e-05,
+      "loss": 0.6808,
+      "step": 7843
+    },
+    {
+      "epoch": 71.96,
+      "learning_rate": 8.835576923076921e-05,
+      "loss": 0.6141,
+      "step": 7844
+    },
+    {
+      "epoch": 71.97,
+      "learning_rate": 8.832692307692307e-05,
+      "loss": 0.5449,
+      "step": 7845
+    },
+    {
+      "epoch": 71.98,
+      "learning_rate": 8.829807692307692e-05,
+      "loss": 0.7118,
+      "step": 7846
+    },
+    {
+      "epoch": 71.99,
+      "learning_rate": 8.826923076923077e-05,
+      "loss": 0.6396,
+      "step": 7847
+    },
+    {
+      "epoch": 72.0,
+      "learning_rate": 8.824038461538461e-05,
+      "loss": 0.753,
+      "step": 7848
+    },
+    {
+      "epoch": 72.01,
+      "learning_rate": 8.821153846153845e-05,
+      "loss": 0.6138,
+      "step": 7849
+    },
+    {
+      "epoch": 72.02,
+      "learning_rate": 8.81826923076923e-05,
+      "loss": 0.583,
+      "step": 7850
+    },
+    {
+      "epoch": 72.03,
+      "learning_rate": 8.815384615384615e-05,
+      "loss": 0.4643,
+      "step": 7851
+    },
+    {
+      "epoch": 72.04,
+      "learning_rate": 8.8125e-05,
+      "loss": 0.5274,
+      "step": 7852
+    },
+    {
+      "epoch": 72.05,
+      "learning_rate": 8.809615384615384e-05,
+      "loss": 0.6749,
+      "step": 7853
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 8.806730769230768e-05,
+      "loss": 0.5758,
+      "step": 7854
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 8.803846153846153e-05,
+      "loss": 0.6262,
+      "step": 7855
+    },
+    {
+      "epoch": 72.07,
+      "learning_rate": 8.800961538461538e-05,
+      "loss": 0.502,
+      "step": 7856
+    },
+    {
+      "epoch": 72.08,
+      "learning_rate": 8.798076923076923e-05,
+      "loss": 0.7337,
+      "step": 7857
+    },
+    {
+      "epoch": 72.09,
+      "learning_rate": 8.795192307692307e-05,
+      "loss": 0.5937,
+      "step": 7858
+    },
+    {
+      "epoch": 72.1,
+      "learning_rate": 8.792307692307691e-05,
+      "loss": 0.626,
+      "step": 7859
+    },
+    {
+      "epoch": 72.11,
+      "learning_rate": 8.789423076923076e-05,
+      "loss": 0.6615,
+      "step": 7860
+    },
+    {
+      "epoch": 72.12,
+      "learning_rate": 8.786538461538461e-05,
+      "loss": 0.4326,
+      "step": 7861
+    },
+    {
+      "epoch": 72.13,
+      "learning_rate": 8.783653846153846e-05,
+      "loss": 0.6174,
+      "step": 7862
+    },
+    {
+      "epoch": 72.14,
+      "learning_rate": 8.78076923076923e-05,
+      "loss": 0.6803,
+      "step": 7863
+    },
+    {
+      "epoch": 72.15,
+      "learning_rate": 8.777884615384614e-05,
+      "loss": 0.6576,
+      "step": 7864
+    },
+    {
+      "epoch": 72.16,
+      "learning_rate": 8.774999999999999e-05,
+      "loss": 0.5443,
+      "step": 7865
+    },
+    {
+      "epoch": 72.17,
+      "learning_rate": 8.772115384615384e-05,
+      "loss": 0.5657,
+      "step": 7866
+    },
+    {
+      "epoch": 72.17,
+      "learning_rate": 8.76923076923077e-05,
+      "loss": 0.749,
+      "step": 7867
+    },
+    {
+      "epoch": 72.18,
+      "learning_rate": 8.766346153846153e-05,
+      "loss": 0.604,
+      "step": 7868
+    },
+    {
+      "epoch": 72.19,
+      "learning_rate": 8.763461538461537e-05,
+      "loss": 0.5983,
+      "step": 7869
+    },
+    {
+      "epoch": 72.2,
+      "learning_rate": 8.760576923076922e-05,
+      "loss": 0.5901,
+      "step": 7870
+    },
+    {
+      "epoch": 72.21,
+      "learning_rate": 8.757692307692306e-05,
+      "loss": 0.5996,
+      "step": 7871
+    },
+    {
+      "epoch": 72.22,
+      "learning_rate": 8.754807692307693e-05,
+      "loss": 0.6016,
+      "step": 7872
+    },
+    {
+      "epoch": 72.23,
+      "learning_rate": 8.751923076923076e-05,
+      "loss": 0.5769,
+      "step": 7873
+    },
+    {
+      "epoch": 72.24,
+      "learning_rate": 8.74903846153846e-05,
+      "loss": 0.7151,
+      "step": 7874
+    },
+    {
+      "epoch": 72.25,
+      "learning_rate": 8.746153846153845e-05,
+      "loss": 0.5813,
+      "step": 7875
+    },
+    {
+      "epoch": 72.26,
+      "learning_rate": 8.743269230769229e-05,
+      "loss": 0.5878,
+      "step": 7876
+    },
+    {
+      "epoch": 72.27,
+      "learning_rate": 8.740384615384616e-05,
+      "loss": 0.7557,
+      "step": 7877
+    },
+    {
+      "epoch": 72.28,
+      "learning_rate": 8.7375e-05,
+      "loss": 0.6447,
+      "step": 7878
+    },
+    {
+      "epoch": 72.28,
+      "learning_rate": 8.734615384615383e-05,
+      "loss": 0.5236,
+      "step": 7879
+    },
+    {
+      "epoch": 72.29,
+      "learning_rate": 8.731730769230769e-05,
+      "loss": 0.6209,
+      "step": 7880
+    },
+    {
+      "epoch": 72.3,
+      "learning_rate": 8.728846153846152e-05,
+      "loss": 0.6774,
+      "step": 7881
+    },
+    {
+      "epoch": 72.31,
+      "learning_rate": 8.725961538461539e-05,
+      "loss": 0.6696,
+      "step": 7882
+    },
+    {
+      "epoch": 72.32,
+      "learning_rate": 8.723076923076923e-05,
+      "loss": 0.7097,
+      "step": 7883
+    },
+    {
+      "epoch": 72.33,
+      "learning_rate": 8.720192307692307e-05,
+      "loss": 0.6724,
+      "step": 7884
+    },
+    {
+      "epoch": 72.34,
+      "learning_rate": 8.717307692307692e-05,
+      "loss": 0.633,
+      "step": 7885
+    },
+    {
+      "epoch": 72.35,
+      "learning_rate": 8.714423076923076e-05,
+      "loss": 0.6094,
+      "step": 7886
+    },
+    {
+      "epoch": 72.36,
+      "learning_rate": 8.711538461538462e-05,
+      "loss": 0.5866,
+      "step": 7887
+    },
+    {
+      "epoch": 72.37,
+      "learning_rate": 8.708653846153846e-05,
+      "loss": 0.6036,
+      "step": 7888
+    },
+    {
+      "epoch": 72.38,
+      "learning_rate": 8.70576923076923e-05,
+      "loss": 0.6522,
+      "step": 7889
+    },
+    {
+      "epoch": 72.39,
+      "learning_rate": 8.702884615384615e-05,
+      "loss": 0.7618,
+      "step": 7890
+    },
+    {
+      "epoch": 72.39,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 0.6977,
+      "step": 7891
+    },
+    {
+      "epoch": 72.4,
+      "learning_rate": 8.697115384615385e-05,
+      "loss": 0.5535,
+      "step": 7892
+    },
+    {
+      "epoch": 72.41,
+      "learning_rate": 8.694230769230769e-05,
+      "loss": 0.5712,
+      "step": 7893
+    },
+    {
+      "epoch": 72.42,
+      "learning_rate": 8.691346153846153e-05,
+      "loss": 0.5968,
+      "step": 7894
+    },
+    {
+      "epoch": 72.43,
+      "learning_rate": 8.688461538461538e-05,
+      "loss": 0.5627,
+      "step": 7895
+    },
+    {
+      "epoch": 72.44,
+      "learning_rate": 8.685576923076922e-05,
+      "loss": 0.6253,
+      "step": 7896
+    },
+    {
+      "epoch": 72.45,
+      "learning_rate": 8.682692307692308e-05,
+      "loss": 0.6536,
+      "step": 7897
+    },
+    {
+      "epoch": 72.46,
+      "learning_rate": 8.679807692307692e-05,
+      "loss": 0.5712,
+      "step": 7898
+    },
+    {
+      "epoch": 72.47,
+      "learning_rate": 8.676923076923076e-05,
+      "loss": 0.5664,
+      "step": 7899
+    },
+    {
+      "epoch": 72.48,
+      "learning_rate": 8.674038461538461e-05,
+      "loss": 0.7033,
+      "step": 7900
+    },
+    {
+      "epoch": 72.49,
+      "learning_rate": 8.671153846153845e-05,
+      "loss": 0.6084,
+      "step": 7901
+    },
+    {
+      "epoch": 72.5,
+      "learning_rate": 8.668269230769229e-05,
+      "loss": 0.5261,
+      "step": 7902
+    },
+    {
+      "epoch": 72.5,
+      "learning_rate": 8.665384615384615e-05,
+      "loss": 0.7268,
+      "step": 7903
+    },
+    {
+      "epoch": 72.51,
+      "learning_rate": 8.662499999999999e-05,
+      "loss": 0.6113,
+      "step": 7904
+    },
+    {
+      "epoch": 72.52,
+      "learning_rate": 8.659615384615384e-05,
+      "loss": 0.5333,
+      "step": 7905
+    },
+    {
+      "epoch": 72.53,
+      "learning_rate": 8.656730769230768e-05,
+      "loss": 0.589,
+      "step": 7906
+    },
+    {
+      "epoch": 72.54,
+      "learning_rate": 8.653846153846152e-05,
+      "loss": 0.669,
+      "step": 7907
+    },
+    {
+      "epoch": 72.55,
+      "learning_rate": 8.650961538461538e-05,
+      "loss": 0.6087,
+      "step": 7908
+    },
+    {
+      "epoch": 72.56,
+      "learning_rate": 8.648076923076922e-05,
+      "loss": 0.5977,
+      "step": 7909
+    },
+    {
+      "epoch": 72.57,
+      "learning_rate": 8.645192307692307e-05,
+      "loss": 0.5939,
+      "step": 7910
+    },
+    {
+      "epoch": 72.58,
+      "learning_rate": 8.642307692307691e-05,
+      "loss": 0.7651,
+      "step": 7911
+    },
+    {
+      "epoch": 72.59,
+      "learning_rate": 8.639423076923075e-05,
+      "loss": 0.7631,
+      "step": 7912
+    },
+    {
+      "epoch": 72.6,
+      "learning_rate": 8.636538461538462e-05,
+      "loss": 0.6696,
+      "step": 7913
+    },
+    {
+      "epoch": 72.61,
+      "learning_rate": 8.633653846153845e-05,
+      "loss": 0.6536,
+      "step": 7914
+    },
+    {
+      "epoch": 72.61,
+      "learning_rate": 8.63076923076923e-05,
+      "loss": 0.5501,
+      "step": 7915
+    },
+    {
+      "epoch": 72.62,
+      "learning_rate": 8.627884615384614e-05,
+      "loss": 0.6133,
+      "step": 7916
+    },
+    {
+      "epoch": 72.63,
+      "learning_rate": 8.624999999999998e-05,
+      "loss": 0.5029,
+      "step": 7917
+    },
+    {
+      "epoch": 72.64,
+      "learning_rate": 8.622115384615385e-05,
+      "loss": 0.6272,
+      "step": 7918
+    },
+    {
+      "epoch": 72.65,
+      "learning_rate": 8.619230769230769e-05,
+      "loss": 0.6087,
+      "step": 7919
+    },
+    {
+      "epoch": 72.66,
+      "learning_rate": 8.616346153846154e-05,
+      "loss": 0.5033,
+      "step": 7920
+    },
+    {
+      "epoch": 72.67,
+      "learning_rate": 8.613461538461538e-05,
+      "loss": 0.7579,
+      "step": 7921
+    },
+    {
+      "epoch": 72.68,
+      "learning_rate": 8.610576923076921e-05,
+      "loss": 0.7218,
+      "step": 7922
+    },
+    {
+      "epoch": 72.69,
+      "learning_rate": 8.607692307692308e-05,
+      "loss": 0.5813,
+      "step": 7923
+    },
+    {
+      "epoch": 72.7,
+      "learning_rate": 8.604807692307692e-05,
+      "loss": 0.6486,
+      "step": 7924
+    },
+    {
+      "epoch": 72.71,
+      "learning_rate": 8.601923076923077e-05,
+      "loss": 0.6025,
+      "step": 7925
+    },
+    {
+      "epoch": 72.72,
+      "learning_rate": 8.599038461538461e-05,
+      "loss": 0.4728,
+      "step": 7926
+    },
+    {
+      "epoch": 72.72,
+      "learning_rate": 8.596153846153845e-05,
+      "loss": 0.7265,
+      "step": 7927
+    },
+    {
+      "epoch": 72.73,
+      "learning_rate": 8.593269230769231e-05,
+      "loss": 0.5944,
+      "step": 7928
+    },
+    {
+      "epoch": 72.74,
+      "learning_rate": 8.590384615384615e-05,
+      "loss": 0.6738,
+      "step": 7929
+    },
+    {
+      "epoch": 72.75,
+      "learning_rate": 8.5875e-05,
+      "loss": 0.6058,
+      "step": 7930
+    },
+    {
+      "epoch": 72.76,
+      "learning_rate": 8.584615384615384e-05,
+      "loss": 0.6409,
+      "step": 7931
+    },
+    {
+      "epoch": 72.77,
+      "learning_rate": 8.581730769230768e-05,
+      "loss": 0.6748,
+      "step": 7932
+    },
+    {
+      "epoch": 72.78,
+      "learning_rate": 8.578846153846154e-05,
+      "loss": 0.7058,
+      "step": 7933
+    },
+    {
+      "epoch": 72.79,
+      "learning_rate": 8.575961538461538e-05,
+      "loss": 0.6427,
+      "step": 7934
+    },
+    {
+      "epoch": 72.8,
+      "learning_rate": 8.573076923076923e-05,
+      "loss": 0.6314,
+      "step": 7935
+    },
+    {
+      "epoch": 72.81,
+      "learning_rate": 8.570192307692307e-05,
+      "loss": 0.6142,
+      "step": 7936
+    },
+    {
+      "epoch": 72.82,
+      "learning_rate": 8.567307692307691e-05,
+      "loss": 0.6236,
+      "step": 7937
+    },
+    {
+      "epoch": 72.83,
+      "learning_rate": 8.564423076923076e-05,
+      "loss": 0.5527,
+      "step": 7938
+    },
+    {
+      "epoch": 72.83,
+      "learning_rate": 8.561538461538461e-05,
+      "loss": 0.5499,
+      "step": 7939
+    },
+    {
+      "epoch": 72.84,
+      "learning_rate": 8.558653846153846e-05,
+      "loss": 0.6299,
+      "step": 7940
+    },
+    {
+      "epoch": 72.85,
+      "learning_rate": 8.55576923076923e-05,
+      "loss": 0.5631,
+      "step": 7941
+    },
+    {
+      "epoch": 72.86,
+      "learning_rate": 8.552884615384614e-05,
+      "loss": 0.538,
+      "step": 7942
+    },
+    {
+      "epoch": 72.87,
+      "learning_rate": 8.549999999999999e-05,
+      "loss": 0.5696,
+      "step": 7943
+    },
+    {
+      "epoch": 72.88,
+      "learning_rate": 8.547115384615384e-05,
+      "loss": 0.5536,
+      "step": 7944
+    },
+    {
+      "epoch": 72.89,
+      "learning_rate": 8.544230769230768e-05,
+      "loss": 0.6332,
+      "step": 7945
+    },
+    {
+      "epoch": 72.9,
+      "learning_rate": 8.541346153846153e-05,
+      "loss": 0.595,
+      "step": 7946
+    },
+    {
+      "epoch": 72.91,
+      "learning_rate": 8.538461538461537e-05,
+      "loss": 0.5634,
+      "step": 7947
+    },
+    {
+      "epoch": 72.92,
+      "learning_rate": 8.535576923076922e-05,
+      "loss": 0.6162,
+      "step": 7948
+    },
+    {
+      "epoch": 72.93,
+      "learning_rate": 8.532692307692307e-05,
+      "loss": 0.6644,
+      "step": 7949
+    },
+    {
+      "epoch": 72.94,
+      "learning_rate": 8.529807692307691e-05,
+      "loss": 0.5858,
+      "step": 7950
+    },
+    {
+      "epoch": 72.94,
+      "learning_rate": 8.526923076923076e-05,
+      "loss": 0.6842,
+      "step": 7951
+    },
+    {
+      "epoch": 72.95,
+      "learning_rate": 8.52403846153846e-05,
+      "loss": 0.6737,
+      "step": 7952
+    },
+    {
+      "epoch": 72.96,
+      "learning_rate": 8.521153846153845e-05,
+      "loss": 0.5602,
+      "step": 7953
+    },
+    {
+      "epoch": 72.97,
+      "learning_rate": 8.51826923076923e-05,
+      "loss": 0.5888,
+      "step": 7954
+    },
+    {
+      "epoch": 72.98,
+      "learning_rate": 8.515384615384614e-05,
+      "loss": 0.7034,
+      "step": 7955
+    },
+    {
+      "epoch": 72.99,
+      "learning_rate": 8.5125e-05,
+      "loss": 0.6068,
+      "step": 7956
+    },
+    {
+      "epoch": 73.0,
+      "learning_rate": 8.509615384615383e-05,
+      "loss": 0.6451,
+      "step": 7957
+    },
+    {
+      "epoch": 73.01,
+      "learning_rate": 8.506730769230769e-05,
+      "loss": 0.591,
+      "step": 7958
+    },
+    {
+      "epoch": 73.02,
+      "learning_rate": 8.503846153846154e-05,
+      "loss": 0.5721,
+      "step": 7959
+    },
+    {
+      "epoch": 73.03,
+      "learning_rate": 8.500961538461538e-05,
+      "loss": 0.607,
+      "step": 7960
+    },
+    {
+      "epoch": 73.04,
+      "learning_rate": 8.498076923076923e-05,
+      "loss": 0.6083,
+      "step": 7961
+    },
+    {
+      "epoch": 73.05,
+      "learning_rate": 8.495192307692307e-05,
+      "loss": 0.7291,
+      "step": 7962
+    },
+    {
+      "epoch": 73.06,
+      "learning_rate": 8.492307692307692e-05,
+      "loss": 0.6328,
+      "step": 7963
+    },
+    {
+      "epoch": 73.06,
+      "learning_rate": 8.489423076923077e-05,
+      "loss": 0.529,
+      "step": 7964
+    },
+    {
+      "epoch": 73.07,
+      "learning_rate": 8.48653846153846e-05,
+      "loss": 0.5947,
+      "step": 7965
+    },
+    {
+      "epoch": 73.08,
+      "learning_rate": 8.483653846153846e-05,
+      "loss": 0.5942,
+      "step": 7966
+    },
+    {
+      "epoch": 73.09,
+      "learning_rate": 8.48076923076923e-05,
+      "loss": 0.5506,
+      "step": 7967
+    },
+    {
+      "epoch": 73.1,
+      "learning_rate": 8.477884615384615e-05,
+      "loss": 0.831,
+      "step": 7968
+    },
+    {
+      "epoch": 73.11,
+      "learning_rate": 8.474999999999999e-05,
+      "loss": 0.5463,
+      "step": 7969
+    },
+    {
+      "epoch": 73.12,
+      "learning_rate": 8.472115384615384e-05,
+      "loss": 0.7109,
+      "step": 7970
+    },
+    {
+      "epoch": 73.13,
+      "learning_rate": 8.469230769230769e-05,
+      "loss": 0.6767,
+      "step": 7971
+    },
+    {
+      "epoch": 73.14,
+      "learning_rate": 8.466346153846153e-05,
+      "loss": 0.5301,
+      "step": 7972
+    },
+    {
+      "epoch": 73.15,
+      "learning_rate": 8.463461538461537e-05,
+      "loss": 0.5487,
+      "step": 7973
+    },
+    {
+      "epoch": 73.16,
+      "learning_rate": 8.460576923076922e-05,
+      "loss": 0.5792,
+      "step": 7974
+    },
+    {
+      "epoch": 73.17,
+      "learning_rate": 8.457692307692307e-05,
+      "loss": 0.6038,
+      "step": 7975
+    },
+    {
+      "epoch": 73.17,
+      "learning_rate": 8.454807692307692e-05,
+      "loss": 0.6072,
+      "step": 7976
+    },
+    {
+      "epoch": 73.18,
+      "learning_rate": 8.451923076923076e-05,
+      "loss": 0.6151,
+      "step": 7977
+    },
+    {
+      "epoch": 73.19,
+      "learning_rate": 8.44903846153846e-05,
+      "loss": 0.5304,
+      "step": 7978
+    },
+    {
+      "epoch": 73.2,
+      "learning_rate": 8.446153846153845e-05,
+      "loss": 0.5821,
+      "step": 7979
+    },
+    {
+      "epoch": 73.21,
+      "learning_rate": 8.44326923076923e-05,
+      "loss": 0.6304,
+      "step": 7980
+    },
+    {
+      "epoch": 73.22,
+      "learning_rate": 8.440384615384615e-05,
+      "loss": 0.5589,
+      "step": 7981
+    },
+    {
+      "epoch": 73.23,
+      "learning_rate": 8.437499999999999e-05,
+      "loss": 0.5739,
+      "step": 7982
+    },
+    {
+      "epoch": 73.24,
+      "learning_rate": 8.434615384615383e-05,
+      "loss": 0.5377,
+      "step": 7983
+    },
+    {
+      "epoch": 73.25,
+      "learning_rate": 8.431730769230768e-05,
+      "loss": 0.6056,
+      "step": 7984
+    },
+    {
+      "epoch": 73.26,
+      "learning_rate": 8.428846153846153e-05,
+      "loss": 0.5913,
+      "step": 7985
+    },
+    {
+      "epoch": 73.27,
+      "learning_rate": 8.425961538461538e-05,
+      "loss": 0.5527,
+      "step": 7986
+    },
+    {
+      "epoch": 73.28,
+      "learning_rate": 8.423076923076922e-05,
+      "loss": 0.6404,
+      "step": 7987
+    },
+    {
+      "epoch": 73.28,
+      "learning_rate": 8.420192307692306e-05,
+      "loss": 0.7003,
+      "step": 7988
+    },
+    {
+      "epoch": 73.29,
+      "learning_rate": 8.417307692307691e-05,
+      "loss": 0.5547,
+      "step": 7989
+    },
+    {
+      "epoch": 73.3,
+      "learning_rate": 8.414423076923076e-05,
+      "loss": 0.579,
+      "step": 7990
+    },
+    {
+      "epoch": 73.31,
+      "learning_rate": 8.411538461538462e-05,
+      "loss": 0.5268,
+      "step": 7991
+    },
+    {
+      "epoch": 73.32,
+      "learning_rate": 8.408653846153845e-05,
+      "loss": 0.7086,
+      "step": 7992
+    },
+    {
+      "epoch": 73.33,
+      "learning_rate": 8.405769230769229e-05,
+      "loss": 0.5756,
+      "step": 7993
+    },
+    {
+      "epoch": 73.34,
+      "learning_rate": 8.402884615384614e-05,
+      "loss": 0.5555,
+      "step": 7994
+    },
+    {
+      "epoch": 73.35,
+      "learning_rate": 8.4e-05,
+      "loss": 0.6176,
+      "step": 7995
+    },
+    {
+      "epoch": 73.36,
+      "learning_rate": 8.397115384615385e-05,
+      "loss": 0.651,
+      "step": 7996
+    },
+    {
+      "epoch": 73.37,
+      "learning_rate": 8.394230769230769e-05,
+      "loss": 0.5492,
+      "step": 7997
+    },
+    {
+      "epoch": 73.38,
+      "learning_rate": 8.391346153846152e-05,
+      "loss": 0.6326,
+      "step": 7998
+    },
+    {
+      "epoch": 73.39,
+      "learning_rate": 8.388461538461538e-05,
+      "loss": 0.6002,
+      "step": 7999
+    },
+    {
+      "epoch": 73.39,
+      "learning_rate": 8.385576923076921e-05,
+      "loss": 0.6985,
+      "step": 8000
+    },
+    {
+      "epoch": 73.39,
+      "eval_cer": 0.08628315267095459,
+      "eval_loss": 0.3626864552497864,
+      "eval_runtime": 13.9827,
+      "eval_samples_per_second": 117.789,
+      "eval_steps_per_second": 1.859,
+      "eval_wer": 0.3614543968951078,
+      "step": 8000
+    },
+    {
+      "epoch": 73.4,
+      "learning_rate": 8.382692307692308e-05,
+      "loss": 0.5993,
+      "step": 8001
+    },
+    {
+      "epoch": 73.41,
+      "learning_rate": 8.379807692307692e-05,
+      "loss": 0.6217,
+      "step": 8002
+    },
+    {
+      "epoch": 73.42,
+      "learning_rate": 8.376923076923075e-05,
+      "loss": 0.5798,
+      "step": 8003
+    },
+    {
+      "epoch": 73.43,
+      "learning_rate": 8.37403846153846e-05,
+      "loss": 0.5891,
+      "step": 8004
+    },
+    {
+      "epoch": 73.44,
+      "learning_rate": 8.371153846153844e-05,
+      "loss": 0.5388,
+      "step": 8005
+    },
+    {
+      "epoch": 73.45,
+      "learning_rate": 8.368269230769231e-05,
+      "loss": 0.4534,
+      "step": 8006
+    },
+    {
+      "epoch": 73.46,
+      "learning_rate": 8.365384615384615e-05,
+      "loss": 0.4634,
+      "step": 8007
+    },
+    {
+      "epoch": 73.47,
+      "learning_rate": 8.362499999999999e-05,
+      "loss": 0.6718,
+      "step": 8008
+    },
+    {
+      "epoch": 73.48,
+      "learning_rate": 8.359615384615384e-05,
+      "loss": 0.6383,
+      "step": 8009
+    },
+    {
+      "epoch": 73.49,
+      "learning_rate": 8.356730769230768e-05,
+      "loss": 0.5482,
+      "step": 8010
+    },
+    {
+      "epoch": 73.5,
+      "learning_rate": 8.353846153846154e-05,
+      "loss": 0.6166,
+      "step": 8011
+    },
+    {
+      "epoch": 73.5,
+      "learning_rate": 8.350961538461538e-05,
+      "loss": 0.6023,
+      "step": 8012
+    },
+    {
+      "epoch": 73.51,
+      "learning_rate": 8.348076923076922e-05,
+      "loss": 0.6168,
+      "step": 8013
+    },
+    {
+      "epoch": 73.52,
+      "learning_rate": 8.345192307692307e-05,
+      "loss": 0.5425,
+      "step": 8014
+    },
+    {
+      "epoch": 73.53,
+      "learning_rate": 8.342307692307691e-05,
+      "loss": 0.4796,
+      "step": 8015
+    },
+    {
+      "epoch": 73.54,
+      "learning_rate": 8.339423076923077e-05,
+      "loss": 0.5674,
+      "step": 8016
+    },
+    {
+      "epoch": 73.55,
+      "learning_rate": 8.336538461538461e-05,
+      "loss": 0.6328,
+      "step": 8017
+    },
+    {
+      "epoch": 73.56,
+      "learning_rate": 8.333653846153845e-05,
+      "loss": 0.5768,
+      "step": 8018
+    },
+    {
+      "epoch": 73.57,
+      "learning_rate": 8.33076923076923e-05,
+      "loss": 0.5155,
+      "step": 8019
+    },
+    {
+      "epoch": 73.58,
+      "learning_rate": 8.327884615384614e-05,
+      "loss": 0.6573,
+      "step": 8020
+    },
+    {
+      "epoch": 73.59,
+      "learning_rate": 8.325e-05,
+      "loss": 0.5311,
+      "step": 8021
+    },
+    {
+      "epoch": 73.6,
+      "learning_rate": 8.322115384615384e-05,
+      "loss": 0.6977,
+      "step": 8022
+    },
+    {
+      "epoch": 73.61,
+      "learning_rate": 8.319230769230768e-05,
+      "loss": 0.6162,
+      "step": 8023
+    },
+    {
+      "epoch": 73.61,
+      "learning_rate": 8.316346153846153e-05,
+      "loss": 0.5743,
+      "step": 8024
+    },
+    {
+      "epoch": 73.62,
+      "learning_rate": 8.313461538461537e-05,
+      "loss": 0.6968,
+      "step": 8025
+    },
+    {
+      "epoch": 73.63,
+      "learning_rate": 8.310576923076924e-05,
+      "loss": 0.5501,
+      "step": 8026
+    },
+    {
+      "epoch": 73.64,
+      "learning_rate": 8.307692307692307e-05,
+      "loss": 0.6888,
+      "step": 8027
+    },
+    {
+      "epoch": 73.65,
+      "learning_rate": 8.304807692307691e-05,
+      "loss": 0.546,
+      "step": 8028
+    },
+    {
+      "epoch": 73.66,
+      "learning_rate": 8.301923076923076e-05,
+      "loss": 0.5793,
+      "step": 8029
+    },
+    {
+      "epoch": 73.67,
+      "learning_rate": 8.29903846153846e-05,
+      "loss": 0.6092,
+      "step": 8030
+    },
+    {
+      "epoch": 73.68,
+      "learning_rate": 8.296153846153844e-05,
+      "loss": 0.7507,
+      "step": 8031
+    },
+    {
+      "epoch": 73.69,
+      "learning_rate": 8.29326923076923e-05,
+      "loss": 0.5243,
+      "step": 8032
+    },
+    {
+      "epoch": 73.7,
+      "learning_rate": 8.290384615384614e-05,
+      "loss": 0.6285,
+      "step": 8033
+    },
+    {
+      "epoch": 73.71,
+      "learning_rate": 8.2875e-05,
+      "loss": 0.6214,
+      "step": 8034
+    },
+    {
+      "epoch": 73.72,
+      "learning_rate": 8.284615384615383e-05,
+      "loss": 0.6281,
+      "step": 8035
+    },
+    {
+      "epoch": 73.72,
+      "learning_rate": 8.281730769230767e-05,
+      "loss": 0.5696,
+      "step": 8036
+    },
+    {
+      "epoch": 73.73,
+      "learning_rate": 8.278846153846154e-05,
+      "loss": 0.6099,
+      "step": 8037
+    },
+    {
+      "epoch": 73.74,
+      "learning_rate": 8.275961538461537e-05,
+      "loss": 0.5334,
+      "step": 8038
+    },
+    {
+      "epoch": 73.75,
+      "learning_rate": 8.273076923076923e-05,
+      "loss": 0.6522,
+      "step": 8039
+    },
+    {
+      "epoch": 73.76,
+      "learning_rate": 8.270192307692306e-05,
+      "loss": 0.5296,
+      "step": 8040
+    },
+    {
+      "epoch": 73.77,
+      "learning_rate": 8.26730769230769e-05,
+      "loss": 0.5958,
+      "step": 8041
+    },
+    {
+      "epoch": 73.78,
+      "learning_rate": 8.264423076923077e-05,
+      "loss": 0.6367,
+      "step": 8042
+    },
+    {
+      "epoch": 73.79,
+      "learning_rate": 8.26153846153846e-05,
+      "loss": 0.6547,
+      "step": 8043
+    },
+    {
+      "epoch": 73.8,
+      "learning_rate": 8.258653846153846e-05,
+      "loss": 0.7122,
+      "step": 8044
+    },
+    {
+      "epoch": 73.81,
+      "learning_rate": 8.25576923076923e-05,
+      "loss": 0.5041,
+      "step": 8045
+    },
+    {
+      "epoch": 73.82,
+      "learning_rate": 8.252884615384613e-05,
+      "loss": 0.6449,
+      "step": 8046
+    },
+    {
+      "epoch": 73.83,
+      "learning_rate": 8.25e-05,
+      "loss": 0.5868,
+      "step": 8047
+    },
+    {
+      "epoch": 73.83,
+      "learning_rate": 8.247115384615384e-05,
+      "loss": 0.5902,
+      "step": 8048
+    },
+    {
+      "epoch": 73.84,
+      "learning_rate": 8.244230769230769e-05,
+      "loss": 0.7134,
+      "step": 8049
+    },
+    {
+      "epoch": 73.85,
+      "learning_rate": 8.241346153846153e-05,
+      "loss": 0.447,
+      "step": 8050
+    },
+    {
+      "epoch": 73.86,
+      "learning_rate": 8.238461538461537e-05,
+      "loss": 0.5817,
+      "step": 8051
+    },
+    {
+      "epoch": 73.87,
+      "learning_rate": 8.235576923076923e-05,
+      "loss": 0.6532,
+      "step": 8052
+    },
+    {
+      "epoch": 73.88,
+      "learning_rate": 8.232692307692307e-05,
+      "loss": 0.5656,
+      "step": 8053
+    },
+    {
+      "epoch": 73.89,
+      "learning_rate": 8.229807692307692e-05,
+      "loss": 0.7104,
+      "step": 8054
+    },
+    {
+      "epoch": 73.9,
+      "learning_rate": 8.226923076923076e-05,
+      "loss": 0.6358,
+      "step": 8055
+    },
+    {
+      "epoch": 73.91,
+      "learning_rate": 8.22403846153846e-05,
+      "loss": 0.5674,
+      "step": 8056
+    },
+    {
+      "epoch": 73.92,
+      "learning_rate": 8.221153846153846e-05,
+      "loss": 0.5553,
+      "step": 8057
+    },
+    {
+      "epoch": 73.93,
+      "learning_rate": 8.21826923076923e-05,
+      "loss": 0.7678,
+      "step": 8058
+    },
+    {
+      "epoch": 73.94,
+      "learning_rate": 8.215384615384615e-05,
+      "loss": 0.5068,
+      "step": 8059
+    },
+    {
+      "epoch": 73.94,
+      "learning_rate": 8.212499999999999e-05,
+      "loss": 0.4567,
+      "step": 8060
+    },
+    {
+      "epoch": 73.95,
+      "learning_rate": 8.209615384615383e-05,
+      "loss": 0.5419,
+      "step": 8061
+    },
+    {
+      "epoch": 73.96,
+      "learning_rate": 8.20673076923077e-05,
+      "loss": 0.7008,
+      "step": 8062
+    },
+    {
+      "epoch": 73.97,
+      "learning_rate": 8.203846153846153e-05,
+      "loss": 0.7061,
+      "step": 8063
+    },
+    {
+      "epoch": 73.98,
+      "learning_rate": 8.200961538461538e-05,
+      "loss": 0.5327,
+      "step": 8064
+    },
+    {
+      "epoch": 73.99,
+      "learning_rate": 8.198076923076922e-05,
+      "loss": 0.6103,
+      "step": 8065
+    },
+    {
+      "epoch": 74.0,
+      "learning_rate": 8.195192307692306e-05,
+      "loss": 0.6168,
+      "step": 8066
+    },
+    {
+      "epoch": 74.01,
+      "learning_rate": 8.192307692307691e-05,
+      "loss": 0.6866,
+      "step": 8067
+    },
+    {
+      "epoch": 74.02,
+      "learning_rate": 8.189423076923076e-05,
+      "loss": 0.6021,
+      "step": 8068
+    },
+    {
+      "epoch": 74.03,
+      "learning_rate": 8.186538461538462e-05,
+      "loss": 0.5643,
+      "step": 8069
+    },
+    {
+      "epoch": 74.04,
+      "learning_rate": 8.183653846153845e-05,
+      "loss": 0.606,
+      "step": 8070
+    },
+    {
+      "epoch": 74.05,
+      "learning_rate": 8.180769230769229e-05,
+      "loss": 0.6039,
+      "step": 8071
+    },
+    {
+      "epoch": 74.06,
+      "learning_rate": 8.177884615384614e-05,
+      "loss": 0.5468,
+      "step": 8072
+    },
+    {
+      "epoch": 74.06,
+      "learning_rate": 8.175e-05,
+      "loss": 0.4513,
+      "step": 8073
+    },
+    {
+      "epoch": 74.07,
+      "learning_rate": 8.172115384615385e-05,
+      "loss": 0.5576,
+      "step": 8074
+    },
+    {
+      "epoch": 74.08,
+      "learning_rate": 8.169230769230768e-05,
+      "loss": 0.6175,
+      "step": 8075
+    },
+    {
+      "epoch": 74.09,
+      "learning_rate": 8.166346153846152e-05,
+      "loss": 0.6659,
+      "step": 8076
+    },
+    {
+      "epoch": 74.1,
+      "learning_rate": 8.163461538461537e-05,
+      "loss": 0.6988,
+      "step": 8077
+    },
+    {
+      "epoch": 74.11,
+      "learning_rate": 8.160576923076923e-05,
+      "loss": 0.5676,
+      "step": 8078
+    },
+    {
+      "epoch": 74.12,
+      "learning_rate": 8.157692307692308e-05,
+      "loss": 0.6624,
+      "step": 8079
+    },
+    {
+      "epoch": 74.13,
+      "learning_rate": 8.154807692307692e-05,
+      "loss": 0.5612,
+      "step": 8080
+    },
+    {
+      "epoch": 74.14,
+      "learning_rate": 8.151923076923075e-05,
+      "loss": 0.5683,
+      "step": 8081
+    },
+    {
+      "epoch": 74.15,
+      "learning_rate": 8.14903846153846e-05,
+      "loss": 0.6173,
+      "step": 8082
+    },
+    {
+      "epoch": 74.16,
+      "learning_rate": 8.146153846153846e-05,
+      "loss": 0.553,
+      "step": 8083
+    },
+    {
+      "epoch": 74.17,
+      "learning_rate": 8.143269230769231e-05,
+      "loss": 0.6413,
+      "step": 8084
+    },
+    {
+      "epoch": 74.17,
+      "learning_rate": 8.140384615384615e-05,
+      "loss": 0.4859,
+      "step": 8085
+    },
+    {
+      "epoch": 74.18,
+      "learning_rate": 8.137499999999999e-05,
+      "loss": 0.5427,
+      "step": 8086
+    },
+    {
+      "epoch": 74.19,
+      "learning_rate": 8.134615384615384e-05,
+      "loss": 0.5767,
+      "step": 8087
+    },
+    {
+      "epoch": 74.2,
+      "learning_rate": 8.131730769230769e-05,
+      "loss": 0.5704,
+      "step": 8088
+    },
+    {
+      "epoch": 74.21,
+      "learning_rate": 8.128846153846154e-05,
+      "loss": 0.5588,
+      "step": 8089
+    },
+    {
+      "epoch": 74.22,
+      "learning_rate": 8.125961538461538e-05,
+      "loss": 0.5325,
+      "step": 8090
+    },
+    {
+      "epoch": 74.23,
+      "learning_rate": 8.123076923076922e-05,
+      "loss": 0.5411,
+      "step": 8091
+    },
+    {
+      "epoch": 74.24,
+      "learning_rate": 8.120192307692307e-05,
+      "loss": 0.6027,
+      "step": 8092
+    },
+    {
+      "epoch": 74.25,
+      "learning_rate": 8.117307692307692e-05,
+      "loss": 0.5423,
+      "step": 8093
+    },
+    {
+      "epoch": 74.26,
+      "learning_rate": 8.114423076923077e-05,
+      "loss": 0.5852,
+      "step": 8094
+    },
+    {
+      "epoch": 74.27,
+      "learning_rate": 8.111538461538461e-05,
+      "loss": 0.6958,
+      "step": 8095
+    },
+    {
+      "epoch": 74.28,
+      "learning_rate": 8.108653846153845e-05,
+      "loss": 0.5116,
+      "step": 8096
+    },
+    {
+      "epoch": 74.28,
+      "learning_rate": 8.10576923076923e-05,
+      "loss": 0.6562,
+      "step": 8097
+    },
+    {
+      "epoch": 74.29,
+      "learning_rate": 8.102884615384614e-05,
+      "loss": 0.5608,
+      "step": 8098
+    },
+    {
+      "epoch": 74.3,
+      "learning_rate": 8.1e-05,
+      "loss": 0.4632,
+      "step": 8099
+    },
+    {
+      "epoch": 74.31,
+      "learning_rate": 8.097115384615384e-05,
+      "loss": 0.5825,
+      "step": 8100
+    },
+    {
+      "epoch": 74.32,
+      "learning_rate": 8.094230769230768e-05,
+      "loss": 0.6414,
+      "step": 8101
+    },
+    {
+      "epoch": 74.33,
+      "learning_rate": 8.091346153846153e-05,
+      "loss": 0.6802,
+      "step": 8102
+    },
+    {
+      "epoch": 74.34,
+      "learning_rate": 8.088461538461537e-05,
+      "loss": 0.5506,
+      "step": 8103
+    },
+    {
+      "epoch": 74.35,
+      "learning_rate": 8.085576923076924e-05,
+      "loss": 0.5388,
+      "step": 8104
+    },
+    {
+      "epoch": 74.36,
+      "learning_rate": 8.082692307692307e-05,
+      "loss": 0.6261,
+      "step": 8105
+    },
+    {
+      "epoch": 74.37,
+      "learning_rate": 8.079807692307691e-05,
+      "loss": 0.5701,
+      "step": 8106
+    },
+    {
+      "epoch": 74.38,
+      "learning_rate": 8.076923076923076e-05,
+      "loss": 0.5683,
+      "step": 8107
+    },
+    {
+      "epoch": 74.39,
+      "learning_rate": 8.07403846153846e-05,
+      "loss": 0.6444,
+      "step": 8108
+    },
+    {
+      "epoch": 74.39,
+      "learning_rate": 8.071153846153847e-05,
+      "loss": 0.6519,
+      "step": 8109
+    },
+    {
+      "epoch": 74.4,
+      "learning_rate": 8.06826923076923e-05,
+      "loss": 0.6425,
+      "step": 8110
+    },
+    {
+      "epoch": 74.41,
+      "learning_rate": 8.065384615384614e-05,
+      "loss": 0.5679,
+      "step": 8111
+    },
+    {
+      "epoch": 74.42,
+      "learning_rate": 8.0625e-05,
+      "loss": 0.5136,
+      "step": 8112
+    },
+    {
+      "epoch": 74.43,
+      "learning_rate": 8.059615384615383e-05,
+      "loss": 0.5819,
+      "step": 8113
+    },
+    {
+      "epoch": 74.44,
+      "learning_rate": 8.05673076923077e-05,
+      "loss": 0.6741,
+      "step": 8114
+    },
+    {
+      "epoch": 74.45,
+      "learning_rate": 8.053846153846154e-05,
+      "loss": 0.5911,
+      "step": 8115
+    },
+    {
+      "epoch": 74.46,
+      "learning_rate": 8.050961538461537e-05,
+      "loss": 0.6064,
+      "step": 8116
+    },
+    {
+      "epoch": 74.47,
+      "learning_rate": 8.048076923076923e-05,
+      "loss": 0.6045,
+      "step": 8117
+    },
+    {
+      "epoch": 74.48,
+      "learning_rate": 8.045192307692306e-05,
+      "loss": 0.597,
+      "step": 8118
+    },
+    {
+      "epoch": 74.49,
+      "learning_rate": 8.042307692307693e-05,
+      "loss": 0.7154,
+      "step": 8119
+    },
+    {
+      "epoch": 74.5,
+      "learning_rate": 8.039423076923077e-05,
+      "loss": 0.6077,
+      "step": 8120
+    },
+    {
+      "epoch": 74.5,
+      "learning_rate": 8.03653846153846e-05,
+      "loss": 0.5319,
+      "step": 8121
+    },
+    {
+      "epoch": 74.51,
+      "learning_rate": 8.033653846153846e-05,
+      "loss": 0.5783,
+      "step": 8122
+    },
+    {
+      "epoch": 74.52,
+      "learning_rate": 8.03076923076923e-05,
+      "loss": 0.6486,
+      "step": 8123
+    },
+    {
+      "epoch": 74.53,
+      "learning_rate": 8.027884615384616e-05,
+      "loss": 0.5271,
+      "step": 8124
+    },
+    {
+      "epoch": 74.54,
+      "learning_rate": 8.025e-05,
+      "loss": 0.6255,
+      "step": 8125
+    },
+    {
+      "epoch": 74.55,
+      "learning_rate": 8.022115384615384e-05,
+      "loss": 0.6215,
+      "step": 8126
+    },
+    {
+      "epoch": 74.56,
+      "learning_rate": 8.019230769230769e-05,
+      "loss": 0.5488,
+      "step": 8127
+    },
+    {
+      "epoch": 74.57,
+      "learning_rate": 8.016346153846153e-05,
+      "loss": 0.6073,
+      "step": 8128
+    },
+    {
+      "epoch": 74.58,
+      "learning_rate": 8.013461538461537e-05,
+      "loss": 0.5758,
+      "step": 8129
+    },
+    {
+      "epoch": 74.59,
+      "learning_rate": 8.010576923076923e-05,
+      "loss": 0.5272,
+      "step": 8130
+    },
+    {
+      "epoch": 74.6,
+      "learning_rate": 8.007692307692307e-05,
+      "loss": 0.7197,
+      "step": 8131
+    },
+    {
+      "epoch": 74.61,
+      "learning_rate": 8.004807692307692e-05,
+      "loss": 0.5626,
+      "step": 8132
+    },
+    {
+      "epoch": 74.61,
+      "learning_rate": 8.001923076923076e-05,
+      "loss": 0.7008,
+      "step": 8133
+    },
+    {
+      "epoch": 74.62,
+      "learning_rate": 7.99903846153846e-05,
+      "loss": 0.6918,
+      "step": 8134
+    },
+    {
+      "epoch": 74.63,
+      "learning_rate": 7.996153846153846e-05,
+      "loss": 0.5762,
+      "step": 8135
+    },
+    {
+      "epoch": 74.64,
+      "learning_rate": 7.99326923076923e-05,
+      "loss": 0.5709,
+      "step": 8136
+    },
+    {
+      "epoch": 74.65,
+      "learning_rate": 7.990384615384615e-05,
+      "loss": 0.5157,
+      "step": 8137
+    },
+    {
+      "epoch": 74.66,
+      "learning_rate": 7.987499999999999e-05,
+      "loss": 0.4981,
+      "step": 8138
+    },
+    {
+      "epoch": 74.67,
+      "learning_rate": 7.984615384615383e-05,
+      "loss": 0.5386,
+      "step": 8139
+    },
+    {
+      "epoch": 74.68,
+      "learning_rate": 7.98173076923077e-05,
+      "loss": 0.6022,
+      "step": 8140
+    },
+    {
+      "epoch": 74.69,
+      "learning_rate": 7.978846153846153e-05,
+      "loss": 0.5278,
+      "step": 8141
+    },
+    {
+      "epoch": 74.7,
+      "learning_rate": 7.975961538461538e-05,
+      "loss": 0.7037,
+      "step": 8142
+    },
+    {
+      "epoch": 74.71,
+      "learning_rate": 7.973076923076922e-05,
+      "loss": 0.5495,
+      "step": 8143
+    },
+    {
+      "epoch": 74.72,
+      "learning_rate": 7.970192307692306e-05,
+      "loss": 0.5769,
+      "step": 8144
+    },
+    {
+      "epoch": 74.72,
+      "learning_rate": 7.967307692307692e-05,
+      "loss": 0.5697,
+      "step": 8145
+    },
+    {
+      "epoch": 74.73,
+      "learning_rate": 7.964423076923076e-05,
+      "loss": 0.615,
+      "step": 8146
+    },
+    {
+      "epoch": 74.74,
+      "learning_rate": 7.961538461538461e-05,
+      "loss": 0.5065,
+      "step": 8147
+    },
+    {
+      "epoch": 74.75,
+      "learning_rate": 7.958653846153845e-05,
+      "loss": 0.6474,
+      "step": 8148
+    },
+    {
+      "epoch": 74.76,
+      "learning_rate": 7.955769230769229e-05,
+      "loss": 0.4611,
+      "step": 8149
+    },
+    {
+      "epoch": 74.77,
+      "learning_rate": 7.952884615384616e-05,
+      "loss": 0.5528,
+      "step": 8150
+    },
+    {
+      "epoch": 74.78,
+      "learning_rate": 7.95e-05,
+      "loss": 0.4706,
+      "step": 8151
+    },
+    {
+      "epoch": 74.79,
+      "learning_rate": 7.947115384615385e-05,
+      "loss": 0.5627,
+      "step": 8152
+    },
+    {
+      "epoch": 74.8,
+      "learning_rate": 7.944230769230768e-05,
+      "loss": 0.6638,
+      "step": 8153
+    },
+    {
+      "epoch": 74.81,
+      "learning_rate": 7.941346153846152e-05,
+      "loss": 0.5701,
+      "step": 8154
+    },
+    {
+      "epoch": 74.82,
+      "learning_rate": 7.938461538461539e-05,
+      "loss": 0.6687,
+      "step": 8155
+    },
+    {
+      "epoch": 74.83,
+      "learning_rate": 7.935576923076923e-05,
+      "loss": 0.74,
+      "step": 8156
+    },
+    {
+      "epoch": 74.83,
+      "learning_rate": 7.932692307692308e-05,
+      "loss": 0.5104,
+      "step": 8157
+    },
+    {
+      "epoch": 74.84,
+      "learning_rate": 7.929807692307692e-05,
+      "loss": 0.4866,
+      "step": 8158
+    },
+    {
+      "epoch": 74.85,
+      "learning_rate": 7.926923076923075e-05,
+      "loss": 0.5204,
+      "step": 8159
+    },
+    {
+      "epoch": 74.86,
+      "learning_rate": 7.924038461538462e-05,
+      "loss": 0.567,
+      "step": 8160
+    },
+    {
+      "epoch": 74.87,
+      "learning_rate": 7.921153846153846e-05,
+      "loss": 0.5682,
+      "step": 8161
+    },
+    {
+      "epoch": 74.88,
+      "learning_rate": 7.918269230769231e-05,
+      "loss": 0.6203,
+      "step": 8162
+    },
+    {
+      "epoch": 74.89,
+      "learning_rate": 7.915384615384615e-05,
+      "loss": 0.6574,
+      "step": 8163
+    },
+    {
+      "epoch": 74.9,
+      "learning_rate": 7.912499999999999e-05,
+      "loss": 0.6438,
+      "step": 8164
+    },
+    {
+      "epoch": 74.91,
+      "learning_rate": 7.909615384615384e-05,
+      "loss": 0.6459,
+      "step": 8165
+    },
+    {
+      "epoch": 74.92,
+      "learning_rate": 7.906730769230769e-05,
+      "loss": 0.5183,
+      "step": 8166
+    },
+    {
+      "epoch": 74.93,
+      "learning_rate": 7.903846153846154e-05,
+      "loss": 0.6075,
+      "step": 8167
+    },
+    {
+      "epoch": 74.94,
+      "learning_rate": 7.900961538461538e-05,
+      "loss": 0.4616,
+      "step": 8168
+    },
+    {
+      "epoch": 74.94,
+      "learning_rate": 7.898076923076922e-05,
+      "loss": 0.6455,
+      "step": 8169
+    },
+    {
+      "epoch": 74.95,
+      "learning_rate": 7.895192307692307e-05,
+      "loss": 0.5658,
+      "step": 8170
+    },
+    {
+      "epoch": 74.96,
+      "learning_rate": 7.892307692307692e-05,
+      "loss": 0.6403,
+      "step": 8171
+    },
+    {
+      "epoch": 74.97,
+      "learning_rate": 7.889423076923077e-05,
+      "loss": 0.7047,
+      "step": 8172
+    },
+    {
+      "epoch": 74.98,
+      "learning_rate": 7.886538461538461e-05,
+      "loss": 0.5373,
+      "step": 8173
+    },
+    {
+      "epoch": 74.99,
+      "learning_rate": 7.883653846153845e-05,
+      "loss": 0.5875,
+      "step": 8174
+    },
+    {
+      "epoch": 75.0,
+      "learning_rate": 7.88076923076923e-05,
+      "loss": 0.6466,
+      "step": 8175
+    },
+    {
+      "epoch": 75.01,
+      "learning_rate": 7.877884615384615e-05,
+      "loss": 0.6658,
+      "step": 8176
+    },
+    {
+      "epoch": 75.02,
+      "learning_rate": 7.874999999999999e-05,
+      "loss": 0.5073,
+      "step": 8177
+    },
+    {
+      "epoch": 75.03,
+      "learning_rate": 7.872115384615384e-05,
+      "loss": 0.6003,
+      "step": 8178
+    },
+    {
+      "epoch": 75.04,
+      "learning_rate": 7.869230769230768e-05,
+      "loss": 0.5283,
+      "step": 8179
+    },
+    {
+      "epoch": 75.05,
+      "learning_rate": 7.866346153846153e-05,
+      "loss": 0.7012,
+      "step": 8180
+    },
+    {
+      "epoch": 75.06,
+      "learning_rate": 7.863461538461538e-05,
+      "loss": 0.6263,
+      "step": 8181
+    },
+    {
+      "epoch": 75.06,
+      "learning_rate": 7.860576923076922e-05,
+      "loss": 0.6425,
+      "step": 8182
+    },
+    {
+      "epoch": 75.07,
+      "learning_rate": 7.857692307692307e-05,
+      "loss": 0.583,
+      "step": 8183
+    },
+    {
+      "epoch": 75.08,
+      "learning_rate": 7.854807692307691e-05,
+      "loss": 0.6006,
+      "step": 8184
+    },
+    {
+      "epoch": 75.09,
+      "learning_rate": 7.851923076923076e-05,
+      "loss": 0.6861,
+      "step": 8185
+    },
+    {
+      "epoch": 75.1,
+      "learning_rate": 7.849038461538461e-05,
+      "loss": 0.6073,
+      "step": 8186
+    },
+    {
+      "epoch": 75.11,
+      "learning_rate": 7.846153846153845e-05,
+      "loss": 0.4677,
+      "step": 8187
+    },
+    {
+      "epoch": 75.12,
+      "learning_rate": 7.84326923076923e-05,
+      "loss": 0.5329,
+      "step": 8188
+    },
+    {
+      "epoch": 75.13,
+      "learning_rate": 7.840384615384614e-05,
+      "loss": 0.4928,
+      "step": 8189
+    },
+    {
+      "epoch": 75.14,
+      "learning_rate": 7.8375e-05,
+      "loss": 0.6303,
+      "step": 8190
+    },
+    {
+      "epoch": 75.15,
+      "learning_rate": 7.834615384615385e-05,
+      "loss": 0.5209,
+      "step": 8191
+    },
+    {
+      "epoch": 75.16,
+      "learning_rate": 7.831730769230768e-05,
+      "loss": 0.48,
+      "step": 8192
+    },
+    {
+      "epoch": 75.17,
+      "learning_rate": 7.828846153846154e-05,
+      "loss": 0.6366,
+      "step": 8193
+    },
+    {
+      "epoch": 75.17,
+      "learning_rate": 7.825961538461537e-05,
+      "loss": 0.5915,
+      "step": 8194
+    },
+    {
+      "epoch": 75.18,
+      "learning_rate": 7.823076923076923e-05,
+      "loss": 0.5515,
+      "step": 8195
+    },
+    {
+      "epoch": 75.19,
+      "learning_rate": 7.820192307692306e-05,
+      "loss": 0.5758,
+      "step": 8196
+    },
+    {
+      "epoch": 75.2,
+      "learning_rate": 7.817307692307692e-05,
+      "loss": 0.5754,
+      "step": 8197
+    },
+    {
+      "epoch": 75.21,
+      "learning_rate": 7.814423076923077e-05,
+      "loss": 0.4875,
+      "step": 8198
+    },
+    {
+      "epoch": 75.22,
+      "learning_rate": 7.81153846153846e-05,
+      "loss": 0.7196,
+      "step": 8199
+    },
+    {
+      "epoch": 75.23,
+      "learning_rate": 7.808653846153846e-05,
+      "loss": 0.566,
+      "step": 8200
+    },
+    {
+      "epoch": 75.24,
+      "learning_rate": 7.80576923076923e-05,
+      "loss": 0.5777,
+      "step": 8201
+    },
+    {
+      "epoch": 75.25,
+      "learning_rate": 7.802884615384615e-05,
+      "loss": 0.576,
+      "step": 8202
+    },
+    {
+      "epoch": 75.26,
+      "learning_rate": 7.8e-05,
+      "loss": 0.5596,
+      "step": 8203
+    },
+    {
+      "epoch": 75.27,
+      "learning_rate": 7.797115384615384e-05,
+      "loss": 0.6266,
+      "step": 8204
+    },
+    {
+      "epoch": 75.28,
+      "learning_rate": 7.794230769230769e-05,
+      "loss": 0.5611,
+      "step": 8205
+    },
+    {
+      "epoch": 75.28,
+      "learning_rate": 7.791346153846153e-05,
+      "loss": 0.6828,
+      "step": 8206
+    },
+    {
+      "epoch": 75.29,
+      "learning_rate": 7.788461538461538e-05,
+      "loss": 0.7055,
+      "step": 8207
+    },
+    {
+      "epoch": 75.3,
+      "learning_rate": 7.785576923076923e-05,
+      "loss": 0.6077,
+      "step": 8208
+    },
+    {
+      "epoch": 75.31,
+      "learning_rate": 7.782692307692307e-05,
+      "loss": 0.55,
+      "step": 8209
+    },
+    {
+      "epoch": 75.32,
+      "learning_rate": 7.77980769230769e-05,
+      "loss": 0.6227,
+      "step": 8210
+    },
+    {
+      "epoch": 75.33,
+      "learning_rate": 7.776923076923076e-05,
+      "loss": 0.6776,
+      "step": 8211
+    },
+    {
+      "epoch": 75.34,
+      "learning_rate": 7.774038461538461e-05,
+      "loss": 0.5763,
+      "step": 8212
+    },
+    {
+      "epoch": 75.35,
+      "learning_rate": 7.771153846153846e-05,
+      "loss": 0.5891,
+      "step": 8213
+    },
+    {
+      "epoch": 75.36,
+      "learning_rate": 7.76826923076923e-05,
+      "loss": 0.6095,
+      "step": 8214
+    },
+    {
+      "epoch": 75.37,
+      "learning_rate": 7.765384615384614e-05,
+      "loss": 0.6834,
+      "step": 8215
+    },
+    {
+      "epoch": 75.38,
+      "learning_rate": 7.762499999999999e-05,
+      "loss": 0.5545,
+      "step": 8216
+    },
+    {
+      "epoch": 75.39,
+      "learning_rate": 7.759615384615384e-05,
+      "loss": 0.5795,
+      "step": 8217
+    },
+    {
+      "epoch": 75.39,
+      "learning_rate": 7.756730769230769e-05,
+      "loss": 0.6604,
+      "step": 8218
+    },
+    {
+      "epoch": 75.4,
+      "learning_rate": 7.753846153846153e-05,
+      "loss": 0.4832,
+      "step": 8219
+    },
+    {
+      "epoch": 75.41,
+      "learning_rate": 7.750961538461537e-05,
+      "loss": 0.6685,
+      "step": 8220
+    },
+    {
+      "epoch": 75.42,
+      "learning_rate": 7.748076923076922e-05,
+      "loss": 0.5781,
+      "step": 8221
+    },
+    {
+      "epoch": 75.43,
+      "learning_rate": 7.745192307692307e-05,
+      "loss": 0.6375,
+      "step": 8222
+    },
+    {
+      "epoch": 75.44,
+      "learning_rate": 7.742307692307692e-05,
+      "loss": 0.5883,
+      "step": 8223
+    },
+    {
+      "epoch": 75.45,
+      "learning_rate": 7.739423076923076e-05,
+      "loss": 0.6645,
+      "step": 8224
+    },
+    {
+      "epoch": 75.46,
+      "learning_rate": 7.73653846153846e-05,
+      "loss": 0.5342,
+      "step": 8225
+    },
+    {
+      "epoch": 75.47,
+      "learning_rate": 7.733653846153845e-05,
+      "loss": 0.5812,
+      "step": 8226
+    },
+    {
+      "epoch": 75.48,
+      "learning_rate": 7.730769230769229e-05,
+      "loss": 0.6234,
+      "step": 8227
+    },
+    {
+      "epoch": 75.49,
+      "learning_rate": 7.727884615384616e-05,
+      "loss": 0.6248,
+      "step": 8228
+    },
+    {
+      "epoch": 75.5,
+      "learning_rate": 7.725e-05,
+      "loss": 0.5864,
+      "step": 8229
+    },
+    {
+      "epoch": 75.5,
+      "learning_rate": 7.722115384615383e-05,
+      "loss": 0.7562,
+      "step": 8230
+    },
+    {
+      "epoch": 75.51,
+      "learning_rate": 7.719230769230768e-05,
+      "loss": 0.6219,
+      "step": 8231
+    },
+    {
+      "epoch": 75.52,
+      "learning_rate": 7.716346153846152e-05,
+      "loss": 0.6213,
+      "step": 8232
+    },
+    {
+      "epoch": 75.53,
+      "learning_rate": 7.713461538461539e-05,
+      "loss": 0.478,
+      "step": 8233
+    },
+    {
+      "epoch": 75.54,
+      "learning_rate": 7.710576923076923e-05,
+      "loss": 0.6316,
+      "step": 8234
+    },
+    {
+      "epoch": 75.55,
+      "learning_rate": 7.707692307692306e-05,
+      "loss": 0.6429,
+      "step": 8235
+    },
+    {
+      "epoch": 75.56,
+      "learning_rate": 7.704807692307692e-05,
+      "loss": 0.5734,
+      "step": 8236
+    },
+    {
+      "epoch": 75.57,
+      "learning_rate": 7.701923076923075e-05,
+      "loss": 0.5266,
+      "step": 8237
+    },
+    {
+      "epoch": 75.58,
+      "learning_rate": 7.699038461538462e-05,
+      "loss": 0.7105,
+      "step": 8238
+    },
+    {
+      "epoch": 75.59,
+      "learning_rate": 7.696153846153846e-05,
+      "loss": 0.598,
+      "step": 8239
+    },
+    {
+      "epoch": 75.6,
+      "learning_rate": 7.69326923076923e-05,
+      "loss": 0.7017,
+      "step": 8240
+    },
+    {
+      "epoch": 75.61,
+      "learning_rate": 7.690384615384615e-05,
+      "loss": 0.563,
+      "step": 8241
+    },
+    {
+      "epoch": 75.61,
+      "learning_rate": 7.687499999999998e-05,
+      "loss": 0.548,
+      "step": 8242
+    },
+    {
+      "epoch": 75.62,
+      "learning_rate": 7.684615384615385e-05,
+      "loss": 0.5211,
+      "step": 8243
+    },
+    {
+      "epoch": 75.63,
+      "learning_rate": 7.681730769230769e-05,
+      "loss": 0.6784,
+      "step": 8244
+    },
+    {
+      "epoch": 75.64,
+      "learning_rate": 7.678846153846153e-05,
+      "loss": 0.5885,
+      "step": 8245
+    },
+    {
+      "epoch": 75.65,
+      "learning_rate": 7.675961538461538e-05,
+      "loss": 0.5955,
+      "step": 8246
+    },
+    {
+      "epoch": 75.66,
+      "learning_rate": 7.673076923076922e-05,
+      "loss": 0.5062,
+      "step": 8247
+    },
+    {
+      "epoch": 75.67,
+      "learning_rate": 7.670192307692308e-05,
+      "loss": 0.608,
+      "step": 8248
+    },
+    {
+      "epoch": 75.68,
+      "learning_rate": 7.667307692307692e-05,
+      "loss": 0.6057,
+      "step": 8249
+    },
+    {
+      "epoch": 75.69,
+      "learning_rate": 7.664423076923076e-05,
+      "loss": 0.762,
+      "step": 8250
+    },
+    {
+      "epoch": 75.7,
+      "learning_rate": 7.661538461538461e-05,
+      "loss": 0.5808,
+      "step": 8251
+    },
+    {
+      "epoch": 75.71,
+      "learning_rate": 7.658653846153845e-05,
+      "loss": 0.5448,
+      "step": 8252
+    },
+    {
+      "epoch": 75.72,
+      "learning_rate": 7.655769230769231e-05,
+      "loss": 0.5392,
+      "step": 8253
+    },
+    {
+      "epoch": 75.72,
+      "learning_rate": 7.652884615384615e-05,
+      "loss": 0.6848,
+      "step": 8254
+    },
+    {
+      "epoch": 75.73,
+      "learning_rate": 7.649999999999999e-05,
+      "loss": 0.5296,
+      "step": 8255
+    },
+    {
+      "epoch": 75.74,
+      "learning_rate": 7.647115384615384e-05,
+      "loss": 0.5494,
+      "step": 8256
+    },
+    {
+      "epoch": 75.75,
+      "learning_rate": 7.644230769230768e-05,
+      "loss": 0.4718,
+      "step": 8257
+    },
+    {
+      "epoch": 75.76,
+      "learning_rate": 7.641346153846154e-05,
+      "loss": 0.6348,
+      "step": 8258
+    },
+    {
+      "epoch": 75.77,
+      "learning_rate": 7.638461538461538e-05,
+      "loss": 0.6475,
+      "step": 8259
+    },
+    {
+      "epoch": 75.78,
+      "learning_rate": 7.635576923076922e-05,
+      "loss": 0.6705,
+      "step": 8260
+    },
+    {
+      "epoch": 75.79,
+      "learning_rate": 7.632692307692307e-05,
+      "loss": 0.504,
+      "step": 8261
+    },
+    {
+      "epoch": 75.8,
+      "learning_rate": 7.629807692307691e-05,
+      "loss": 0.5695,
+      "step": 8262
+    },
+    {
+      "epoch": 75.81,
+      "learning_rate": 7.626923076923075e-05,
+      "loss": 0.6012,
+      "step": 8263
+    },
+    {
+      "epoch": 75.82,
+      "learning_rate": 7.624038461538461e-05,
+      "loss": 0.5739,
+      "step": 8264
+    },
+    {
+      "epoch": 75.83,
+      "learning_rate": 7.621153846153845e-05,
+      "loss": 0.5162,
+      "step": 8265
+    },
+    {
+      "epoch": 75.83,
+      "learning_rate": 7.61826923076923e-05,
+      "loss": 0.6176,
+      "step": 8266
+    },
+    {
+      "epoch": 75.84,
+      "learning_rate": 7.615384615384614e-05,
+      "loss": 0.5983,
+      "step": 8267
+    },
+    {
+      "epoch": 75.85,
+      "learning_rate": 7.612499999999998e-05,
+      "loss": 0.5299,
+      "step": 8268
+    },
+    {
+      "epoch": 75.86,
+      "learning_rate": 7.609615384615385e-05,
+      "loss": 0.492,
+      "step": 8269
+    },
+    {
+      "epoch": 75.87,
+      "learning_rate": 7.606730769230768e-05,
+      "loss": 0.6121,
+      "step": 8270
+    },
+    {
+      "epoch": 75.88,
+      "learning_rate": 7.603846153846154e-05,
+      "loss": 0.5383,
+      "step": 8271
+    },
+    {
+      "epoch": 75.89,
+      "learning_rate": 7.600961538461537e-05,
+      "loss": 0.543,
+      "step": 8272
+    },
+    {
+      "epoch": 75.9,
+      "learning_rate": 7.598076923076921e-05,
+      "loss": 0.4762,
+      "step": 8273
+    },
+    {
+      "epoch": 75.91,
+      "learning_rate": 7.595192307692308e-05,
+      "loss": 0.5862,
+      "step": 8274
+    },
+    {
+      "epoch": 75.92,
+      "learning_rate": 7.592307692307691e-05,
+      "loss": 0.4527,
+      "step": 8275
+    },
+    {
+      "epoch": 75.93,
+      "learning_rate": 7.589423076923077e-05,
+      "loss": 0.5257,
+      "step": 8276
+    },
+    {
+      "epoch": 75.94,
+      "learning_rate": 7.58653846153846e-05,
+      "loss": 0.6925,
+      "step": 8277
+    },
+    {
+      "epoch": 75.94,
+      "learning_rate": 7.583653846153844e-05,
+      "loss": 0.6231,
+      "step": 8278
+    },
+    {
+      "epoch": 75.95,
+      "learning_rate": 7.580769230769231e-05,
+      "loss": 0.5645,
+      "step": 8279
+    },
+    {
+      "epoch": 75.96,
+      "learning_rate": 7.577884615384615e-05,
+      "loss": 0.708,
+      "step": 8280
+    },
+    {
+      "epoch": 75.97,
+      "learning_rate": 7.575e-05,
+      "loss": 0.7105,
+      "step": 8281
+    },
+    {
+      "epoch": 75.98,
+      "learning_rate": 7.572115384615384e-05,
+      "loss": 0.5468,
+      "step": 8282
+    },
+    {
+      "epoch": 75.99,
+      "learning_rate": 7.569230769230767e-05,
+      "loss": 0.64,
+      "step": 8283
+    },
+    {
+      "epoch": 76.0,
+      "learning_rate": 7.566346153846154e-05,
+      "loss": 0.5442,
+      "step": 8284
+    },
+    {
+      "epoch": 76.01,
+      "learning_rate": 7.563461538461538e-05,
+      "loss": 0.5813,
+      "step": 8285
+    },
+    {
+      "epoch": 76.02,
+      "learning_rate": 7.560576923076923e-05,
+      "loss": 0.5189,
+      "step": 8286
+    },
+    {
+      "epoch": 76.03,
+      "learning_rate": 7.557692307692307e-05,
+      "loss": 0.5786,
+      "step": 8287
+    },
+    {
+      "epoch": 76.04,
+      "learning_rate": 7.55480769230769e-05,
+      "loss": 0.4668,
+      "step": 8288
+    },
+    {
+      "epoch": 76.05,
+      "learning_rate": 7.551923076923077e-05,
+      "loss": 0.6088,
+      "step": 8289
+    },
+    {
+      "epoch": 76.06,
+      "learning_rate": 7.549038461538461e-05,
+      "loss": 0.6329,
+      "step": 8290
+    },
+    {
+      "epoch": 76.06,
+      "learning_rate": 7.546153846153846e-05,
+      "loss": 0.5899,
+      "step": 8291
+    },
+    {
+      "epoch": 76.07,
+      "learning_rate": 7.54326923076923e-05,
+      "loss": 0.4897,
+      "step": 8292
+    },
+    {
+      "epoch": 76.08,
+      "learning_rate": 7.540384615384614e-05,
+      "loss": 0.5991,
+      "step": 8293
+    },
+    {
+      "epoch": 76.09,
+      "learning_rate": 7.537499999999999e-05,
+      "loss": 0.538,
+      "step": 8294
+    },
+    {
+      "epoch": 76.1,
+      "learning_rate": 7.534615384615384e-05,
+      "loss": 0.5499,
+      "step": 8295
+    },
+    {
+      "epoch": 76.11,
+      "learning_rate": 7.531730769230769e-05,
+      "loss": 0.594,
+      "step": 8296
+    },
+    {
+      "epoch": 76.12,
+      "learning_rate": 7.528846153846153e-05,
+      "loss": 0.6474,
+      "step": 8297
+    },
+    {
+      "epoch": 76.13,
+      "learning_rate": 7.525961538461537e-05,
+      "loss": 0.6496,
+      "step": 8298
+    },
+    {
+      "epoch": 76.14,
+      "learning_rate": 7.523076923076922e-05,
+      "loss": 0.516,
+      "step": 8299
+    },
+    {
+      "epoch": 76.15,
+      "learning_rate": 7.520192307692307e-05,
+      "loss": 0.5277,
+      "step": 8300
+    },
+    {
+      "epoch": 76.16,
+      "learning_rate": 7.517307692307692e-05,
+      "loss": 0.6343,
+      "step": 8301
+    },
+    {
+      "epoch": 76.17,
+      "learning_rate": 7.514423076923076e-05,
+      "loss": 0.5271,
+      "step": 8302
+    },
+    {
+      "epoch": 76.17,
+      "learning_rate": 7.51153846153846e-05,
+      "loss": 0.6111,
+      "step": 8303
+    },
+    {
+      "epoch": 76.18,
+      "learning_rate": 7.508653846153845e-05,
+      "loss": 0.5434,
+      "step": 8304
+    },
+    {
+      "epoch": 76.19,
+      "learning_rate": 7.50576923076923e-05,
+      "loss": 0.5626,
+      "step": 8305
+    },
+    {
+      "epoch": 76.2,
+      "learning_rate": 7.502884615384616e-05,
+      "loss": 0.6835,
+      "step": 8306
+    },
+    {
+      "epoch": 76.21,
+      "learning_rate": 7.5e-05,
+      "loss": 0.5644,
+      "step": 8307
+    },
+    {
+      "epoch": 76.22,
+      "learning_rate": 7.497115384615383e-05,
+      "loss": 0.5306,
+      "step": 8308
+    },
+    {
+      "epoch": 76.23,
+      "learning_rate": 7.494230769230768e-05,
+      "loss": 0.3743,
+      "step": 8309
+    },
+    {
+      "epoch": 76.24,
+      "learning_rate": 7.491346153846154e-05,
+      "loss": 0.6739,
+      "step": 8310
+    },
+    {
+      "epoch": 76.25,
+      "learning_rate": 7.488461538461539e-05,
+      "loss": 0.6091,
+      "step": 8311
+    },
+    {
+      "epoch": 76.26,
+      "learning_rate": 7.485576923076922e-05,
+      "loss": 0.5479,
+      "step": 8312
+    },
+    {
+      "epoch": 76.27,
+      "learning_rate": 7.482692307692306e-05,
+      "loss": 0.591,
+      "step": 8313
+    },
+    {
+      "epoch": 76.28,
+      "learning_rate": 7.479807692307691e-05,
+      "loss": 0.5671,
+      "step": 8314
+    },
+    {
+      "epoch": 76.28,
+      "learning_rate": 7.476923076923077e-05,
+      "loss": 0.5435,
+      "step": 8315
+    },
+    {
+      "epoch": 76.29,
+      "learning_rate": 7.47403846153846e-05,
+      "loss": 0.5013,
+      "step": 8316
+    },
+    {
+      "epoch": 76.3,
+      "learning_rate": 7.471153846153846e-05,
+      "loss": 0.5447,
+      "step": 8317
+    },
+    {
+      "epoch": 76.31,
+      "learning_rate": 7.46826923076923e-05,
+      "loss": 0.5271,
+      "step": 8318
+    },
+    {
+      "epoch": 76.32,
+      "learning_rate": 7.465384615384615e-05,
+      "loss": 0.5633,
+      "step": 8319
+    },
+    {
+      "epoch": 76.33,
+      "learning_rate": 7.4625e-05,
+      "loss": 0.5908,
+      "step": 8320
+    },
+    {
+      "epoch": 76.34,
+      "learning_rate": 7.459615384615384e-05,
+      "loss": 0.4056,
+      "step": 8321
+    },
+    {
+      "epoch": 76.35,
+      "learning_rate": 7.456730769230769e-05,
+      "loss": 0.5429,
+      "step": 8322
+    },
+    {
+      "epoch": 76.36,
+      "learning_rate": 7.453846153846153e-05,
+      "loss": 0.6507,
+      "step": 8323
+    },
+    {
+      "epoch": 76.37,
+      "learning_rate": 7.450961538461538e-05,
+      "loss": 0.6136,
+      "step": 8324
+    },
+    {
+      "epoch": 76.38,
+      "learning_rate": 7.448076923076923e-05,
+      "loss": 0.5413,
+      "step": 8325
+    },
+    {
+      "epoch": 76.39,
+      "learning_rate": 7.445192307692307e-05,
+      "loss": 0.5255,
+      "step": 8326
+    },
+    {
+      "epoch": 76.39,
+      "learning_rate": 7.442307692307692e-05,
+      "loss": 0.6019,
+      "step": 8327
+    },
+    {
+      "epoch": 76.4,
+      "learning_rate": 7.439423076923076e-05,
+      "loss": 0.5336,
+      "step": 8328
+    },
+    {
+      "epoch": 76.41,
+      "learning_rate": 7.436538461538461e-05,
+      "loss": 0.6533,
+      "step": 8329
+    },
+    {
+      "epoch": 76.42,
+      "learning_rate": 7.433653846153846e-05,
+      "loss": 0.5527,
+      "step": 8330
+    },
+    {
+      "epoch": 76.43,
+      "learning_rate": 7.43076923076923e-05,
+      "loss": 0.5991,
+      "step": 8331
+    },
+    {
+      "epoch": 76.44,
+      "learning_rate": 7.427884615384615e-05,
+      "loss": 0.6333,
+      "step": 8332
+    },
+    {
+      "epoch": 76.45,
+      "learning_rate": 7.424999999999999e-05,
+      "loss": 0.5926,
+      "step": 8333
+    },
+    {
+      "epoch": 76.46,
+      "learning_rate": 7.422115384615384e-05,
+      "loss": 0.6007,
+      "step": 8334
+    },
+    {
+      "epoch": 76.47,
+      "learning_rate": 7.419230769230769e-05,
+      "loss": 0.5763,
+      "step": 8335
+    },
+    {
+      "epoch": 76.48,
+      "learning_rate": 7.416346153846153e-05,
+      "loss": 0.5503,
+      "step": 8336
+    },
+    {
+      "epoch": 76.49,
+      "learning_rate": 7.413461538461538e-05,
+      "loss": 0.5802,
+      "step": 8337
+    },
+    {
+      "epoch": 76.5,
+      "learning_rate": 7.410576923076922e-05,
+      "loss": 0.5764,
+      "step": 8338
+    },
+    {
+      "epoch": 76.5,
+      "learning_rate": 7.410576923076922e-05,
+      "loss": 0.4488,
+      "step": 8339
+    },
+    {
+      "epoch": 76.51,
+      "learning_rate": 7.407692307692307e-05,
+      "loss": 0.5357,
+      "step": 8340
+    },
+    {
+      "epoch": 76.52,
+      "learning_rate": 7.404807692307692e-05,
+      "loss": 0.5506,
+      "step": 8341
+    },
+    {
+      "epoch": 76.53,
+      "learning_rate": 7.401923076923076e-05,
+      "loss": 0.5296,
+      "step": 8342
+    },
+    {
+      "epoch": 76.54,
+      "learning_rate": 7.399038461538461e-05,
+      "loss": 0.6456,
+      "step": 8343
+    },
+    {
+      "epoch": 76.55,
+      "learning_rate": 7.396153846153845e-05,
+      "loss": 0.6531,
+      "step": 8344
+    },
+    {
+      "epoch": 76.56,
+      "learning_rate": 7.39326923076923e-05,
+      "loss": 0.5977,
+      "step": 8345
+    },
+    {
+      "epoch": 76.57,
+      "learning_rate": 7.390384615384616e-05,
+      "loss": 0.6236,
+      "step": 8346
+    },
+    {
+      "epoch": 76.58,
+      "learning_rate": 7.387499999999999e-05,
+      "loss": 0.4664,
+      "step": 8347
+    },
+    {
+      "epoch": 76.59,
+      "learning_rate": 7.384615384615384e-05,
+      "loss": 0.5338,
+      "step": 8348
+    },
+    {
+      "epoch": 76.6,
+      "learning_rate": 7.381730769230768e-05,
+      "loss": 0.5704,
+      "step": 8349
+    },
+    {
+      "epoch": 76.61,
+      "learning_rate": 7.378846153846153e-05,
+      "loss": 0.5996,
+      "step": 8350
+    },
+    {
+      "epoch": 76.61,
+      "learning_rate": 7.375961538461539e-05,
+      "loss": 0.672,
+      "step": 8351
+    },
+    {
+      "epoch": 76.62,
+      "learning_rate": 7.373076923076922e-05,
+      "loss": 0.6176,
+      "step": 8352
+    },
+    {
+      "epoch": 76.63,
+      "learning_rate": 7.370192307692306e-05,
+      "loss": 0.5626,
+      "step": 8353
+    },
+    {
+      "epoch": 76.64,
+      "learning_rate": 7.367307692307691e-05,
+      "loss": 0.5427,
+      "step": 8354
+    },
+    {
+      "epoch": 76.65,
+      "learning_rate": 7.364423076923077e-05,
+      "loss": 0.7366,
+      "step": 8355
+    },
+    {
+      "epoch": 76.66,
+      "learning_rate": 7.361538461538462e-05,
+      "loss": 0.5613,
+      "step": 8356
+    },
+    {
+      "epoch": 76.67,
+      "learning_rate": 7.358653846153846e-05,
+      "loss": 0.5895,
+      "step": 8357
+    },
+    {
+      "epoch": 76.68,
+      "learning_rate": 7.35576923076923e-05,
+      "loss": 0.5278,
+      "step": 8358
+    },
+    {
+      "epoch": 76.69,
+      "learning_rate": 7.352884615384615e-05,
+      "loss": 0.4545,
+      "step": 8359
+    },
+    {
+      "epoch": 76.7,
+      "learning_rate": 7.35e-05,
+      "loss": 0.5263,
+      "step": 8360
+    },
+    {
+      "epoch": 76.71,
+      "learning_rate": 7.347115384615385e-05,
+      "loss": 0.5561,
+      "step": 8361
+    },
+    {
+      "epoch": 76.72,
+      "learning_rate": 7.344230769230769e-05,
+      "loss": 0.521,
+      "step": 8362
+    },
+    {
+      "epoch": 76.72,
+      "learning_rate": 7.341346153846153e-05,
+      "loss": 0.5937,
+      "step": 8363
+    },
+    {
+      "epoch": 76.73,
+      "learning_rate": 7.338461538461538e-05,
+      "loss": 0.5691,
+      "step": 8364
+    },
+    {
+      "epoch": 76.74,
+      "learning_rate": 7.335576923076923e-05,
+      "loss": 0.5437,
+      "step": 8365
+    },
+    {
+      "epoch": 76.75,
+      "learning_rate": 7.332692307692307e-05,
+      "loss": 0.5727,
+      "step": 8366
+    },
+    {
+      "epoch": 76.76,
+      "learning_rate": 7.329807692307692e-05,
+      "loss": 0.6312,
+      "step": 8367
+    },
+    {
+      "epoch": 76.77,
+      "learning_rate": 7.326923076923076e-05,
+      "loss": 0.6836,
+      "step": 8368
+    },
+    {
+      "epoch": 76.78,
+      "learning_rate": 7.324038461538461e-05,
+      "loss": 0.6305,
+      "step": 8369
+    },
+    {
+      "epoch": 76.79,
+      "learning_rate": 7.321153846153846e-05,
+      "loss": 0.5189,
+      "step": 8370
+    },
+    {
+      "epoch": 76.8,
+      "learning_rate": 7.31826923076923e-05,
+      "loss": 0.5863,
+      "step": 8371
+    },
+    {
+      "epoch": 76.81,
+      "learning_rate": 7.315384615384615e-05,
+      "loss": 0.5862,
+      "step": 8372
+    },
+    {
+      "epoch": 76.82,
+      "learning_rate": 7.312499999999999e-05,
+      "loss": 0.4871,
+      "step": 8373
+    },
+    {
+      "epoch": 76.83,
+      "learning_rate": 7.309615384615384e-05,
+      "loss": 0.5525,
+      "step": 8374
+    },
+    {
+      "epoch": 76.83,
+      "learning_rate": 7.306730769230769e-05,
+      "loss": 0.6395,
+      "step": 8375
+    },
+    {
+      "epoch": 76.84,
+      "learning_rate": 7.303846153846153e-05,
+      "loss": 0.5992,
+      "step": 8376
+    },
+    {
+      "epoch": 76.85,
+      "learning_rate": 7.300961538461538e-05,
+      "loss": 0.6343,
+      "step": 8377
+    },
+    {
+      "epoch": 76.86,
+      "learning_rate": 7.298076923076922e-05,
+      "loss": 0.6367,
+      "step": 8378
+    },
+    {
+      "epoch": 76.87,
+      "learning_rate": 7.295192307692307e-05,
+      "loss": 0.611,
+      "step": 8379
+    },
+    {
+      "epoch": 76.88,
+      "learning_rate": 7.292307692307692e-05,
+      "loss": 0.5226,
+      "step": 8380
+    },
+    {
+      "epoch": 76.89,
+      "learning_rate": 7.289423076923076e-05,
+      "loss": 0.5957,
+      "step": 8381
+    },
+    {
+      "epoch": 76.9,
+      "learning_rate": 7.286538461538461e-05,
+      "loss": 0.6261,
+      "step": 8382
+    },
+    {
+      "epoch": 76.91,
+      "learning_rate": 7.283653846153845e-05,
+      "loss": 0.6294,
+      "step": 8383
+    },
+    {
+      "epoch": 76.92,
+      "learning_rate": 7.28076923076923e-05,
+      "loss": 0.5287,
+      "step": 8384
+    },
+    {
+      "epoch": 76.93,
+      "learning_rate": 7.277884615384615e-05,
+      "loss": 0.6918,
+      "step": 8385
+    },
+    {
+      "epoch": 76.94,
+      "learning_rate": 7.274999999999999e-05,
+      "loss": 0.6349,
+      "step": 8386
+    },
+    {
+      "epoch": 76.94,
+      "learning_rate": 7.272115384615384e-05,
+      "loss": 0.6635,
+      "step": 8387
+    },
+    {
+      "epoch": 76.95,
+      "learning_rate": 7.269230769230768e-05,
+      "loss": 0.7184,
+      "step": 8388
+    },
+    {
+      "epoch": 76.96,
+      "learning_rate": 7.266346153846153e-05,
+      "loss": 0.6831,
+      "step": 8389
+    },
+    {
+      "epoch": 76.97,
+      "learning_rate": 7.263461538461539e-05,
+      "loss": 0.5455,
+      "step": 8390
+    },
+    {
+      "epoch": 76.98,
+      "learning_rate": 7.260576923076922e-05,
+      "loss": 0.5715,
+      "step": 8391
+    },
+    {
+      "epoch": 76.99,
+      "learning_rate": 7.257692307692308e-05,
+      "loss": 0.6104,
+      "step": 8392
+    },
+    {
+      "epoch": 77.0,
+      "learning_rate": 7.254807692307691e-05,
+      "loss": 0.6013,
+      "step": 8393
+    },
+    {
+      "epoch": 77.01,
+      "learning_rate": 7.251923076923077e-05,
+      "loss": 0.5885,
+      "step": 8394
+    },
+    {
+      "epoch": 77.02,
+      "learning_rate": 7.249038461538462e-05,
+      "loss": 0.6472,
+      "step": 8395
+    },
+    {
+      "epoch": 77.03,
+      "learning_rate": 7.246153846153846e-05,
+      "loss": 0.6325,
+      "step": 8396
+    },
+    {
+      "epoch": 77.04,
+      "learning_rate": 7.243269230769231e-05,
+      "loss": 0.5831,
+      "step": 8397
+    },
+    {
+      "epoch": 77.05,
+      "learning_rate": 7.240384615384615e-05,
+      "loss": 0.5981,
+      "step": 8398
+    },
+    {
+      "epoch": 77.06,
+      "learning_rate": 7.2375e-05,
+      "loss": 0.6205,
+      "step": 8399
+    },
+    {
+      "epoch": 77.06,
+      "learning_rate": 7.234615384615385e-05,
+      "loss": 0.5984,
+      "step": 8400
+    },
+    {
+      "epoch": 77.07,
+      "learning_rate": 7.231730769230769e-05,
+      "loss": 0.5139,
+      "step": 8401
+    },
+    {
+      "epoch": 77.08,
+      "learning_rate": 7.228846153846153e-05,
+      "loss": 0.5781,
+      "step": 8402
+    },
+    {
+      "epoch": 77.09,
+      "learning_rate": 7.225961538461538e-05,
+      "loss": 0.6307,
+      "step": 8403
+    },
+    {
+      "epoch": 77.1,
+      "learning_rate": 7.223076923076923e-05,
+      "loss": 0.5765,
+      "step": 8404
+    },
+    {
+      "epoch": 77.11,
+      "learning_rate": 7.220192307692308e-05,
+      "loss": 0.6032,
+      "step": 8405
+    },
+    {
+      "epoch": 77.12,
+      "learning_rate": 7.217307692307692e-05,
+      "loss": 0.5379,
+      "step": 8406
+    },
+    {
+      "epoch": 77.13,
+      "learning_rate": 7.214423076923076e-05,
+      "loss": 0.6194,
+      "step": 8407
+    },
+    {
+      "epoch": 77.14,
+      "learning_rate": 7.211538461538461e-05,
+      "loss": 0.5653,
+      "step": 8408
+    },
+    {
+      "epoch": 77.15,
+      "learning_rate": 7.208653846153846e-05,
+      "loss": 0.5758,
+      "step": 8409
+    },
+    {
+      "epoch": 77.16,
+      "learning_rate": 7.20576923076923e-05,
+      "loss": 0.6182,
+      "step": 8410
+    },
+    {
+      "epoch": 77.17,
+      "learning_rate": 7.202884615384615e-05,
+      "loss": 0.5834,
+      "step": 8411
+    },
+    {
+      "epoch": 77.17,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 0.6221,
+      "step": 8412
+    },
+    {
+      "epoch": 77.18,
+      "learning_rate": 7.197115384615384e-05,
+      "loss": 0.5069,
+      "step": 8413
+    },
+    {
+      "epoch": 77.19,
+      "learning_rate": 7.194230769230769e-05,
+      "loss": 0.6728,
+      "step": 8414
+    },
+    {
+      "epoch": 77.2,
+      "learning_rate": 7.191346153846153e-05,
+      "loss": 0.6329,
+      "step": 8415
+    },
+    {
+      "epoch": 77.21,
+      "learning_rate": 7.188461538461538e-05,
+      "loss": 0.5928,
+      "step": 8416
+    },
+    {
+      "epoch": 77.22,
+      "learning_rate": 7.185576923076922e-05,
+      "loss": 0.5744,
+      "step": 8417
+    },
+    {
+      "epoch": 77.23,
+      "learning_rate": 7.182692307692307e-05,
+      "loss": 0.4723,
+      "step": 8418
+    },
+    {
+      "epoch": 77.24,
+      "learning_rate": 7.179807692307692e-05,
+      "loss": 0.5281,
+      "step": 8419
+    },
+    {
+      "epoch": 77.25,
+      "learning_rate": 7.176923076923076e-05,
+      "loss": 0.7136,
+      "step": 8420
+    },
+    {
+      "epoch": 77.26,
+      "learning_rate": 7.174038461538461e-05,
+      "loss": 0.5006,
+      "step": 8421
+    },
+    {
+      "epoch": 77.27,
+      "learning_rate": 7.171153846153845e-05,
+      "loss": 0.592,
+      "step": 8422
+    },
+    {
+      "epoch": 77.28,
+      "learning_rate": 7.16826923076923e-05,
+      "loss": 0.6278,
+      "step": 8423
+    },
+    {
+      "epoch": 77.28,
+      "learning_rate": 7.165384615384615e-05,
+      "loss": 0.5058,
+      "step": 8424
+    },
+    {
+      "epoch": 77.29,
+      "learning_rate": 7.162499999999999e-05,
+      "loss": 0.6422,
+      "step": 8425
+    },
+    {
+      "epoch": 77.3,
+      "learning_rate": 7.159615384615384e-05,
+      "loss": 0.5219,
+      "step": 8426
+    },
+    {
+      "epoch": 77.31,
+      "learning_rate": 7.156730769230768e-05,
+      "loss": 0.5216,
+      "step": 8427
+    },
+    {
+      "epoch": 77.32,
+      "learning_rate": 7.153846153846153e-05,
+      "loss": 0.47,
+      "step": 8428
+    },
+    {
+      "epoch": 77.33,
+      "learning_rate": 7.150961538461537e-05,
+      "loss": 0.4444,
+      "step": 8429
+    },
+    {
+      "epoch": 77.34,
+      "learning_rate": 7.148076923076922e-05,
+      "loss": 0.6064,
+      "step": 8430
+    },
+    {
+      "epoch": 77.35,
+      "learning_rate": 7.145192307692308e-05,
+      "loss": 0.5608,
+      "step": 8431
+    },
+    {
+      "epoch": 77.36,
+      "learning_rate": 7.142307692307691e-05,
+      "loss": 0.6174,
+      "step": 8432
+    },
+    {
+      "epoch": 77.37,
+      "learning_rate": 7.139423076923077e-05,
+      "loss": 0.6017,
+      "step": 8433
+    },
+    {
+      "epoch": 77.38,
+      "learning_rate": 7.13653846153846e-05,
+      "loss": 0.5839,
+      "step": 8434
+    },
+    {
+      "epoch": 77.39,
+      "learning_rate": 7.133653846153846e-05,
+      "loss": 0.5803,
+      "step": 8435
+    },
+    {
+      "epoch": 77.39,
+      "learning_rate": 7.130769230769231e-05,
+      "loss": 0.602,
+      "step": 8436
+    },
+    {
+      "epoch": 77.4,
+      "learning_rate": 7.127884615384615e-05,
+      "loss": 0.6327,
+      "step": 8437
+    },
+    {
+      "epoch": 77.41,
+      "learning_rate": 7.125e-05,
+      "loss": 0.4755,
+      "step": 8438
+    },
+    {
+      "epoch": 77.42,
+      "learning_rate": 7.122115384615384e-05,
+      "loss": 0.5103,
+      "step": 8439
+    },
+    {
+      "epoch": 77.43,
+      "learning_rate": 7.119230769230769e-05,
+      "loss": 0.52,
+      "step": 8440
+    },
+    {
+      "epoch": 77.44,
+      "learning_rate": 7.116346153846154e-05,
+      "loss": 0.5505,
+      "step": 8441
+    },
+    {
+      "epoch": 77.45,
+      "learning_rate": 7.113461538461538e-05,
+      "loss": 0.5726,
+      "step": 8442
+    },
+    {
+      "epoch": 77.46,
+      "learning_rate": 7.110576923076922e-05,
+      "loss": 0.6093,
+      "step": 8443
+    },
+    {
+      "epoch": 77.47,
+      "learning_rate": 7.107692307692307e-05,
+      "loss": 0.6055,
+      "step": 8444
+    },
+    {
+      "epoch": 77.48,
+      "learning_rate": 7.104807692307692e-05,
+      "loss": 0.5604,
+      "step": 8445
+    },
+    {
+      "epoch": 77.49,
+      "learning_rate": 7.101923076923077e-05,
+      "loss": 0.5713,
+      "step": 8446
+    },
+    {
+      "epoch": 77.5,
+      "learning_rate": 7.099038461538461e-05,
+      "loss": 0.4701,
+      "step": 8447
+    },
+    {
+      "epoch": 77.5,
+      "learning_rate": 7.096153846153845e-05,
+      "loss": 0.586,
+      "step": 8448
+    },
+    {
+      "epoch": 77.51,
+      "learning_rate": 7.09326923076923e-05,
+      "loss": 0.6574,
+      "step": 8449
+    },
+    {
+      "epoch": 77.52,
+      "learning_rate": 7.090384615384615e-05,
+      "loss": 0.5879,
+      "step": 8450
+    },
+    {
+      "epoch": 77.53,
+      "learning_rate": 7.087499999999999e-05,
+      "loss": 0.463,
+      "step": 8451
+    },
+    {
+      "epoch": 77.54,
+      "learning_rate": 7.084615384615384e-05,
+      "loss": 0.6814,
+      "step": 8452
+    },
+    {
+      "epoch": 77.55,
+      "learning_rate": 7.081730769230768e-05,
+      "loss": 0.5353,
+      "step": 8453
+    },
+    {
+      "epoch": 77.56,
+      "learning_rate": 7.078846153846153e-05,
+      "loss": 0.4992,
+      "step": 8454
+    },
+    {
+      "epoch": 77.57,
+      "learning_rate": 7.075961538461538e-05,
+      "loss": 0.599,
+      "step": 8455
+    },
+    {
+      "epoch": 77.58,
+      "learning_rate": 7.073076923076922e-05,
+      "loss": 0.5617,
+      "step": 8456
+    },
+    {
+      "epoch": 77.59,
+      "learning_rate": 7.070192307692307e-05,
+      "loss": 0.582,
+      "step": 8457
+    },
+    {
+      "epoch": 77.6,
+      "learning_rate": 7.067307692307691e-05,
+      "loss": 0.6223,
+      "step": 8458
+    },
+    {
+      "epoch": 77.61,
+      "learning_rate": 7.064423076923076e-05,
+      "loss": 0.595,
+      "step": 8459
+    },
+    {
+      "epoch": 77.61,
+      "learning_rate": 7.061538461538461e-05,
+      "loss": 0.6303,
+      "step": 8460
+    },
+    {
+      "epoch": 77.62,
+      "learning_rate": 7.058653846153845e-05,
+      "loss": 0.6432,
+      "step": 8461
+    },
+    {
+      "epoch": 77.63,
+      "learning_rate": 7.05576923076923e-05,
+      "loss": 0.6465,
+      "step": 8462
+    },
+    {
+      "epoch": 77.64,
+      "learning_rate": 7.052884615384614e-05,
+      "loss": 0.6423,
+      "step": 8463
+    },
+    {
+      "epoch": 77.65,
+      "learning_rate": 7.049999999999999e-05,
+      "loss": 0.5489,
+      "step": 8464
+    },
+    {
+      "epoch": 77.66,
+      "learning_rate": 7.047115384615384e-05,
+      "loss": 0.671,
+      "step": 8465
+    },
+    {
+      "epoch": 77.67,
+      "learning_rate": 7.044230769230768e-05,
+      "loss": 0.6117,
+      "step": 8466
+    },
+    {
+      "epoch": 77.68,
+      "learning_rate": 7.041346153846153e-05,
+      "loss": 0.5092,
+      "step": 8467
+    },
+    {
+      "epoch": 77.69,
+      "learning_rate": 7.038461538461537e-05,
+      "loss": 0.5014,
+      "step": 8468
+    },
+    {
+      "epoch": 77.7,
+      "learning_rate": 7.035576923076922e-05,
+      "loss": 0.6494,
+      "step": 8469
+    },
+    {
+      "epoch": 77.71,
+      "learning_rate": 7.032692307692308e-05,
+      "loss": 0.6122,
+      "step": 8470
+    },
+    {
+      "epoch": 77.72,
+      "learning_rate": 7.029807692307691e-05,
+      "loss": 0.5832,
+      "step": 8471
+    },
+    {
+      "epoch": 77.72,
+      "learning_rate": 7.026923076923077e-05,
+      "loss": 0.5377,
+      "step": 8472
+    },
+    {
+      "epoch": 77.73,
+      "learning_rate": 7.02403846153846e-05,
+      "loss": 0.6976,
+      "step": 8473
+    },
+    {
+      "epoch": 77.74,
+      "learning_rate": 7.021153846153846e-05,
+      "loss": 0.5587,
+      "step": 8474
+    },
+    {
+      "epoch": 77.75,
+      "learning_rate": 7.018269230769231e-05,
+      "loss": 0.5486,
+      "step": 8475
+    },
+    {
+      "epoch": 77.76,
+      "learning_rate": 7.015384615384615e-05,
+      "loss": 0.5595,
+      "step": 8476
+    },
+    {
+      "epoch": 77.77,
+      "learning_rate": 7.0125e-05,
+      "loss": 0.517,
+      "step": 8477
+    },
+    {
+      "epoch": 77.78,
+      "learning_rate": 7.009615384615384e-05,
+      "loss": 0.6149,
+      "step": 8478
+    },
+    {
+      "epoch": 77.79,
+      "learning_rate": 7.006730769230769e-05,
+      "loss": 0.6406,
+      "step": 8479
+    },
+    {
+      "epoch": 77.8,
+      "learning_rate": 7.003846153846154e-05,
+      "loss": 0.6234,
+      "step": 8480
+    },
+    {
+      "epoch": 77.81,
+      "learning_rate": 7.000961538461538e-05,
+      "loss": 0.541,
+      "step": 8481
+    },
+    {
+      "epoch": 77.82,
+      "learning_rate": 6.998076923076921e-05,
+      "loss": 0.5122,
+      "step": 8482
+    },
+    {
+      "epoch": 77.83,
+      "learning_rate": 6.995192307692307e-05,
+      "loss": 0.5537,
+      "step": 8483
+    },
+    {
+      "epoch": 77.83,
+      "learning_rate": 6.992307692307692e-05,
+      "loss": 0.5814,
+      "step": 8484
+    },
+    {
+      "epoch": 77.84,
+      "learning_rate": 6.989423076923077e-05,
+      "loss": 0.6482,
+      "step": 8485
+    },
+    {
+      "epoch": 77.85,
+      "learning_rate": 6.986538461538461e-05,
+      "loss": 0.5651,
+      "step": 8486
+    },
+    {
+      "epoch": 77.86,
+      "learning_rate": 6.983653846153845e-05,
+      "loss": 0.6307,
+      "step": 8487
+    },
+    {
+      "epoch": 77.87,
+      "learning_rate": 6.98076923076923e-05,
+      "loss": 0.6596,
+      "step": 8488
+    },
+    {
+      "epoch": 77.88,
+      "learning_rate": 6.977884615384615e-05,
+      "loss": 0.6503,
+      "step": 8489
+    },
+    {
+      "epoch": 77.89,
+      "learning_rate": 6.975e-05,
+      "loss": 0.5342,
+      "step": 8490
+    },
+    {
+      "epoch": 77.9,
+      "learning_rate": 6.972115384615384e-05,
+      "loss": 0.5671,
+      "step": 8491
+    },
+    {
+      "epoch": 77.91,
+      "learning_rate": 6.969230769230768e-05,
+      "loss": 0.6739,
+      "step": 8492
+    },
+    {
+      "epoch": 77.92,
+      "learning_rate": 6.966346153846153e-05,
+      "loss": 0.4363,
+      "step": 8493
+    },
+    {
+      "epoch": 77.93,
+      "learning_rate": 6.963461538461538e-05,
+      "loss": 0.5403,
+      "step": 8494
+    },
+    {
+      "epoch": 77.94,
+      "learning_rate": 6.960576923076922e-05,
+      "loss": 0.5557,
+      "step": 8495
+    },
+    {
+      "epoch": 77.94,
+      "learning_rate": 6.957692307692307e-05,
+      "loss": 0.6018,
+      "step": 8496
+    },
+    {
+      "epoch": 77.95,
+      "learning_rate": 6.954807692307691e-05,
+      "loss": 0.6484,
+      "step": 8497
+    },
+    {
+      "epoch": 77.96,
+      "learning_rate": 6.951923076923076e-05,
+      "loss": 0.4887,
+      "step": 8498
+    },
+    {
+      "epoch": 77.97,
+      "learning_rate": 6.949038461538461e-05,
+      "loss": 0.6319,
+      "step": 8499
+    },
+    {
+      "epoch": 77.98,
+      "learning_rate": 6.946153846153845e-05,
+      "loss": 0.5889,
+      "step": 8500
+    },
+    {
+      "epoch": 77.98,
+      "eval_cer": 0.08583337081515464,
+      "eval_loss": 0.37054380774497986,
+      "eval_runtime": 14.0437,
+      "eval_samples_per_second": 117.276,
+      "eval_steps_per_second": 1.851,
+      "eval_wer": 0.354815647022776,
+      "step": 8500
+    },
+    {
+      "epoch": 77.99,
+      "learning_rate": 6.94326923076923e-05,
+      "loss": 0.4487,
+      "step": 8501
+    },
+    {
+      "epoch": 78.0,
+      "learning_rate": 6.940384615384614e-05,
+      "loss": 0.3932,
+      "step": 8502
+    },
+    {
+      "epoch": 78.01,
+      "learning_rate": 6.937499999999999e-05,
+      "loss": 0.5939,
+      "step": 8503
+    },
+    {
+      "epoch": 78.02,
+      "learning_rate": 6.934615384615384e-05,
+      "loss": 0.6632,
+      "step": 8504
+    },
+    {
+      "epoch": 78.03,
+      "learning_rate": 6.931730769230768e-05,
+      "loss": 0.5598,
+      "step": 8505
+    },
+    {
+      "epoch": 78.04,
+      "learning_rate": 6.928846153846153e-05,
+      "loss": 0.5073,
+      "step": 8506
+    },
+    {
+      "epoch": 78.05,
+      "learning_rate": 6.925961538461537e-05,
+      "loss": 0.5105,
+      "step": 8507
+    },
+    {
+      "epoch": 78.06,
+      "learning_rate": 6.923076923076922e-05,
+      "loss": 0.4008,
+      "step": 8508
+    },
+    {
+      "epoch": 78.06,
+      "learning_rate": 6.920192307692308e-05,
+      "loss": 0.675,
+      "step": 8509
+    },
+    {
+      "epoch": 78.07,
+      "learning_rate": 6.917307692307691e-05,
+      "loss": 0.4865,
+      "step": 8510
+    },
+    {
+      "epoch": 78.08,
+      "learning_rate": 6.914423076923077e-05,
+      "loss": 0.5279,
+      "step": 8511
+    },
+    {
+      "epoch": 78.09,
+      "learning_rate": 6.91153846153846e-05,
+      "loss": 0.5049,
+      "step": 8512
+    },
+    {
+      "epoch": 78.1,
+      "learning_rate": 6.908653846153846e-05,
+      "loss": 0.5586,
+      "step": 8513
+    },
+    {
+      "epoch": 78.11,
+      "learning_rate": 6.905769230769231e-05,
+      "loss": 0.595,
+      "step": 8514
+    },
+    {
+      "epoch": 78.12,
+      "learning_rate": 6.902884615384614e-05,
+      "loss": 0.6184,
+      "step": 8515
+    },
+    {
+      "epoch": 78.13,
+      "learning_rate": 6.9e-05,
+      "loss": 0.509,
+      "step": 8516
+    },
+    {
+      "epoch": 78.14,
+      "learning_rate": 6.897115384615383e-05,
+      "loss": 0.6811,
+      "step": 8517
+    },
+    {
+      "epoch": 78.15,
+      "learning_rate": 6.894230769230769e-05,
+      "loss": 0.6539,
+      "step": 8518
+    },
+    {
+      "epoch": 78.16,
+      "learning_rate": 6.891346153846154e-05,
+      "loss": 0.6555,
+      "step": 8519
+    },
+    {
+      "epoch": 78.17,
+      "learning_rate": 6.888461538461538e-05,
+      "loss": 0.6032,
+      "step": 8520
+    },
+    {
+      "epoch": 78.17,
+      "learning_rate": 6.885576923076923e-05,
+      "loss": 0.5425,
+      "step": 8521
+    },
+    {
+      "epoch": 78.18,
+      "learning_rate": 6.882692307692307e-05,
+      "loss": 0.6088,
+      "step": 8522
+    },
+    {
+      "epoch": 78.19,
+      "learning_rate": 6.879807692307692e-05,
+      "loss": 0.5176,
+      "step": 8523
+    },
+    {
+      "epoch": 78.2,
+      "learning_rate": 6.876923076923077e-05,
+      "loss": 0.6198,
+      "step": 8524
+    },
+    {
+      "epoch": 78.21,
+      "learning_rate": 6.874038461538461e-05,
+      "loss": 0.5699,
+      "step": 8525
+    },
+    {
+      "epoch": 78.22,
+      "learning_rate": 6.871153846153846e-05,
+      "loss": 0.5556,
+      "step": 8526
+    },
+    {
+      "epoch": 78.23,
+      "learning_rate": 6.86826923076923e-05,
+      "loss": 0.4946,
+      "step": 8527
+    },
+    {
+      "epoch": 78.24,
+      "learning_rate": 6.865384615384615e-05,
+      "loss": 0.6119,
+      "step": 8528
+    },
+    {
+      "epoch": 78.25,
+      "learning_rate": 6.8625e-05,
+      "loss": 0.5349,
+      "step": 8529
+    },
+    {
+      "epoch": 78.26,
+      "learning_rate": 6.859615384615384e-05,
+      "loss": 0.5931,
+      "step": 8530
+    },
+    {
+      "epoch": 78.27,
+      "learning_rate": 6.856730769230768e-05,
+      "loss": 0.5871,
+      "step": 8531
+    },
+    {
+      "epoch": 78.28,
+      "learning_rate": 6.853846153846153e-05,
+      "loss": 0.6301,
+      "step": 8532
+    },
+    {
+      "epoch": 78.28,
+      "learning_rate": 6.850961538461538e-05,
+      "loss": 0.619,
+      "step": 8533
+    },
+    {
+      "epoch": 78.29,
+      "learning_rate": 6.848076923076923e-05,
+      "loss": 0.4722,
+      "step": 8534
+    },
+    {
+      "epoch": 78.3,
+      "learning_rate": 6.845192307692307e-05,
+      "loss": 0.586,
+      "step": 8535
+    },
+    {
+      "epoch": 78.31,
+      "learning_rate": 6.842307692307691e-05,
+      "loss": 0.5707,
+      "step": 8536
+    },
+    {
+      "epoch": 78.32,
+      "learning_rate": 6.839423076923076e-05,
+      "loss": 0.5066,
+      "step": 8537
+    },
+    {
+      "epoch": 78.33,
+      "learning_rate": 6.836538461538461e-05,
+      "loss": 0.5796,
+      "step": 8538
+    },
+    {
+      "epoch": 78.34,
+      "learning_rate": 6.833653846153846e-05,
+      "loss": 0.5557,
+      "step": 8539
+    },
+    {
+      "epoch": 78.35,
+      "learning_rate": 6.83076923076923e-05,
+      "loss": 0.6605,
+      "step": 8540
+    },
+    {
+      "epoch": 78.36,
+      "learning_rate": 6.827884615384614e-05,
+      "loss": 0.6038,
+      "step": 8541
+    },
+    {
+      "epoch": 78.37,
+      "learning_rate": 6.824999999999999e-05,
+      "loss": 0.622,
+      "step": 8542
+    },
+    {
+      "epoch": 78.38,
+      "learning_rate": 6.822115384615384e-05,
+      "loss": 0.4884,
+      "step": 8543
+    },
+    {
+      "epoch": 78.39,
+      "learning_rate": 6.819230769230768e-05,
+      "loss": 0.594,
+      "step": 8544
+    },
+    {
+      "epoch": 78.39,
+      "learning_rate": 6.816346153846153e-05,
+      "loss": 0.4968,
+      "step": 8545
+    },
+    {
+      "epoch": 78.4,
+      "learning_rate": 6.813461538461537e-05,
+      "loss": 0.547,
+      "step": 8546
+    },
+    {
+      "epoch": 78.41,
+      "learning_rate": 6.810576923076922e-05,
+      "loss": 0.6497,
+      "step": 8547
+    },
+    {
+      "epoch": 78.42,
+      "learning_rate": 6.807692307692308e-05,
+      "loss": 0.5494,
+      "step": 8548
+    },
+    {
+      "epoch": 78.43,
+      "learning_rate": 6.804807692307691e-05,
+      "loss": 0.6532,
+      "step": 8549
+    },
+    {
+      "epoch": 78.44,
+      "learning_rate": 6.801923076923076e-05,
+      "loss": 0.6243,
+      "step": 8550
+    },
+    {
+      "epoch": 78.45,
+      "learning_rate": 6.79903846153846e-05,
+      "loss": 0.5271,
+      "step": 8551
+    },
+    {
+      "epoch": 78.46,
+      "learning_rate": 6.796153846153845e-05,
+      "loss": 0.5217,
+      "step": 8552
+    },
+    {
+      "epoch": 78.47,
+      "learning_rate": 6.79326923076923e-05,
+      "loss": 0.5111,
+      "step": 8553
+    },
+    {
+      "epoch": 78.48,
+      "learning_rate": 6.790384615384614e-05,
+      "loss": 0.6869,
+      "step": 8554
+    },
+    {
+      "epoch": 78.49,
+      "learning_rate": 6.7875e-05,
+      "loss": 0.5152,
+      "step": 8555
+    },
+    {
+      "epoch": 78.5,
+      "learning_rate": 6.784615384615383e-05,
+      "loss": 0.5968,
+      "step": 8556
+    },
+    {
+      "epoch": 78.5,
+      "learning_rate": 6.781730769230769e-05,
+      "loss": 0.6099,
+      "step": 8557
+    },
+    {
+      "epoch": 78.51,
+      "learning_rate": 6.778846153846154e-05,
+      "loss": 0.6848,
+      "step": 8558
+    },
+    {
+      "epoch": 78.52,
+      "learning_rate": 6.775961538461538e-05,
+      "loss": 0.6234,
+      "step": 8559
+    },
+    {
+      "epoch": 78.53,
+      "learning_rate": 6.773076923076923e-05,
+      "loss": 0.6586,
+      "step": 8560
+    },
+    {
+      "epoch": 78.54,
+      "learning_rate": 6.770192307692307e-05,
+      "loss": 0.5647,
+      "step": 8561
+    },
+    {
+      "epoch": 78.55,
+      "learning_rate": 6.767307692307692e-05,
+      "loss": 0.5956,
+      "step": 8562
+    },
+    {
+      "epoch": 78.56,
+      "learning_rate": 6.764423076923077e-05,
+      "loss": 0.6215,
+      "step": 8563
+    },
+    {
+      "epoch": 78.57,
+      "learning_rate": 6.761538461538461e-05,
+      "loss": 0.4834,
+      "step": 8564
+    },
+    {
+      "epoch": 78.58,
+      "learning_rate": 6.758653846153846e-05,
+      "loss": 0.5676,
+      "step": 8565
+    },
+    {
+      "epoch": 78.59,
+      "learning_rate": 6.75576923076923e-05,
+      "loss": 0.6582,
+      "step": 8566
+    },
+    {
+      "epoch": 78.6,
+      "learning_rate": 6.752884615384615e-05,
+      "loss": 0.5338,
+      "step": 8567
+    },
+    {
+      "epoch": 78.61,
+      "learning_rate": 6.75e-05,
+      "loss": 0.5297,
+      "step": 8568
+    },
+    {
+      "epoch": 78.61,
+      "learning_rate": 6.747115384615384e-05,
+      "loss": 0.5721,
+      "step": 8569
+    },
+    {
+      "epoch": 78.62,
+      "learning_rate": 6.744230769230769e-05,
+      "loss": 0.5847,
+      "step": 8570
+    },
+    {
+      "epoch": 78.63,
+      "learning_rate": 6.741346153846153e-05,
+      "loss": 0.5791,
+      "step": 8571
+    },
+    {
+      "epoch": 78.64,
+      "learning_rate": 6.738461538461538e-05,
+      "loss": 0.6195,
+      "step": 8572
+    },
+    {
+      "epoch": 78.65,
+      "learning_rate": 6.735576923076923e-05,
+      "loss": 0.5156,
+      "step": 8573
+    },
+    {
+      "epoch": 78.66,
+      "learning_rate": 6.732692307692307e-05,
+      "loss": 0.4869,
+      "step": 8574
+    },
+    {
+      "epoch": 78.67,
+      "learning_rate": 6.729807692307692e-05,
+      "loss": 0.5401,
+      "step": 8575
+    },
+    {
+      "epoch": 78.68,
+      "learning_rate": 6.726923076923076e-05,
+      "loss": 0.6162,
+      "step": 8576
+    },
+    {
+      "epoch": 78.69,
+      "learning_rate": 6.724038461538461e-05,
+      "loss": 0.456,
+      "step": 8577
+    },
+    {
+      "epoch": 78.7,
+      "learning_rate": 6.721153846153846e-05,
+      "loss": 0.5926,
+      "step": 8578
+    },
+    {
+      "epoch": 78.71,
+      "learning_rate": 6.71826923076923e-05,
+      "loss": 0.4643,
+      "step": 8579
+    },
+    {
+      "epoch": 78.72,
+      "learning_rate": 6.715384615384614e-05,
+      "loss": 0.673,
+      "step": 8580
+    },
+    {
+      "epoch": 78.72,
+      "learning_rate": 6.712499999999999e-05,
+      "loss": 0.5574,
+      "step": 8581
+    },
+    {
+      "epoch": 78.73,
+      "learning_rate": 6.709615384615384e-05,
+      "loss": 0.5648,
+      "step": 8582
+    },
+    {
+      "epoch": 78.74,
+      "learning_rate": 6.70673076923077e-05,
+      "loss": 0.6451,
+      "step": 8583
+    },
+    {
+      "epoch": 78.75,
+      "learning_rate": 6.703846153846153e-05,
+      "loss": 0.5411,
+      "step": 8584
+    },
+    {
+      "epoch": 78.76,
+      "learning_rate": 6.700961538461537e-05,
+      "loss": 0.5627,
+      "step": 8585
+    },
+    {
+      "epoch": 78.77,
+      "learning_rate": 6.698076923076922e-05,
+      "loss": 0.602,
+      "step": 8586
+    },
+    {
+      "epoch": 78.78,
+      "learning_rate": 6.695192307692307e-05,
+      "loss": 0.5553,
+      "step": 8587
+    },
+    {
+      "epoch": 78.79,
+      "learning_rate": 6.692307692307693e-05,
+      "loss": 0.5932,
+      "step": 8588
+    },
+    {
+      "epoch": 78.8,
+      "learning_rate": 6.689423076923076e-05,
+      "loss": 0.6359,
+      "step": 8589
+    },
+    {
+      "epoch": 78.81,
+      "learning_rate": 6.68653846153846e-05,
+      "loss": 0.548,
+      "step": 8590
+    },
+    {
+      "epoch": 78.82,
+      "learning_rate": 6.683653846153845e-05,
+      "loss": 0.645,
+      "step": 8591
+    },
+    {
+      "epoch": 78.83,
+      "learning_rate": 6.68076923076923e-05,
+      "loss": 0.5814,
+      "step": 8592
+    },
+    {
+      "epoch": 78.83,
+      "learning_rate": 6.677884615384614e-05,
+      "loss": 0.5251,
+      "step": 8593
+    },
+    {
+      "epoch": 78.84,
+      "learning_rate": 6.675e-05,
+      "loss": 0.576,
+      "step": 8594
+    },
+    {
+      "epoch": 78.85,
+      "learning_rate": 6.672115384615383e-05,
+      "loss": 0.6271,
+      "step": 8595
+    },
+    {
+      "epoch": 78.86,
+      "learning_rate": 6.669230769230769e-05,
+      "loss": 0.6073,
+      "step": 8596
+    },
+    {
+      "epoch": 78.87,
+      "learning_rate": 6.666346153846154e-05,
+      "loss": 0.6189,
+      "step": 8597
+    },
+    {
+      "epoch": 78.88,
+      "learning_rate": 6.663461538461538e-05,
+      "loss": 0.5366,
+      "step": 8598
+    },
+    {
+      "epoch": 78.89,
+      "learning_rate": 6.660576923076923e-05,
+      "loss": 0.539,
+      "step": 8599
+    },
+    {
+      "epoch": 78.9,
+      "learning_rate": 6.657692307692307e-05,
+      "loss": 0.5436,
+      "step": 8600
+    },
+    {
+      "epoch": 78.91,
+      "learning_rate": 6.654807692307692e-05,
+      "loss": 0.5067,
+      "step": 8601
+    },
+    {
+      "epoch": 78.92,
+      "learning_rate": 6.651923076923077e-05,
+      "loss": 0.4958,
+      "step": 8602
+    },
+    {
+      "epoch": 78.93,
+      "learning_rate": 6.649038461538461e-05,
+      "loss": 0.599,
+      "step": 8603
+    },
+    {
+      "epoch": 78.94,
+      "learning_rate": 6.646153846153846e-05,
+      "loss": 0.4958,
+      "step": 8604
+    },
+    {
+      "epoch": 78.94,
+      "learning_rate": 6.64326923076923e-05,
+      "loss": 0.7004,
+      "step": 8605
+    },
+    {
+      "epoch": 78.95,
+      "learning_rate": 6.640384615384615e-05,
+      "loss": 0.5507,
+      "step": 8606
+    },
+    {
+      "epoch": 78.96,
+      "learning_rate": 6.6375e-05,
+      "loss": 0.5765,
+      "step": 8607
+    },
+    {
+      "epoch": 78.97,
+      "learning_rate": 6.634615384615384e-05,
+      "loss": 0.5256,
+      "step": 8608
+    },
+    {
+      "epoch": 78.98,
+      "learning_rate": 6.631730769230769e-05,
+      "loss": 0.47,
+      "step": 8609
+    },
+    {
+      "epoch": 78.99,
+      "learning_rate": 6.628846153846153e-05,
+      "loss": 0.631,
+      "step": 8610
+    },
+    {
+      "epoch": 79.0,
+      "learning_rate": 6.625961538461538e-05,
+      "loss": 0.5897,
+      "step": 8611
+    },
+    {
+      "epoch": 79.01,
+      "learning_rate": 6.623076923076923e-05,
+      "loss": 0.617,
+      "step": 8612
+    },
+    {
+      "epoch": 79.02,
+      "learning_rate": 6.620192307692307e-05,
+      "loss": 0.5531,
+      "step": 8613
+    },
+    {
+      "epoch": 79.03,
+      "learning_rate": 6.617307692307692e-05,
+      "loss": 0.5602,
+      "step": 8614
+    },
+    {
+      "epoch": 79.04,
+      "learning_rate": 6.614423076923076e-05,
+      "loss": 0.5437,
+      "step": 8615
+    },
+    {
+      "epoch": 79.05,
+      "learning_rate": 6.611538461538461e-05,
+      "loss": 0.5715,
+      "step": 8616
+    },
+    {
+      "epoch": 79.06,
+      "learning_rate": 6.608653846153846e-05,
+      "loss": 0.5643,
+      "step": 8617
+    },
+    {
+      "epoch": 79.06,
+      "learning_rate": 6.60576923076923e-05,
+      "loss": 0.5827,
+      "step": 8618
+    },
+    {
+      "epoch": 79.07,
+      "learning_rate": 6.602884615384615e-05,
+      "loss": 0.3995,
+      "step": 8619
+    },
+    {
+      "epoch": 79.08,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 0.6241,
+      "step": 8620
+    },
+    {
+      "epoch": 79.09,
+      "learning_rate": 6.597115384615384e-05,
+      "loss": 0.5967,
+      "step": 8621
+    },
+    {
+      "epoch": 79.1,
+      "learning_rate": 6.59423076923077e-05,
+      "loss": 0.5504,
+      "step": 8622
+    },
+    {
+      "epoch": 79.11,
+      "learning_rate": 6.591346153846153e-05,
+      "loss": 0.6242,
+      "step": 8623
+    },
+    {
+      "epoch": 79.12,
+      "learning_rate": 6.588461538461538e-05,
+      "loss": 0.6079,
+      "step": 8624
+    },
+    {
+      "epoch": 79.13,
+      "learning_rate": 6.585576923076922e-05,
+      "loss": 0.6044,
+      "step": 8625
+    },
+    {
+      "epoch": 79.14,
+      "learning_rate": 6.582692307692307e-05,
+      "loss": 0.537,
+      "step": 8626
+    },
+    {
+      "epoch": 79.15,
+      "learning_rate": 6.579807692307693e-05,
+      "loss": 0.5722,
+      "step": 8627
+    },
+    {
+      "epoch": 79.16,
+      "learning_rate": 6.576923076923076e-05,
+      "loss": 0.5234,
+      "step": 8628
+    },
+    {
+      "epoch": 79.17,
+      "learning_rate": 6.57403846153846e-05,
+      "loss": 0.6253,
+      "step": 8629
+    },
+    {
+      "epoch": 79.17,
+      "learning_rate": 6.571153846153845e-05,
+      "loss": 0.5854,
+      "step": 8630
+    },
+    {
+      "epoch": 79.18,
+      "learning_rate": 6.56826923076923e-05,
+      "loss": 0.5099,
+      "step": 8631
+    },
+    {
+      "epoch": 79.19,
+      "learning_rate": 6.565384615384616e-05,
+      "loss": 0.5758,
+      "step": 8632
+    },
+    {
+      "epoch": 79.2,
+      "learning_rate": 6.5625e-05,
+      "loss": 0.4792,
+      "step": 8633
+    },
+    {
+      "epoch": 79.21,
+      "learning_rate": 6.559615384615383e-05,
+      "loss": 0.6,
+      "step": 8634
+    },
+    {
+      "epoch": 79.22,
+      "learning_rate": 6.556730769230769e-05,
+      "loss": 0.6035,
+      "step": 8635
+    },
+    {
+      "epoch": 79.23,
+      "learning_rate": 6.553846153846154e-05,
+      "loss": 0.5899,
+      "step": 8636
+    },
+    {
+      "epoch": 79.24,
+      "learning_rate": 6.550961538461539e-05,
+      "loss": 0.5922,
+      "step": 8637
+    },
+    {
+      "epoch": 79.25,
+      "learning_rate": 6.548076923076923e-05,
+      "loss": 0.5545,
+      "step": 8638
+    },
+    {
+      "epoch": 79.26,
+      "learning_rate": 6.545192307692307e-05,
+      "loss": 0.4648,
+      "step": 8639
+    },
+    {
+      "epoch": 79.27,
+      "learning_rate": 6.542307692307692e-05,
+      "loss": 0.627,
+      "step": 8640
+    },
+    {
+      "epoch": 79.28,
+      "learning_rate": 6.539423076923077e-05,
+      "loss": 0.6324,
+      "step": 8641
+    },
+    {
+      "epoch": 79.28,
+      "learning_rate": 6.536538461538461e-05,
+      "loss": 0.6143,
+      "step": 8642
+    },
+    {
+      "epoch": 79.29,
+      "learning_rate": 6.533653846153846e-05,
+      "loss": 0.5772,
+      "step": 8643
+    },
+    {
+      "epoch": 79.3,
+      "learning_rate": 6.53076923076923e-05,
+      "loss": 0.7335,
+      "step": 8644
+    },
+    {
+      "epoch": 79.31,
+      "learning_rate": 6.527884615384615e-05,
+      "loss": 0.5001,
+      "step": 8645
+    },
+    {
+      "epoch": 79.32,
+      "learning_rate": 6.525e-05,
+      "loss": 0.6496,
+      "step": 8646
+    },
+    {
+      "epoch": 79.33,
+      "learning_rate": 6.522115384615384e-05,
+      "loss": 0.6554,
+      "step": 8647
+    },
+    {
+      "epoch": 79.34,
+      "learning_rate": 6.519230769230769e-05,
+      "loss": 0.7093,
+      "step": 8648
+    },
+    {
+      "epoch": 79.35,
+      "learning_rate": 6.516346153846153e-05,
+      "loss": 0.5851,
+      "step": 8649
+    },
+    {
+      "epoch": 79.36,
+      "learning_rate": 6.513461538461538e-05,
+      "loss": 0.6051,
+      "step": 8650
+    },
+    {
+      "epoch": 79.37,
+      "learning_rate": 6.510576923076923e-05,
+      "loss": 0.5581,
+      "step": 8651
+    },
+    {
+      "epoch": 79.38,
+      "learning_rate": 6.507692307692307e-05,
+      "loss": 0.55,
+      "step": 8652
+    },
+    {
+      "epoch": 79.39,
+      "learning_rate": 6.504807692307692e-05,
+      "loss": 0.5825,
+      "step": 8653
+    },
+    {
+      "epoch": 79.39,
+      "learning_rate": 6.501923076923076e-05,
+      "loss": 0.6289,
+      "step": 8654
+    },
+    {
+      "epoch": 79.4,
+      "learning_rate": 6.499038461538461e-05,
+      "loss": 0.4888,
+      "step": 8655
+    },
+    {
+      "epoch": 79.41,
+      "learning_rate": 6.496153846153846e-05,
+      "loss": 0.465,
+      "step": 8656
+    },
+    {
+      "epoch": 79.42,
+      "learning_rate": 6.49326923076923e-05,
+      "loss": 0.5022,
+      "step": 8657
+    },
+    {
+      "epoch": 79.43,
+      "learning_rate": 6.490384615384615e-05,
+      "loss": 0.6591,
+      "step": 8658
+    },
+    {
+      "epoch": 79.44,
+      "learning_rate": 6.487499999999999e-05,
+      "loss": 0.6396,
+      "step": 8659
+    },
+    {
+      "epoch": 79.45,
+      "learning_rate": 6.484615384615384e-05,
+      "loss": 0.6742,
+      "step": 8660
+    },
+    {
+      "epoch": 79.46,
+      "learning_rate": 6.481730769230768e-05,
+      "loss": 0.6878,
+      "step": 8661
+    },
+    {
+      "epoch": 79.47,
+      "learning_rate": 6.478846153846153e-05,
+      "loss": 0.6056,
+      "step": 8662
+    },
+    {
+      "epoch": 79.48,
+      "learning_rate": 6.475961538461538e-05,
+      "loss": 0.5006,
+      "step": 8663
+    },
+    {
+      "epoch": 79.49,
+      "learning_rate": 6.473076923076922e-05,
+      "loss": 0.5609,
+      "step": 8664
+    },
+    {
+      "epoch": 79.5,
+      "learning_rate": 6.470192307692307e-05,
+      "loss": 0.6303,
+      "step": 8665
+    },
+    {
+      "epoch": 79.5,
+      "learning_rate": 6.467307692307691e-05,
+      "loss": 0.547,
+      "step": 8666
+    },
+    {
+      "epoch": 79.51,
+      "learning_rate": 6.464423076923076e-05,
+      "loss": 0.5606,
+      "step": 8667
+    },
+    {
+      "epoch": 79.52,
+      "learning_rate": 6.461538461538462e-05,
+      "loss": 0.5969,
+      "step": 8668
+    },
+    {
+      "epoch": 79.53,
+      "learning_rate": 6.458653846153845e-05,
+      "loss": 0.5365,
+      "step": 8669
+    },
+    {
+      "epoch": 79.54,
+      "learning_rate": 6.45576923076923e-05,
+      "loss": 0.606,
+      "step": 8670
+    },
+    {
+      "epoch": 79.55,
+      "learning_rate": 6.452884615384614e-05,
+      "loss": 0.4848,
+      "step": 8671
+    },
+    {
+      "epoch": 79.56,
+      "learning_rate": 6.45e-05,
+      "loss": 0.6897,
+      "step": 8672
+    },
+    {
+      "epoch": 79.57,
+      "learning_rate": 6.447115384615385e-05,
+      "loss": 0.5693,
+      "step": 8673
+    },
+    {
+      "epoch": 79.58,
+      "learning_rate": 6.444230769230769e-05,
+      "loss": 0.6127,
+      "step": 8674
+    },
+    {
+      "epoch": 79.59,
+      "learning_rate": 6.441346153846152e-05,
+      "loss": 0.5046,
+      "step": 8675
+    },
+    {
+      "epoch": 79.6,
+      "learning_rate": 6.438461538461538e-05,
+      "loss": 0.521,
+      "step": 8676
+    },
+    {
+      "epoch": 79.61,
+      "learning_rate": 6.435576923076923e-05,
+      "loss": 0.5319,
+      "step": 8677
+    },
+    {
+      "epoch": 79.61,
+      "learning_rate": 6.432692307692307e-05,
+      "loss": 0.5903,
+      "step": 8678
+    },
+    {
+      "epoch": 79.62,
+      "learning_rate": 6.429807692307692e-05,
+      "loss": 0.5936,
+      "step": 8679
+    },
+    {
+      "epoch": 79.63,
+      "learning_rate": 6.426923076923076e-05,
+      "loss": 0.5502,
+      "step": 8680
+    },
+    {
+      "epoch": 79.64,
+      "learning_rate": 6.424038461538461e-05,
+      "loss": 0.5539,
+      "step": 8681
+    },
+    {
+      "epoch": 79.65,
+      "learning_rate": 6.421153846153846e-05,
+      "loss": 0.6046,
+      "step": 8682
+    },
+    {
+      "epoch": 79.66,
+      "learning_rate": 6.41826923076923e-05,
+      "loss": 0.4703,
+      "step": 8683
+    },
+    {
+      "epoch": 79.67,
+      "learning_rate": 6.415384615384615e-05,
+      "loss": 0.6341,
+      "step": 8684
+    },
+    {
+      "epoch": 79.68,
+      "learning_rate": 6.412499999999999e-05,
+      "loss": 0.6563,
+      "step": 8685
+    },
+    {
+      "epoch": 79.69,
+      "learning_rate": 6.409615384615384e-05,
+      "loss": 0.5325,
+      "step": 8686
+    },
+    {
+      "epoch": 79.7,
+      "learning_rate": 6.406730769230769e-05,
+      "loss": 0.5446,
+      "step": 8687
+    },
+    {
+      "epoch": 79.71,
+      "learning_rate": 6.403846153846153e-05,
+      "loss": 0.5062,
+      "step": 8688
+    },
+    {
+      "epoch": 79.72,
+      "learning_rate": 6.400961538461538e-05,
+      "loss": 0.5371,
+      "step": 8689
+    },
+    {
+      "epoch": 79.72,
+      "learning_rate": 6.398076923076922e-05,
+      "loss": 0.4774,
+      "step": 8690
+    },
+    {
+      "epoch": 79.73,
+      "learning_rate": 6.395192307692307e-05,
+      "loss": 0.564,
+      "step": 8691
+    },
+    {
+      "epoch": 79.74,
+      "learning_rate": 6.392307692307692e-05,
+      "loss": 0.5523,
+      "step": 8692
+    },
+    {
+      "epoch": 79.75,
+      "learning_rate": 6.389423076923076e-05,
+      "loss": 0.5327,
+      "step": 8693
+    },
+    {
+      "epoch": 79.76,
+      "learning_rate": 6.386538461538461e-05,
+      "loss": 0.5239,
+      "step": 8694
+    },
+    {
+      "epoch": 79.77,
+      "learning_rate": 6.383653846153845e-05,
+      "loss": 0.5522,
+      "step": 8695
+    },
+    {
+      "epoch": 79.78,
+      "learning_rate": 6.38076923076923e-05,
+      "loss": 0.6148,
+      "step": 8696
+    },
+    {
+      "epoch": 79.79,
+      "learning_rate": 6.377884615384615e-05,
+      "loss": 0.602,
+      "step": 8697
+    },
+    {
+      "epoch": 79.8,
+      "learning_rate": 6.374999999999999e-05,
+      "loss": 0.5636,
+      "step": 8698
+    },
+    {
+      "epoch": 79.81,
+      "learning_rate": 6.372115384615384e-05,
+      "loss": 0.4472,
+      "step": 8699
+    },
+    {
+      "epoch": 79.82,
+      "learning_rate": 6.369230769230768e-05,
+      "loss": 0.6438,
+      "step": 8700
+    },
+    {
+      "epoch": 79.83,
+      "learning_rate": 6.366346153846153e-05,
+      "loss": 0.5056,
+      "step": 8701
+    },
+    {
+      "epoch": 79.83,
+      "learning_rate": 6.363461538461538e-05,
+      "loss": 0.632,
+      "step": 8702
+    },
+    {
+      "epoch": 79.84,
+      "learning_rate": 6.360576923076922e-05,
+      "loss": 0.5995,
+      "step": 8703
+    },
+    {
+      "epoch": 79.85,
+      "learning_rate": 6.357692307692307e-05,
+      "loss": 0.4294,
+      "step": 8704
+    },
+    {
+      "epoch": 79.86,
+      "learning_rate": 6.354807692307691e-05,
+      "loss": 0.6644,
+      "step": 8705
+    },
+    {
+      "epoch": 79.87,
+      "learning_rate": 6.351923076923076e-05,
+      "loss": 0.5596,
+      "step": 8706
+    },
+    {
+      "epoch": 79.88,
+      "learning_rate": 6.349038461538462e-05,
+      "loss": 0.6136,
+      "step": 8707
+    },
+    {
+      "epoch": 79.89,
+      "learning_rate": 6.346153846153845e-05,
+      "loss": 0.5632,
+      "step": 8708
+    },
+    {
+      "epoch": 79.9,
+      "learning_rate": 6.343269230769229e-05,
+      "loss": 0.5109,
+      "step": 8709
+    },
+    {
+      "epoch": 79.91,
+      "learning_rate": 6.340384615384614e-05,
+      "loss": 0.6238,
+      "step": 8710
+    },
+    {
+      "epoch": 79.92,
+      "learning_rate": 6.3375e-05,
+      "loss": 0.4784,
+      "step": 8711
+    },
+    {
+      "epoch": 79.93,
+      "learning_rate": 6.334615384615385e-05,
+      "loss": 0.5075,
+      "step": 8712
+    },
+    {
+      "epoch": 79.94,
+      "learning_rate": 6.331730769230769e-05,
+      "loss": 0.5315,
+      "step": 8713
+    },
+    {
+      "epoch": 79.94,
+      "learning_rate": 6.328846153846152e-05,
+      "loss": 0.4513,
+      "step": 8714
+    },
+    {
+      "epoch": 79.95,
+      "learning_rate": 6.325961538461538e-05,
+      "loss": 0.6405,
+      "step": 8715
+    },
+    {
+      "epoch": 79.96,
+      "learning_rate": 6.323076923076923e-05,
+      "loss": 0.5565,
+      "step": 8716
+    },
+    {
+      "epoch": 79.97,
+      "learning_rate": 6.320192307692308e-05,
+      "loss": 0.5876,
+      "step": 8717
+    },
+    {
+      "epoch": 79.98,
+      "learning_rate": 6.317307692307692e-05,
+      "loss": 0.5612,
+      "step": 8718
+    },
+    {
+      "epoch": 79.99,
+      "learning_rate": 6.314423076923075e-05,
+      "loss": 0.5238,
+      "step": 8719
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 6.31153846153846e-05,
+      "loss": 0.5234,
+      "step": 8720
+    },
+    {
+      "epoch": 80.01,
+      "learning_rate": 6.308653846153846e-05,
+      "loss": 0.5937,
+      "step": 8721
+    },
+    {
+      "epoch": 80.02,
+      "learning_rate": 6.305769230769231e-05,
+      "loss": 0.5183,
+      "step": 8722
+    },
+    {
+      "epoch": 80.03,
+      "learning_rate": 6.302884615384615e-05,
+      "loss": 0.3959,
+      "step": 8723
+    },
+    {
+      "epoch": 80.04,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 0.4972,
+      "step": 8724
+    },
+    {
+      "epoch": 80.05,
+      "learning_rate": 6.297115384615384e-05,
+      "loss": 0.5967,
+      "step": 8725
+    },
+    {
+      "epoch": 80.06,
+      "learning_rate": 6.294230769230769e-05,
+      "loss": 0.5022,
+      "step": 8726
+    },
+    {
+      "epoch": 80.06,
+      "learning_rate": 6.291346153846153e-05,
+      "loss": 0.5565,
+      "step": 8727
+    },
+    {
+      "epoch": 80.07,
+      "learning_rate": 6.288461538461538e-05,
+      "loss": 0.4458,
+      "step": 8728
+    },
+    {
+      "epoch": 80.08,
+      "learning_rate": 6.285576923076922e-05,
+      "loss": 0.5257,
+      "step": 8729
+    },
+    {
+      "epoch": 80.09,
+      "learning_rate": 6.282692307692307e-05,
+      "loss": 0.5497,
+      "step": 8730
+    },
+    {
+      "epoch": 80.1,
+      "learning_rate": 6.279807692307692e-05,
+      "loss": 0.5452,
+      "step": 8731
+    },
+    {
+      "epoch": 80.11,
+      "learning_rate": 6.276923076923076e-05,
+      "loss": 0.6295,
+      "step": 8732
+    },
+    {
+      "epoch": 80.12,
+      "learning_rate": 6.274038461538461e-05,
+      "loss": 0.4846,
+      "step": 8733
+    },
+    {
+      "epoch": 80.13,
+      "learning_rate": 6.271153846153845e-05,
+      "loss": 0.6439,
+      "step": 8734
+    },
+    {
+      "epoch": 80.14,
+      "learning_rate": 6.26826923076923e-05,
+      "loss": 0.4587,
+      "step": 8735
+    },
+    {
+      "epoch": 80.15,
+      "learning_rate": 6.265384615384615e-05,
+      "loss": 0.6111,
+      "step": 8736
+    },
+    {
+      "epoch": 80.16,
+      "learning_rate": 6.262499999999999e-05,
+      "loss": 0.6006,
+      "step": 8737
+    },
+    {
+      "epoch": 80.17,
+      "learning_rate": 6.259615384615384e-05,
+      "loss": 0.5377,
+      "step": 8738
+    },
+    {
+      "epoch": 80.17,
+      "learning_rate": 6.256730769230768e-05,
+      "loss": 0.6079,
+      "step": 8739
+    },
+    {
+      "epoch": 80.18,
+      "learning_rate": 6.253846153846153e-05,
+      "loss": 0.5757,
+      "step": 8740
+    },
+    {
+      "epoch": 80.19,
+      "learning_rate": 6.250961538461538e-05,
+      "loss": 0.4982,
+      "step": 8741
+    },
+    {
+      "epoch": 80.2,
+      "learning_rate": 6.248076923076922e-05,
+      "loss": 0.4913,
+      "step": 8742
+    },
+    {
+      "epoch": 80.21,
+      "learning_rate": 6.245192307692307e-05,
+      "loss": 0.6236,
+      "step": 8743
+    },
+    {
+      "epoch": 80.22,
+      "learning_rate": 6.242307692307691e-05,
+      "loss": 0.5844,
+      "step": 8744
+    },
+    {
+      "epoch": 80.23,
+      "learning_rate": 6.239423076923076e-05,
+      "loss": 0.4513,
+      "step": 8745
+    },
+    {
+      "epoch": 80.24,
+      "learning_rate": 6.236538461538462e-05,
+      "loss": 0.6571,
+      "step": 8746
+    },
+    {
+      "epoch": 80.25,
+      "learning_rate": 6.233653846153845e-05,
+      "loss": 0.4773,
+      "step": 8747
+    },
+    {
+      "epoch": 80.26,
+      "learning_rate": 6.23076923076923e-05,
+      "loss": 0.5668,
+      "step": 8748
+    },
+    {
+      "epoch": 80.27,
+      "learning_rate": 6.227884615384614e-05,
+      "loss": 0.6101,
+      "step": 8749
+    },
+    {
+      "epoch": 80.28,
+      "learning_rate": 6.225e-05,
+      "loss": 0.5832,
+      "step": 8750
+    },
+    {
+      "epoch": 80.28,
+      "learning_rate": 6.222115384615385e-05,
+      "loss": 0.5846,
+      "step": 8751
+    },
+    {
+      "epoch": 80.29,
+      "learning_rate": 6.219230769230769e-05,
+      "loss": 0.6856,
+      "step": 8752
+    },
+    {
+      "epoch": 80.3,
+      "learning_rate": 6.216346153846154e-05,
+      "loss": 0.5346,
+      "step": 8753
+    },
+    {
+      "epoch": 80.31,
+      "learning_rate": 6.213461538461537e-05,
+      "loss": 0.4911,
+      "step": 8754
+    },
+    {
+      "epoch": 80.32,
+      "learning_rate": 6.210576923076923e-05,
+      "loss": 0.4852,
+      "step": 8755
+    },
+    {
+      "epoch": 80.33,
+      "learning_rate": 6.207692307692308e-05,
+      "loss": 0.4848,
+      "step": 8756
+    },
+    {
+      "epoch": 80.34,
+      "learning_rate": 6.204807692307692e-05,
+      "loss": 0.4777,
+      "step": 8757
+    },
+    {
+      "epoch": 80.35,
+      "learning_rate": 6.201923076923075e-05,
+      "loss": 0.7069,
+      "step": 8758
+    },
+    {
+      "epoch": 80.36,
+      "learning_rate": 6.19903846153846e-05,
+      "loss": 0.543,
+      "step": 8759
+    },
+    {
+      "epoch": 80.37,
+      "learning_rate": 6.196153846153846e-05,
+      "loss": 0.5259,
+      "step": 8760
+    },
+    {
+      "epoch": 80.38,
+      "learning_rate": 6.193269230769231e-05,
+      "loss": 0.5589,
+      "step": 8761
+    },
+    {
+      "epoch": 80.39,
+      "learning_rate": 6.190384615384615e-05,
+      "loss": 0.5248,
+      "step": 8762
+    },
+    {
+      "epoch": 80.39,
+      "learning_rate": 6.187499999999999e-05,
+      "loss": 0.5448,
+      "step": 8763
+    },
+    {
+      "epoch": 80.4,
+      "learning_rate": 6.184615384615384e-05,
+      "loss": 0.3521,
+      "step": 8764
+    },
+    {
+      "epoch": 80.41,
+      "learning_rate": 6.181730769230769e-05,
+      "loss": 0.6337,
+      "step": 8765
+    },
+    {
+      "epoch": 80.42,
+      "learning_rate": 6.178846153846154e-05,
+      "loss": 0.3845,
+      "step": 8766
+    },
+    {
+      "epoch": 80.43,
+      "learning_rate": 6.175961538461538e-05,
+      "loss": 0.4994,
+      "step": 8767
+    },
+    {
+      "epoch": 80.44,
+      "learning_rate": 6.173076923076922e-05,
+      "loss": 0.554,
+      "step": 8768
+    },
+    {
+      "epoch": 80.45,
+      "learning_rate": 6.170192307692307e-05,
+      "loss": 0.5588,
+      "step": 8769
+    },
+    {
+      "epoch": 80.46,
+      "learning_rate": 6.167307692307692e-05,
+      "loss": 0.5831,
+      "step": 8770
+    },
+    {
+      "epoch": 80.47,
+      "learning_rate": 6.164423076923077e-05,
+      "loss": 0.5816,
+      "step": 8771
+    },
+    {
+      "epoch": 80.48,
+      "learning_rate": 6.161538461538461e-05,
+      "loss": 0.5786,
+      "step": 8772
+    },
+    {
+      "epoch": 80.49,
+      "learning_rate": 6.158653846153845e-05,
+      "loss": 0.5303,
+      "step": 8773
+    },
+    {
+      "epoch": 80.5,
+      "learning_rate": 6.15576923076923e-05,
+      "loss": 0.6145,
+      "step": 8774
+    },
+    {
+      "epoch": 80.5,
+      "learning_rate": 6.152884615384615e-05,
+      "loss": 0.6503,
+      "step": 8775
+    },
+    {
+      "epoch": 80.51,
+      "learning_rate": 6.149999999999999e-05,
+      "loss": 0.7308,
+      "step": 8776
+    },
+    {
+      "epoch": 80.52,
+      "learning_rate": 6.147115384615384e-05,
+      "loss": 0.6108,
+      "step": 8777
+    },
+    {
+      "epoch": 80.53,
+      "learning_rate": 6.144230769230768e-05,
+      "loss": 0.5283,
+      "step": 8778
+    },
+    {
+      "epoch": 80.54,
+      "learning_rate": 6.141346153846153e-05,
+      "loss": 0.5175,
+      "step": 8779
+    },
+    {
+      "epoch": 80.55,
+      "learning_rate": 6.138461538461538e-05,
+      "loss": 0.5832,
+      "step": 8780
+    },
+    {
+      "epoch": 80.56,
+      "learning_rate": 6.135576923076922e-05,
+      "loss": 0.4491,
+      "step": 8781
+    },
+    {
+      "epoch": 80.57,
+      "learning_rate": 6.132692307692307e-05,
+      "loss": 0.5683,
+      "step": 8782
+    },
+    {
+      "epoch": 80.58,
+      "learning_rate": 6.129807692307691e-05,
+      "loss": 0.4997,
+      "step": 8783
+    },
+    {
+      "epoch": 80.59,
+      "learning_rate": 6.126923076923076e-05,
+      "loss": 0.5272,
+      "step": 8784
+    },
+    {
+      "epoch": 80.6,
+      "learning_rate": 6.124038461538462e-05,
+      "loss": 0.6029,
+      "step": 8785
+    },
+    {
+      "epoch": 80.61,
+      "learning_rate": 6.121153846153845e-05,
+      "loss": 0.5444,
+      "step": 8786
+    },
+    {
+      "epoch": 80.61,
+      "learning_rate": 6.11826923076923e-05,
+      "loss": 0.6225,
+      "step": 8787
+    },
+    {
+      "epoch": 80.62,
+      "learning_rate": 6.115384615384614e-05,
+      "loss": 0.4809,
+      "step": 8788
+    },
+    {
+      "epoch": 80.63,
+      "learning_rate": 6.1125e-05,
+      "loss": 0.4465,
+      "step": 8789
+    },
+    {
+      "epoch": 80.64,
+      "learning_rate": 6.109615384615385e-05,
+      "loss": 0.6303,
+      "step": 8790
+    },
+    {
+      "epoch": 80.65,
+      "learning_rate": 6.106730769230768e-05,
+      "loss": 0.6261,
+      "step": 8791
+    },
+    {
+      "epoch": 80.66,
+      "learning_rate": 6.103846153846154e-05,
+      "loss": 0.6515,
+      "step": 8792
+    },
+    {
+      "epoch": 80.67,
+      "learning_rate": 6.100961538461538e-05,
+      "loss": 0.5023,
+      "step": 8793
+    },
+    {
+      "epoch": 80.68,
+      "learning_rate": 6.098076923076922e-05,
+      "loss": 0.5637,
+      "step": 8794
+    },
+    {
+      "epoch": 80.69,
+      "learning_rate": 6.095192307692307e-05,
+      "loss": 0.493,
+      "step": 8795
+    },
+    {
+      "epoch": 80.7,
+      "learning_rate": 6.0923076923076916e-05,
+      "loss": 0.6755,
+      "step": 8796
+    },
+    {
+      "epoch": 80.71,
+      "learning_rate": 6.089423076923077e-05,
+      "loss": 0.628,
+      "step": 8797
+    },
+    {
+      "epoch": 80.72,
+      "learning_rate": 6.086538461538461e-05,
+      "loss": 0.5955,
+      "step": 8798
+    },
+    {
+      "epoch": 80.72,
+      "learning_rate": 6.083653846153845e-05,
+      "loss": 0.5711,
+      "step": 8799
+    },
+    {
+      "epoch": 80.73,
+      "learning_rate": 6.08076923076923e-05,
+      "loss": 0.5601,
+      "step": 8800
+    },
+    {
+      "epoch": 80.74,
+      "learning_rate": 6.077884615384615e-05,
+      "loss": 0.3976,
+      "step": 8801
+    },
+    {
+      "epoch": 80.75,
+      "learning_rate": 6.075e-05,
+      "loss": 0.5785,
+      "step": 8802
+    },
+    {
+      "epoch": 80.76,
+      "learning_rate": 6.0721153846153844e-05,
+      "loss": 0.5182,
+      "step": 8803
+    },
+    {
+      "epoch": 80.77,
+      "learning_rate": 6.069230769230768e-05,
+      "loss": 0.5753,
+      "step": 8804
+    },
+    {
+      "epoch": 80.78,
+      "learning_rate": 6.0663461538461534e-05,
+      "loss": 0.5962,
+      "step": 8805
+    },
+    {
+      "epoch": 80.79,
+      "learning_rate": 6.063461538461538e-05,
+      "loss": 0.4998,
+      "step": 8806
+    },
+    {
+      "epoch": 80.8,
+      "learning_rate": 6.0605769230769224e-05,
+      "loss": 0.4267,
+      "step": 8807
+    },
+    {
+      "epoch": 80.81,
+      "learning_rate": 6.0576923076923076e-05,
+      "loss": 0.4508,
+      "step": 8808
+    },
+    {
+      "epoch": 80.82,
+      "learning_rate": 6.0548076923076914e-05,
+      "loss": 0.4723,
+      "step": 8809
+    },
+    {
+      "epoch": 80.83,
+      "learning_rate": 6.0519230769230766e-05,
+      "loss": 0.6136,
+      "step": 8810
+    },
+    {
+      "epoch": 80.83,
+      "learning_rate": 6.049038461538461e-05,
+      "loss": 0.475,
+      "step": 8811
+    },
+    {
+      "epoch": 80.84,
+      "learning_rate": 6.0461538461538456e-05,
+      "loss": 0.5652,
+      "step": 8812
+    },
+    {
+      "epoch": 80.85,
+      "learning_rate": 6.043269230769231e-05,
+      "loss": 0.5701,
+      "step": 8813
+    },
+    {
+      "epoch": 80.86,
+      "learning_rate": 6.0403846153846145e-05,
+      "loss": 0.6232,
+      "step": 8814
+    },
+    {
+      "epoch": 80.87,
+      "learning_rate": 6.0375e-05,
+      "loss": 0.5137,
+      "step": 8815
+    },
+    {
+      "epoch": 80.88,
+      "learning_rate": 6.034615384615384e-05,
+      "loss": 0.5624,
+      "step": 8816
+    },
+    {
+      "epoch": 80.89,
+      "learning_rate": 6.031730769230769e-05,
+      "loss": 0.5188,
+      "step": 8817
+    },
+    {
+      "epoch": 80.9,
+      "learning_rate": 6.028846153846154e-05,
+      "loss": 0.5876,
+      "step": 8818
+    },
+    {
+      "epoch": 80.91,
+      "learning_rate": 6.025961538461538e-05,
+      "loss": 0.4517,
+      "step": 8819
+    },
+    {
+      "epoch": 80.92,
+      "learning_rate": 6.023076923076922e-05,
+      "loss": 0.4221,
+      "step": 8820
+    },
+    {
+      "epoch": 80.93,
+      "learning_rate": 6.0201923076923073e-05,
+      "loss": 0.5459,
+      "step": 8821
+    },
+    {
+      "epoch": 80.94,
+      "learning_rate": 6.017307692307692e-05,
+      "loss": 0.6348,
+      "step": 8822
+    },
+    {
+      "epoch": 80.94,
+      "learning_rate": 6.014423076923077e-05,
+      "loss": 0.5268,
+      "step": 8823
+    },
+    {
+      "epoch": 80.95,
+      "learning_rate": 6.011538461538461e-05,
+      "loss": 0.6058,
+      "step": 8824
+    },
+    {
+      "epoch": 80.96,
+      "learning_rate": 6.008653846153845e-05,
+      "loss": 0.6272,
+      "step": 8825
+    },
+    {
+      "epoch": 80.97,
+      "learning_rate": 6.0057692307692305e-05,
+      "loss": 0.7155,
+      "step": 8826
+    },
+    {
+      "epoch": 80.98,
+      "learning_rate": 6.002884615384615e-05,
+      "loss": 0.558,
+      "step": 8827
+    },
+    {
+      "epoch": 80.99,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 0.5073,
+      "step": 8828
+    },
+    {
+      "epoch": 81.0,
+      "learning_rate": 5.997115384615384e-05,
+      "loss": 0.541,
+      "step": 8829
+    },
+    {
+      "epoch": 81.01,
+      "learning_rate": 5.9942307692307685e-05,
+      "loss": 0.6544,
+      "step": 8830
+    },
+    {
+      "epoch": 81.02,
+      "learning_rate": 5.9913461538461536e-05,
+      "loss": 0.577,
+      "step": 8831
+    },
+    {
+      "epoch": 81.03,
+      "learning_rate": 5.988461538461538e-05,
+      "loss": 0.5103,
+      "step": 8832
+    },
+    {
+      "epoch": 81.04,
+      "learning_rate": 5.9855769230769226e-05,
+      "loss": 0.615,
+      "step": 8833
+    },
+    {
+      "epoch": 81.05,
+      "learning_rate": 5.982692307692307e-05,
+      "loss": 0.666,
+      "step": 8834
+    },
+    {
+      "epoch": 81.06,
+      "learning_rate": 5.9798076923076916e-05,
+      "loss": 0.4915,
+      "step": 8835
+    },
+    {
+      "epoch": 81.06,
+      "learning_rate": 5.976923076923077e-05,
+      "loss": 0.5166,
+      "step": 8836
+    },
+    {
+      "epoch": 81.07,
+      "learning_rate": 5.974038461538461e-05,
+      "loss": 0.6324,
+      "step": 8837
+    },
+    {
+      "epoch": 81.08,
+      "learning_rate": 5.971153846153845e-05,
+      "loss": 0.4574,
+      "step": 8838
+    },
+    {
+      "epoch": 81.09,
+      "learning_rate": 5.96826923076923e-05,
+      "loss": 0.5475,
+      "step": 8839
+    },
+    {
+      "epoch": 81.1,
+      "learning_rate": 5.965384615384615e-05,
+      "loss": 0.5491,
+      "step": 8840
+    },
+    {
+      "epoch": 81.11,
+      "learning_rate": 5.9625e-05,
+      "loss": 0.5257,
+      "step": 8841
+    },
+    {
+      "epoch": 81.12,
+      "learning_rate": 5.959615384615384e-05,
+      "loss": 0.5484,
+      "step": 8842
+    },
+    {
+      "epoch": 81.13,
+      "learning_rate": 5.956730769230768e-05,
+      "loss": 0.62,
+      "step": 8843
+    },
+    {
+      "epoch": 81.14,
+      "learning_rate": 5.9538461538461534e-05,
+      "loss": 0.5762,
+      "step": 8844
+    },
+    {
+      "epoch": 81.15,
+      "learning_rate": 5.950961538461538e-05,
+      "loss": 0.4593,
+      "step": 8845
+    },
+    {
+      "epoch": 81.16,
+      "learning_rate": 5.948076923076923e-05,
+      "loss": 0.5826,
+      "step": 8846
+    },
+    {
+      "epoch": 81.17,
+      "learning_rate": 5.945192307692307e-05,
+      "loss": 0.6696,
+      "step": 8847
+    },
+    {
+      "epoch": 81.17,
+      "learning_rate": 5.9423076923076914e-05,
+      "loss": 0.5475,
+      "step": 8848
+    },
+    {
+      "epoch": 81.18,
+      "learning_rate": 5.9394230769230765e-05,
+      "loss": 0.5375,
+      "step": 8849
+    },
+    {
+      "epoch": 81.19,
+      "learning_rate": 5.936538461538461e-05,
+      "loss": 0.5082,
+      "step": 8850
+    },
+    {
+      "epoch": 81.2,
+      "learning_rate": 5.933653846153846e-05,
+      "loss": 0.5623,
+      "step": 8851
+    },
+    {
+      "epoch": 81.21,
+      "learning_rate": 5.93076923076923e-05,
+      "loss": 0.4872,
+      "step": 8852
+    },
+    {
+      "epoch": 81.22,
+      "learning_rate": 5.9278846153846145e-05,
+      "loss": 0.5753,
+      "step": 8853
+    },
+    {
+      "epoch": 81.23,
+      "learning_rate": 5.925e-05,
+      "loss": 0.5206,
+      "step": 8854
+    },
+    {
+      "epoch": 81.24,
+      "learning_rate": 5.922115384615384e-05,
+      "loss": 0.5569,
+      "step": 8855
+    },
+    {
+      "epoch": 81.25,
+      "learning_rate": 5.919230769230769e-05,
+      "loss": 0.4755,
+      "step": 8856
+    },
+    {
+      "epoch": 81.26,
+      "learning_rate": 5.916346153846153e-05,
+      "loss": 0.5685,
+      "step": 8857
+    },
+    {
+      "epoch": 81.27,
+      "learning_rate": 5.9134615384615377e-05,
+      "loss": 0.5963,
+      "step": 8858
+    },
+    {
+      "epoch": 81.28,
+      "learning_rate": 5.910576923076923e-05,
+      "loss": 0.5909,
+      "step": 8859
+    },
+    {
+      "epoch": 81.28,
+      "learning_rate": 5.907692307692307e-05,
+      "loss": 0.4653,
+      "step": 8860
+    },
+    {
+      "epoch": 81.29,
+      "learning_rate": 5.904807692307691e-05,
+      "loss": 0.5715,
+      "step": 8861
+    },
+    {
+      "epoch": 81.3,
+      "learning_rate": 5.901923076923076e-05,
+      "loss": 0.5783,
+      "step": 8862
+    },
+    {
+      "epoch": 81.31,
+      "learning_rate": 5.899038461538461e-05,
+      "loss": 0.5483,
+      "step": 8863
+    },
+    {
+      "epoch": 81.32,
+      "learning_rate": 5.896153846153846e-05,
+      "loss": 0.5147,
+      "step": 8864
+    },
+    {
+      "epoch": 81.33,
+      "learning_rate": 5.8932692307692305e-05,
+      "loss": 0.45,
+      "step": 8865
+    },
+    {
+      "epoch": 81.34,
+      "learning_rate": 5.890384615384614e-05,
+      "loss": 0.6029,
+      "step": 8866
+    },
+    {
+      "epoch": 81.35,
+      "learning_rate": 5.8874999999999995e-05,
+      "loss": 0.5568,
+      "step": 8867
+    },
+    {
+      "epoch": 81.36,
+      "learning_rate": 5.884615384615384e-05,
+      "loss": 0.545,
+      "step": 8868
+    },
+    {
+      "epoch": 81.37,
+      "learning_rate": 5.8817307692307684e-05,
+      "loss": 0.5658,
+      "step": 8869
+    },
+    {
+      "epoch": 81.38,
+      "learning_rate": 5.8788461538461536e-05,
+      "loss": 0.5727,
+      "step": 8870
+    },
+    {
+      "epoch": 81.39,
+      "learning_rate": 5.8759615384615374e-05,
+      "loss": 0.5941,
+      "step": 8871
+    },
+    {
+      "epoch": 81.39,
+      "learning_rate": 5.8730769230769226e-05,
+      "loss": 0.4391,
+      "step": 8872
+    },
+    {
+      "epoch": 81.4,
+      "learning_rate": 5.870192307692307e-05,
+      "loss": 0.5651,
+      "step": 8873
+    },
+    {
+      "epoch": 81.41,
+      "learning_rate": 5.8673076923076916e-05,
+      "loss": 0.5795,
+      "step": 8874
+    },
+    {
+      "epoch": 81.42,
+      "learning_rate": 5.864423076923077e-05,
+      "loss": 0.5151,
+      "step": 8875
+    },
+    {
+      "epoch": 81.43,
+      "learning_rate": 5.8615384615384606e-05,
+      "loss": 0.4994,
+      "step": 8876
+    },
+    {
+      "epoch": 81.44,
+      "learning_rate": 5.858653846153846e-05,
+      "loss": 0.5385,
+      "step": 8877
+    },
+    {
+      "epoch": 81.45,
+      "learning_rate": 5.85576923076923e-05,
+      "loss": 0.428,
+      "step": 8878
+    },
+    {
+      "epoch": 81.46,
+      "learning_rate": 5.852884615384615e-05,
+      "loss": 0.375,
+      "step": 8879
+    },
+    {
+      "epoch": 81.47,
+      "learning_rate": 5.85e-05,
+      "loss": 0.6334,
+      "step": 8880
+    },
+    {
+      "epoch": 81.48,
+      "learning_rate": 5.847115384615384e-05,
+      "loss": 0.5589,
+      "step": 8881
+    },
+    {
+      "epoch": 81.49,
+      "learning_rate": 5.844230769230769e-05,
+      "loss": 0.5041,
+      "step": 8882
+    },
+    {
+      "epoch": 81.5,
+      "learning_rate": 5.8413461538461534e-05,
+      "loss": 0.5619,
+      "step": 8883
+    },
+    {
+      "epoch": 81.5,
+      "learning_rate": 5.838461538461538e-05,
+      "loss": 0.4965,
+      "step": 8884
+    },
+    {
+      "epoch": 81.51,
+      "learning_rate": 5.835576923076923e-05,
+      "loss": 0.5398,
+      "step": 8885
+    },
+    {
+      "epoch": 81.52,
+      "learning_rate": 5.832692307692307e-05,
+      "loss": 0.5794,
+      "step": 8886
+    },
+    {
+      "epoch": 81.53,
+      "learning_rate": 5.8298076923076913e-05,
+      "loss": 0.6009,
+      "step": 8887
+    },
+    {
+      "epoch": 81.54,
+      "learning_rate": 5.8269230769230765e-05,
+      "loss": 0.5541,
+      "step": 8888
+    },
+    {
+      "epoch": 81.55,
+      "learning_rate": 5.824038461538461e-05,
+      "loss": 0.6089,
+      "step": 8889
+    },
+    {
+      "epoch": 81.56,
+      "learning_rate": 5.821153846153846e-05,
+      "loss": 0.5002,
+      "step": 8890
+    },
+    {
+      "epoch": 81.57,
+      "learning_rate": 5.81826923076923e-05,
+      "loss": 0.4559,
+      "step": 8891
+    },
+    {
+      "epoch": 81.58,
+      "learning_rate": 5.8153846153846145e-05,
+      "loss": 0.4485,
+      "step": 8892
+    },
+    {
+      "epoch": 81.59,
+      "learning_rate": 5.8124999999999997e-05,
+      "loss": 0.5327,
+      "step": 8893
+    },
+    {
+      "epoch": 81.6,
+      "learning_rate": 5.809615384615384e-05,
+      "loss": 0.4338,
+      "step": 8894
+    },
+    {
+      "epoch": 81.61,
+      "learning_rate": 5.806730769230769e-05,
+      "loss": 0.5244,
+      "step": 8895
+    },
+    {
+      "epoch": 81.61,
+      "learning_rate": 5.803846153846153e-05,
+      "loss": 0.5612,
+      "step": 8896
+    },
+    {
+      "epoch": 81.62,
+      "learning_rate": 5.8009615384615376e-05,
+      "loss": 0.6102,
+      "step": 8897
+    },
+    {
+      "epoch": 81.63,
+      "learning_rate": 5.798076923076923e-05,
+      "loss": 0.6171,
+      "step": 8898
+    },
+    {
+      "epoch": 81.64,
+      "learning_rate": 5.795192307692307e-05,
+      "loss": 0.6254,
+      "step": 8899
+    },
+    {
+      "epoch": 81.65,
+      "learning_rate": 5.7923076923076925e-05,
+      "loss": 0.5355,
+      "step": 8900
+    },
+    {
+      "epoch": 81.66,
+      "learning_rate": 5.789423076923076e-05,
+      "loss": 0.5458,
+      "step": 8901
+    },
+    {
+      "epoch": 81.67,
+      "learning_rate": 5.786538461538461e-05,
+      "loss": 0.5743,
+      "step": 8902
+    },
+    {
+      "epoch": 81.68,
+      "learning_rate": 5.783653846153846e-05,
+      "loss": 0.6438,
+      "step": 8903
+    },
+    {
+      "epoch": 81.69,
+      "learning_rate": 5.7807692307692304e-05,
+      "loss": 0.6746,
+      "step": 8904
+    },
+    {
+      "epoch": 81.7,
+      "learning_rate": 5.777884615384614e-05,
+      "loss": 0.4743,
+      "step": 8905
+    },
+    {
+      "epoch": 81.71,
+      "learning_rate": 5.7749999999999994e-05,
+      "loss": 0.5563,
+      "step": 8906
+    },
+    {
+      "epoch": 81.72,
+      "learning_rate": 5.772115384615384e-05,
+      "loss": 0.5307,
+      "step": 8907
+    },
+    {
+      "epoch": 81.72,
+      "learning_rate": 5.769230769230769e-05,
+      "loss": 0.6555,
+      "step": 8908
+    },
+    {
+      "epoch": 81.73,
+      "learning_rate": 5.7663461538461536e-05,
+      "loss": 0.6427,
+      "step": 8909
+    },
+    {
+      "epoch": 81.74,
+      "learning_rate": 5.7634615384615374e-05,
+      "loss": 0.4863,
+      "step": 8910
+    },
+    {
+      "epoch": 81.75,
+      "learning_rate": 5.7605769230769226e-05,
+      "loss": 0.6186,
+      "step": 8911
+    },
+    {
+      "epoch": 81.76,
+      "learning_rate": 5.757692307692307e-05,
+      "loss": 0.4974,
+      "step": 8912
+    },
+    {
+      "epoch": 81.77,
+      "learning_rate": 5.754807692307692e-05,
+      "loss": 0.5412,
+      "step": 8913
+    },
+    {
+      "epoch": 81.78,
+      "learning_rate": 5.751923076923077e-05,
+      "loss": 0.5713,
+      "step": 8914
+    },
+    {
+      "epoch": 81.79,
+      "learning_rate": 5.7490384615384605e-05,
+      "loss": 0.5204,
+      "step": 8915
+    },
+    {
+      "epoch": 81.8,
+      "learning_rate": 5.746153846153846e-05,
+      "loss": 0.6251,
+      "step": 8916
+    },
+    {
+      "epoch": 81.81,
+      "learning_rate": 5.74326923076923e-05,
+      "loss": 0.5968,
+      "step": 8917
+    },
+    {
+      "epoch": 81.82,
+      "learning_rate": 5.740384615384615e-05,
+      "loss": 0.6744,
+      "step": 8918
+    },
+    {
+      "epoch": 81.83,
+      "learning_rate": 5.7375e-05,
+      "loss": 0.4962,
+      "step": 8919
+    },
+    {
+      "epoch": 81.83,
+      "learning_rate": 5.734615384615384e-05,
+      "loss": 0.4971,
+      "step": 8920
+    },
+    {
+      "epoch": 81.84,
+      "learning_rate": 5.731730769230769e-05,
+      "loss": 0.5193,
+      "step": 8921
+    },
+    {
+      "epoch": 81.85,
+      "learning_rate": 5.7288461538461533e-05,
+      "loss": 0.5261,
+      "step": 8922
+    },
+    {
+      "epoch": 81.86,
+      "learning_rate": 5.725961538461538e-05,
+      "loss": 0.4087,
+      "step": 8923
+    },
+    {
+      "epoch": 81.87,
+      "learning_rate": 5.723076923076923e-05,
+      "loss": 0.5696,
+      "step": 8924
+    },
+    {
+      "epoch": 81.88,
+      "learning_rate": 5.720192307692307e-05,
+      "loss": 0.6075,
+      "step": 8925
+    },
+    {
+      "epoch": 81.89,
+      "learning_rate": 5.717307692307692e-05,
+      "loss": 0.5547,
+      "step": 8926
+    },
+    {
+      "epoch": 81.9,
+      "learning_rate": 5.7144230769230765e-05,
+      "loss": 0.4721,
+      "step": 8927
+    },
+    {
+      "epoch": 81.91,
+      "learning_rate": 5.711538461538461e-05,
+      "loss": 0.4958,
+      "step": 8928
+    },
+    {
+      "epoch": 81.92,
+      "learning_rate": 5.708653846153846e-05,
+      "loss": 0.5311,
+      "step": 8929
+    },
+    {
+      "epoch": 81.93,
+      "learning_rate": 5.70576923076923e-05,
+      "loss": 0.632,
+      "step": 8930
+    },
+    {
+      "epoch": 81.94,
+      "learning_rate": 5.702884615384615e-05,
+      "loss": 0.5856,
+      "step": 8931
+    },
+    {
+      "epoch": 81.94,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 0.5289,
+      "step": 8932
+    },
+    {
+      "epoch": 81.95,
+      "learning_rate": 5.697115384615384e-05,
+      "loss": 0.536,
+      "step": 8933
+    },
+    {
+      "epoch": 81.96,
+      "learning_rate": 5.694230769230769e-05,
+      "loss": 0.5889,
+      "step": 8934
+    },
+    {
+      "epoch": 81.97,
+      "learning_rate": 5.691346153846153e-05,
+      "loss": 0.6234,
+      "step": 8935
+    },
+    {
+      "epoch": 81.98,
+      "learning_rate": 5.6884615384615376e-05,
+      "loss": 0.5628,
+      "step": 8936
+    },
+    {
+      "epoch": 81.99,
+      "learning_rate": 5.685576923076923e-05,
+      "loss": 0.4275,
+      "step": 8937
+    },
+    {
+      "epoch": 82.0,
+      "learning_rate": 5.682692307692307e-05,
+      "loss": 0.4889,
+      "step": 8938
+    },
+    {
+      "epoch": 82.01,
+      "learning_rate": 5.6798076923076924e-05,
+      "loss": 0.6848,
+      "step": 8939
+    },
+    {
+      "epoch": 82.02,
+      "learning_rate": 5.676923076923076e-05,
+      "loss": 0.6016,
+      "step": 8940
+    },
+    {
+      "epoch": 82.03,
+      "learning_rate": 5.674038461538461e-05,
+      "loss": 0.5107,
+      "step": 8941
+    },
+    {
+      "epoch": 82.04,
+      "learning_rate": 5.671153846153846e-05,
+      "loss": 0.4519,
+      "step": 8942
+    },
+    {
+      "epoch": 82.05,
+      "learning_rate": 5.6682692307692304e-05,
+      "loss": 0.5645,
+      "step": 8943
+    },
+    {
+      "epoch": 82.06,
+      "learning_rate": 5.665384615384615e-05,
+      "loss": 0.6105,
+      "step": 8944
+    },
+    {
+      "epoch": 82.06,
+      "learning_rate": 5.6624999999999994e-05,
+      "loss": 0.6247,
+      "step": 8945
+    },
+    {
+      "epoch": 82.07,
+      "learning_rate": 5.659615384615384e-05,
+      "loss": 0.5611,
+      "step": 8946
+    },
+    {
+      "epoch": 82.08,
+      "learning_rate": 5.656730769230769e-05,
+      "loss": 0.6474,
+      "step": 8947
+    },
+    {
+      "epoch": 82.09,
+      "learning_rate": 5.6538461538461536e-05,
+      "loss": 0.5849,
+      "step": 8948
+    },
+    {
+      "epoch": 82.1,
+      "learning_rate": 5.650961538461538e-05,
+      "loss": 0.5877,
+      "step": 8949
+    },
+    {
+      "epoch": 82.11,
+      "learning_rate": 5.6480769230769225e-05,
+      "loss": 0.5053,
+      "step": 8950
+    },
+    {
+      "epoch": 82.12,
+      "learning_rate": 5.645192307692307e-05,
+      "loss": 0.4928,
+      "step": 8951
+    },
+    {
+      "epoch": 82.13,
+      "learning_rate": 5.642307692307692e-05,
+      "loss": 0.6787,
+      "step": 8952
+    },
+    {
+      "epoch": 82.14,
+      "learning_rate": 5.639423076923077e-05,
+      "loss": 0.5043,
+      "step": 8953
+    },
+    {
+      "epoch": 82.15,
+      "learning_rate": 5.6365384615384605e-05,
+      "loss": 0.5596,
+      "step": 8954
+    },
+    {
+      "epoch": 82.16,
+      "learning_rate": 5.633653846153846e-05,
+      "loss": 0.6549,
+      "step": 8955
+    },
+    {
+      "epoch": 82.17,
+      "learning_rate": 5.63076923076923e-05,
+      "loss": 0.4968,
+      "step": 8956
+    },
+    {
+      "epoch": 82.17,
+      "learning_rate": 5.6278846153846154e-05,
+      "loss": 0.4854,
+      "step": 8957
+    },
+    {
+      "epoch": 82.18,
+      "learning_rate": 5.625e-05,
+      "loss": 0.4989,
+      "step": 8958
+    },
+    {
+      "epoch": 82.19,
+      "learning_rate": 5.622115384615384e-05,
+      "loss": 0.524,
+      "step": 8959
+    },
+    {
+      "epoch": 82.2,
+      "learning_rate": 5.619230769230769e-05,
+      "loss": 0.6284,
+      "step": 8960
+    },
+    {
+      "epoch": 82.21,
+      "learning_rate": 5.616346153846153e-05,
+      "loss": 0.6916,
+      "step": 8961
+    },
+    {
+      "epoch": 82.22,
+      "learning_rate": 5.6134615384615385e-05,
+      "loss": 0.5859,
+      "step": 8962
+    },
+    {
+      "epoch": 82.23,
+      "learning_rate": 5.610576923076922e-05,
+      "loss": 0.4393,
+      "step": 8963
+    },
+    {
+      "epoch": 82.24,
+      "learning_rate": 5.607692307692307e-05,
+      "loss": 0.5862,
+      "step": 8964
+    },
+    {
+      "epoch": 82.25,
+      "learning_rate": 5.604807692307692e-05,
+      "loss": 0.5432,
+      "step": 8965
+    },
+    {
+      "epoch": 82.26,
+      "learning_rate": 5.6019230769230765e-05,
+      "loss": 0.6082,
+      "step": 8966
+    },
+    {
+      "epoch": 82.27,
+      "learning_rate": 5.599038461538461e-05,
+      "loss": 0.5442,
+      "step": 8967
+    },
+    {
+      "epoch": 82.28,
+      "learning_rate": 5.5961538461538455e-05,
+      "loss": 0.7122,
+      "step": 8968
+    },
+    {
+      "epoch": 82.28,
+      "learning_rate": 5.59326923076923e-05,
+      "loss": 0.6059,
+      "step": 8969
+    },
+    {
+      "epoch": 82.29,
+      "learning_rate": 5.590384615384615e-05,
+      "loss": 0.5208,
+      "step": 8970
+    },
+    {
+      "epoch": 82.3,
+      "learning_rate": 5.5874999999999996e-05,
+      "loss": 0.5928,
+      "step": 8971
+    },
+    {
+      "epoch": 82.31,
+      "learning_rate": 5.584615384615384e-05,
+      "loss": 0.5319,
+      "step": 8972
+    },
+    {
+      "epoch": 82.32,
+      "learning_rate": 5.5817307692307686e-05,
+      "loss": 0.7097,
+      "step": 8973
+    },
+    {
+      "epoch": 82.33,
+      "learning_rate": 5.578846153846153e-05,
+      "loss": 0.4482,
+      "step": 8974
+    },
+    {
+      "epoch": 82.34,
+      "learning_rate": 5.575961538461538e-05,
+      "loss": 0.6035,
+      "step": 8975
+    },
+    {
+      "epoch": 82.35,
+      "learning_rate": 5.573076923076923e-05,
+      "loss": 0.6336,
+      "step": 8976
+    },
+    {
+      "epoch": 82.36,
+      "learning_rate": 5.5701923076923066e-05,
+      "loss": 0.5275,
+      "step": 8977
+    },
+    {
+      "epoch": 82.37,
+      "learning_rate": 5.567307692307692e-05,
+      "loss": 0.552,
+      "step": 8978
+    },
+    {
+      "epoch": 82.38,
+      "learning_rate": 5.564423076923076e-05,
+      "loss": 0.4643,
+      "step": 8979
+    },
+    {
+      "epoch": 82.39,
+      "learning_rate": 5.5615384615384614e-05,
+      "loss": 0.6068,
+      "step": 8980
+    },
+    {
+      "epoch": 82.39,
+      "learning_rate": 5.558653846153846e-05,
+      "loss": 0.5454,
+      "step": 8981
+    },
+    {
+      "epoch": 82.4,
+      "learning_rate": 5.55576923076923e-05,
+      "loss": 0.5242,
+      "step": 8982
+    },
+    {
+      "epoch": 82.41,
+      "learning_rate": 5.552884615384615e-05,
+      "loss": 0.5965,
+      "step": 8983
+    },
+    {
+      "epoch": 82.42,
+      "learning_rate": 5.5499999999999994e-05,
+      "loss": 0.6193,
+      "step": 8984
+    },
+    {
+      "epoch": 82.43,
+      "learning_rate": 5.547115384615384e-05,
+      "loss": 0.5063,
+      "step": 8985
+    },
+    {
+      "epoch": 82.44,
+      "learning_rate": 5.544230769230769e-05,
+      "loss": 0.5202,
+      "step": 8986
+    },
+    {
+      "epoch": 82.45,
+      "learning_rate": 5.541346153846153e-05,
+      "loss": 0.6042,
+      "step": 8987
+    },
+    {
+      "epoch": 82.46,
+      "learning_rate": 5.538461538461538e-05,
+      "loss": 0.5657,
+      "step": 8988
+    },
+    {
+      "epoch": 82.47,
+      "learning_rate": 5.5355769230769225e-05,
+      "loss": 0.604,
+      "step": 8989
+    },
+    {
+      "epoch": 82.48,
+      "learning_rate": 5.532692307692307e-05,
+      "loss": 0.5704,
+      "step": 8990
+    },
+    {
+      "epoch": 82.49,
+      "learning_rate": 5.529807692307692e-05,
+      "loss": 0.5135,
+      "step": 8991
+    },
+    {
+      "epoch": 82.5,
+      "learning_rate": 5.526923076923076e-05,
+      "loss": 0.5558,
+      "step": 8992
+    },
+    {
+      "epoch": 82.5,
+      "learning_rate": 5.524038461538461e-05,
+      "loss": 0.6188,
+      "step": 8993
+    },
+    {
+      "epoch": 82.51,
+      "learning_rate": 5.521153846153846e-05,
+      "loss": 0.3868,
+      "step": 8994
+    },
+    {
+      "epoch": 82.52,
+      "learning_rate": 5.51826923076923e-05,
+      "loss": 0.6044,
+      "step": 8995
+    },
+    {
+      "epoch": 82.53,
+      "learning_rate": 5.515384615384615e-05,
+      "loss": 0.5624,
+      "step": 8996
+    },
+    {
+      "epoch": 82.54,
+      "learning_rate": 5.512499999999999e-05,
+      "loss": 0.4905,
+      "step": 8997
+    },
+    {
+      "epoch": 82.55,
+      "learning_rate": 5.509615384615384e-05,
+      "loss": 0.595,
+      "step": 8998
+    },
+    {
+      "epoch": 82.56,
+      "learning_rate": 5.506730769230769e-05,
+      "loss": 0.6031,
+      "step": 8999
+    },
+    {
+      "epoch": 82.57,
+      "learning_rate": 5.503846153846153e-05,
+      "loss": 0.5447,
+      "step": 9000
+    },
+    {
+      "epoch": 82.57,
+      "eval_cer": 0.08535360350230138,
+      "eval_loss": 0.36778390407562256,
+      "eval_runtime": 14.0359,
+      "eval_samples_per_second": 117.342,
+      "eval_steps_per_second": 1.852,
+      "eval_wer": 0.3533857624348892,
+      "step": 9000
+    },
+    {
+      "epoch": 82.58,
+      "learning_rate": 5.5009615384615385e-05,
+      "loss": 0.6522,
+      "step": 9001
+    },
+    {
+      "epoch": 82.59,
+      "learning_rate": 5.498076923076922e-05,
+      "loss": 0.4619,
+      "step": 9002
+    },
+    {
+      "epoch": 82.6,
+      "learning_rate": 5.495192307692307e-05,
+      "loss": 0.5566,
+      "step": 9003
+    },
+    {
+      "epoch": 82.61,
+      "learning_rate": 5.492307692307692e-05,
+      "loss": 0.6345,
+      "step": 9004
+    },
+    {
+      "epoch": 82.61,
+      "learning_rate": 5.4894230769230764e-05,
+      "loss": 0.5885,
+      "step": 9005
+    },
+    {
+      "epoch": 82.62,
+      "learning_rate": 5.4865384615384616e-05,
+      "loss": 0.5768,
+      "step": 9006
+    },
+    {
+      "epoch": 82.63,
+      "learning_rate": 5.4836538461538454e-05,
+      "loss": 0.5308,
+      "step": 9007
+    },
+    {
+      "epoch": 82.64,
+      "learning_rate": 5.48076923076923e-05,
+      "loss": 0.5525,
+      "step": 9008
+    },
+    {
+      "epoch": 82.65,
+      "learning_rate": 5.477884615384615e-05,
+      "loss": 0.5479,
+      "step": 9009
+    },
+    {
+      "epoch": 82.66,
+      "learning_rate": 5.4749999999999996e-05,
+      "loss": 0.4495,
+      "step": 9010
+    },
+    {
+      "epoch": 82.67,
+      "learning_rate": 5.472115384615385e-05,
+      "loss": 0.5731,
+      "step": 9011
+    },
+    {
+      "epoch": 82.68,
+      "learning_rate": 5.4692307692307686e-05,
+      "loss": 0.5291,
+      "step": 9012
+    },
+    {
+      "epoch": 82.69,
+      "learning_rate": 5.466346153846153e-05,
+      "loss": 0.4912,
+      "step": 9013
+    },
+    {
+      "epoch": 82.7,
+      "learning_rate": 5.463461538461538e-05,
+      "loss": 0.5903,
+      "step": 9014
+    },
+    {
+      "epoch": 82.71,
+      "learning_rate": 5.460576923076923e-05,
+      "loss": 0.5982,
+      "step": 9015
+    },
+    {
+      "epoch": 82.72,
+      "learning_rate": 5.4576923076923065e-05,
+      "loss": 0.4609,
+      "step": 9016
+    },
+    {
+      "epoch": 82.72,
+      "learning_rate": 5.454807692307692e-05,
+      "loss": 0.5762,
+      "step": 9017
+    },
+    {
+      "epoch": 82.73,
+      "learning_rate": 5.451923076923076e-05,
+      "loss": 0.5105,
+      "step": 9018
+    },
+    {
+      "epoch": 82.74,
+      "learning_rate": 5.4490384615384614e-05,
+      "loss": 0.6182,
+      "step": 9019
+    },
+    {
+      "epoch": 82.75,
+      "learning_rate": 5.446153846153846e-05,
+      "loss": 0.5451,
+      "step": 9020
+    },
+    {
+      "epoch": 82.76,
+      "learning_rate": 5.44326923076923e-05,
+      "loss": 0.5631,
+      "step": 9021
+    },
+    {
+      "epoch": 82.77,
+      "learning_rate": 5.440384615384615e-05,
+      "loss": 0.6002,
+      "step": 9022
+    },
+    {
+      "epoch": 82.78,
+      "learning_rate": 5.4374999999999994e-05,
+      "loss": 0.552,
+      "step": 9023
+    },
+    {
+      "epoch": 82.79,
+      "learning_rate": 5.4346153846153845e-05,
+      "loss": 0.5754,
+      "step": 9024
+    },
+    {
+      "epoch": 82.8,
+      "learning_rate": 5.431730769230769e-05,
+      "loss": 0.5442,
+      "step": 9025
+    },
+    {
+      "epoch": 82.81,
+      "learning_rate": 5.428846153846153e-05,
+      "loss": 0.4074,
+      "step": 9026
+    },
+    {
+      "epoch": 82.82,
+      "learning_rate": 5.425961538461538e-05,
+      "loss": 0.5805,
+      "step": 9027
+    },
+    {
+      "epoch": 82.83,
+      "learning_rate": 5.4230769230769225e-05,
+      "loss": 0.4435,
+      "step": 9028
+    },
+    {
+      "epoch": 82.83,
+      "learning_rate": 5.420192307692308e-05,
+      "loss": 0.5389,
+      "step": 9029
+    },
+    {
+      "epoch": 82.84,
+      "learning_rate": 5.417307692307692e-05,
+      "loss": 0.7087,
+      "step": 9030
+    },
+    {
+      "epoch": 82.85,
+      "learning_rate": 5.414423076923076e-05,
+      "loss": 0.6572,
+      "step": 9031
+    },
+    {
+      "epoch": 82.86,
+      "learning_rate": 5.411538461538461e-05,
+      "loss": 0.5383,
+      "step": 9032
+    },
+    {
+      "epoch": 82.87,
+      "learning_rate": 5.4086538461538456e-05,
+      "loss": 0.5199,
+      "step": 9033
+    },
+    {
+      "epoch": 82.88,
+      "learning_rate": 5.40576923076923e-05,
+      "loss": 0.516,
+      "step": 9034
+    },
+    {
+      "epoch": 82.89,
+      "learning_rate": 5.402884615384615e-05,
+      "loss": 0.4602,
+      "step": 9035
+    },
+    {
+      "epoch": 82.9,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 0.434,
+      "step": 9036
+    },
+    {
+      "epoch": 82.91,
+      "learning_rate": 5.397115384615384e-05,
+      "loss": 0.4868,
+      "step": 9037
+    },
+    {
+      "epoch": 82.92,
+      "learning_rate": 5.394230769230769e-05,
+      "loss": 0.4055,
+      "step": 9038
+    },
+    {
+      "epoch": 82.93,
+      "learning_rate": 5.391346153846153e-05,
+      "loss": 0.5059,
+      "step": 9039
+    },
+    {
+      "epoch": 82.94,
+      "learning_rate": 5.3884615384615384e-05,
+      "loss": 0.5892,
+      "step": 9040
+    },
+    {
+      "epoch": 82.94,
+      "learning_rate": 5.385576923076922e-05,
+      "loss": 0.439,
+      "step": 9041
+    },
+    {
+      "epoch": 82.95,
+      "learning_rate": 5.3826923076923074e-05,
+      "loss": 0.5381,
+      "step": 9042
+    },
+    {
+      "epoch": 82.96,
+      "learning_rate": 5.379807692307692e-05,
+      "loss": 0.5789,
+      "step": 9043
+    },
+    {
+      "epoch": 82.97,
+      "learning_rate": 5.3769230769230764e-05,
+      "loss": 0.4851,
+      "step": 9044
+    },
+    {
+      "epoch": 82.98,
+      "learning_rate": 5.3740384615384616e-05,
+      "loss": 0.5665,
+      "step": 9045
+    },
+    {
+      "epoch": 82.99,
+      "learning_rate": 5.3711538461538454e-05,
+      "loss": 0.5216,
+      "step": 9046
+    },
+    {
+      "epoch": 83.0,
+      "learning_rate": 5.3682692307692306e-05,
+      "loss": 0.5147,
+      "step": 9047
+    },
+    {
+      "epoch": 83.01,
+      "learning_rate": 5.365384615384615e-05,
+      "loss": 0.623,
+      "step": 9048
+    },
+    {
+      "epoch": 83.02,
+      "learning_rate": 5.3624999999999996e-05,
+      "loss": 0.5208,
+      "step": 9049
+    },
+    {
+      "epoch": 83.03,
+      "learning_rate": 5.359615384615385e-05,
+      "loss": 0.6174,
+      "step": 9050
+    },
+    {
+      "epoch": 83.04,
+      "learning_rate": 5.3567307692307686e-05,
+      "loss": 0.5022,
+      "step": 9051
+    },
+    {
+      "epoch": 83.05,
+      "learning_rate": 5.353846153846153e-05,
+      "loss": 0.7608,
+      "step": 9052
+    },
+    {
+      "epoch": 83.06,
+      "learning_rate": 5.350961538461538e-05,
+      "loss": 0.3913,
+      "step": 9053
+    },
+    {
+      "epoch": 83.06,
+      "learning_rate": 5.348076923076923e-05,
+      "loss": 0.5096,
+      "step": 9054
+    },
+    {
+      "epoch": 83.07,
+      "learning_rate": 5.345192307692308e-05,
+      "loss": 0.4809,
+      "step": 9055
+    },
+    {
+      "epoch": 83.08,
+      "learning_rate": 5.342307692307692e-05,
+      "loss": 0.5951,
+      "step": 9056
+    },
+    {
+      "epoch": 83.09,
+      "learning_rate": 5.339423076923076e-05,
+      "loss": 0.6312,
+      "step": 9057
+    },
+    {
+      "epoch": 83.1,
+      "learning_rate": 5.3365384615384614e-05,
+      "loss": 0.55,
+      "step": 9058
+    },
+    {
+      "epoch": 83.11,
+      "learning_rate": 5.333653846153846e-05,
+      "loss": 0.6264,
+      "step": 9059
+    },
+    {
+      "epoch": 83.12,
+      "learning_rate": 5.3307692307692303e-05,
+      "loss": 0.5751,
+      "step": 9060
+    },
+    {
+      "epoch": 83.13,
+      "learning_rate": 5.327884615384615e-05,
+      "loss": 0.5617,
+      "step": 9061
+    },
+    {
+      "epoch": 83.14,
+      "learning_rate": 5.324999999999999e-05,
+      "loss": 0.6398,
+      "step": 9062
+    },
+    {
+      "epoch": 83.15,
+      "learning_rate": 5.3221153846153845e-05,
+      "loss": 0.5811,
+      "step": 9063
+    },
+    {
+      "epoch": 83.16,
+      "learning_rate": 5.319230769230769e-05,
+      "loss": 0.5898,
+      "step": 9064
+    },
+    {
+      "epoch": 83.17,
+      "learning_rate": 5.316346153846153e-05,
+      "loss": 0.5202,
+      "step": 9065
+    },
+    {
+      "epoch": 83.17,
+      "learning_rate": 5.313461538461538e-05,
+      "loss": 0.5462,
+      "step": 9066
+    },
+    {
+      "epoch": 83.18,
+      "learning_rate": 5.3105769230769225e-05,
+      "loss": 0.651,
+      "step": 9067
+    },
+    {
+      "epoch": 83.19,
+      "learning_rate": 5.3076923076923076e-05,
+      "loss": 0.5005,
+      "step": 9068
+    },
+    {
+      "epoch": 83.2,
+      "learning_rate": 5.304807692307692e-05,
+      "loss": 0.5362,
+      "step": 9069
+    },
+    {
+      "epoch": 83.21,
+      "learning_rate": 5.301923076923076e-05,
+      "loss": 0.5298,
+      "step": 9070
+    },
+    {
+      "epoch": 83.22,
+      "learning_rate": 5.299038461538461e-05,
+      "loss": 0.6238,
+      "step": 9071
+    },
+    {
+      "epoch": 83.23,
+      "learning_rate": 5.2961538461538456e-05,
+      "loss": 0.5024,
+      "step": 9072
+    },
+    {
+      "epoch": 83.24,
+      "learning_rate": 5.293269230769231e-05,
+      "loss": 0.5305,
+      "step": 9073
+    },
+    {
+      "epoch": 83.25,
+      "learning_rate": 5.290384615384615e-05,
+      "loss": 0.4571,
+      "step": 9074
+    },
+    {
+      "epoch": 83.26,
+      "learning_rate": 5.287499999999999e-05,
+      "loss": 0.5067,
+      "step": 9075
+    },
+    {
+      "epoch": 83.27,
+      "learning_rate": 5.284615384615384e-05,
+      "loss": 0.5924,
+      "step": 9076
+    },
+    {
+      "epoch": 83.28,
+      "learning_rate": 5.281730769230769e-05,
+      "loss": 0.5935,
+      "step": 9077
+    },
+    {
+      "epoch": 83.28,
+      "learning_rate": 5.278846153846154e-05,
+      "loss": 0.5719,
+      "step": 9078
+    },
+    {
+      "epoch": 83.29,
+      "learning_rate": 5.275961538461538e-05,
+      "loss": 0.5127,
+      "step": 9079
+    },
+    {
+      "epoch": 83.3,
+      "learning_rate": 5.273076923076922e-05,
+      "loss": 0.5561,
+      "step": 9080
+    },
+    {
+      "epoch": 83.31,
+      "learning_rate": 5.2701923076923074e-05,
+      "loss": 0.5353,
+      "step": 9081
+    },
+    {
+      "epoch": 83.32,
+      "learning_rate": 5.267307692307692e-05,
+      "loss": 0.509,
+      "step": 9082
+    },
+    {
+      "epoch": 83.33,
+      "learning_rate": 5.2644230769230764e-05,
+      "loss": 0.5394,
+      "step": 9083
+    },
+    {
+      "epoch": 83.34,
+      "learning_rate": 5.261538461538461e-05,
+      "loss": 0.6529,
+      "step": 9084
+    },
+    {
+      "epoch": 83.35,
+      "learning_rate": 5.2586538461538454e-05,
+      "loss": 0.7497,
+      "step": 9085
+    },
+    {
+      "epoch": 83.36,
+      "learning_rate": 5.2557692307692306e-05,
+      "loss": 0.5838,
+      "step": 9086
+    },
+    {
+      "epoch": 83.37,
+      "learning_rate": 5.252884615384615e-05,
+      "loss": 0.5949,
+      "step": 9087
+    },
+    {
+      "epoch": 83.38,
+      "learning_rate": 5.2499999999999995e-05,
+      "loss": 0.5016,
+      "step": 9088
+    },
+    {
+      "epoch": 83.39,
+      "learning_rate": 5.247115384615384e-05,
+      "loss": 0.5199,
+      "step": 9089
+    },
+    {
+      "epoch": 83.39,
+      "learning_rate": 5.2442307692307685e-05,
+      "loss": 0.6104,
+      "step": 9090
+    },
+    {
+      "epoch": 83.4,
+      "learning_rate": 5.241346153846154e-05,
+      "loss": 0.4664,
+      "step": 9091
+    },
+    {
+      "epoch": 83.41,
+      "learning_rate": 5.238461538461538e-05,
+      "loss": 0.5834,
+      "step": 9092
+    },
+    {
+      "epoch": 83.42,
+      "learning_rate": 5.235576923076922e-05,
+      "loss": 0.595,
+      "step": 9093
+    },
+    {
+      "epoch": 83.43,
+      "learning_rate": 5.232692307692307e-05,
+      "loss": 0.4103,
+      "step": 9094
+    },
+    {
+      "epoch": 83.44,
+      "learning_rate": 5.229807692307692e-05,
+      "loss": 0.4888,
+      "step": 9095
+    },
+    {
+      "epoch": 83.45,
+      "learning_rate": 5.226923076923077e-05,
+      "loss": 0.6319,
+      "step": 9096
+    },
+    {
+      "epoch": 83.46,
+      "learning_rate": 5.224038461538461e-05,
+      "loss": 0.5963,
+      "step": 9097
+    },
+    {
+      "epoch": 83.47,
+      "learning_rate": 5.221153846153845e-05,
+      "loss": 0.6113,
+      "step": 9098
+    },
+    {
+      "epoch": 83.48,
+      "learning_rate": 5.21826923076923e-05,
+      "loss": 0.5777,
+      "step": 9099
+    },
+    {
+      "epoch": 83.49,
+      "learning_rate": 5.215384615384615e-05,
+      "loss": 0.5587,
+      "step": 9100
+    },
+    {
+      "epoch": 83.5,
+      "learning_rate": 5.212499999999999e-05,
+      "loss": 0.5399,
+      "step": 9101
+    },
+    {
+      "epoch": 83.5,
+      "learning_rate": 5.2096153846153845e-05,
+      "loss": 0.4972,
+      "step": 9102
+    },
+    {
+      "epoch": 83.51,
+      "learning_rate": 5.206730769230768e-05,
+      "loss": 0.6844,
+      "step": 9103
+    },
+    {
+      "epoch": 83.52,
+      "learning_rate": 5.2038461538461535e-05,
+      "loss": 0.5743,
+      "step": 9104
+    },
+    {
+      "epoch": 83.53,
+      "learning_rate": 5.200961538461538e-05,
+      "loss": 0.4948,
+      "step": 9105
+    },
+    {
+      "epoch": 83.54,
+      "learning_rate": 5.1980769230769224e-05,
+      "loss": 0.618,
+      "step": 9106
+    },
+    {
+      "epoch": 83.55,
+      "learning_rate": 5.1951923076923076e-05,
+      "loss": 0.5431,
+      "step": 9107
+    },
+    {
+      "epoch": 83.56,
+      "learning_rate": 5.1923076923076914e-05,
+      "loss": 0.6196,
+      "step": 9108
+    },
+    {
+      "epoch": 83.57,
+      "learning_rate": 5.1894230769230766e-05,
+      "loss": 0.4723,
+      "step": 9109
+    },
+    {
+      "epoch": 83.58,
+      "learning_rate": 5.186538461538461e-05,
+      "loss": 0.4636,
+      "step": 9110
+    },
+    {
+      "epoch": 83.59,
+      "learning_rate": 5.1836538461538456e-05,
+      "loss": 0.5408,
+      "step": 9111
+    },
+    {
+      "epoch": 83.6,
+      "learning_rate": 5.180769230769231e-05,
+      "loss": 0.4915,
+      "step": 9112
+    },
+    {
+      "epoch": 83.61,
+      "learning_rate": 5.1778846153846146e-05,
+      "loss": 0.4729,
+      "step": 9113
+    },
+    {
+      "epoch": 83.61,
+      "learning_rate": 5.174999999999999e-05,
+      "loss": 0.4412,
+      "step": 9114
+    },
+    {
+      "epoch": 83.62,
+      "learning_rate": 5.172115384615384e-05,
+      "loss": 0.5869,
+      "step": 9115
+    },
+    {
+      "epoch": 83.63,
+      "learning_rate": 5.169230769230769e-05,
+      "loss": 0.5052,
+      "step": 9116
+    },
+    {
+      "epoch": 83.64,
+      "learning_rate": 5.166346153846154e-05,
+      "loss": 0.5753,
+      "step": 9117
+    },
+    {
+      "epoch": 83.65,
+      "learning_rate": 5.163461538461538e-05,
+      "loss": 0.5716,
+      "step": 9118
+    },
+    {
+      "epoch": 83.66,
+      "learning_rate": 5.160576923076922e-05,
+      "loss": 0.5558,
+      "step": 9119
+    },
+    {
+      "epoch": 83.67,
+      "learning_rate": 5.1576923076923074e-05,
+      "loss": 0.5455,
+      "step": 9120
+    },
+    {
+      "epoch": 83.68,
+      "learning_rate": 5.154807692307692e-05,
+      "loss": 0.485,
+      "step": 9121
+    },
+    {
+      "epoch": 83.69,
+      "learning_rate": 5.151923076923077e-05,
+      "loss": 0.4971,
+      "step": 9122
+    },
+    {
+      "epoch": 83.7,
+      "learning_rate": 5.149038461538461e-05,
+      "loss": 0.5206,
+      "step": 9123
+    },
+    {
+      "epoch": 83.71,
+      "learning_rate": 5.1461538461538454e-05,
+      "loss": 0.5778,
+      "step": 9124
+    },
+    {
+      "epoch": 83.72,
+      "learning_rate": 5.1432692307692305e-05,
+      "loss": 0.5141,
+      "step": 9125
+    },
+    {
+      "epoch": 83.72,
+      "learning_rate": 5.140384615384615e-05,
+      "loss": 0.5852,
+      "step": 9126
+    },
+    {
+      "epoch": 83.73,
+      "learning_rate": 5.1375e-05,
+      "loss": 0.5204,
+      "step": 9127
+    },
+    {
+      "epoch": 83.74,
+      "learning_rate": 5.134615384615384e-05,
+      "loss": 0.5581,
+      "step": 9128
+    },
+    {
+      "epoch": 83.75,
+      "learning_rate": 5.1317307692307685e-05,
+      "loss": 0.6206,
+      "step": 9129
+    },
+    {
+      "epoch": 83.76,
+      "learning_rate": 5.128846153846154e-05,
+      "loss": 0.5469,
+      "step": 9130
+    },
+    {
+      "epoch": 83.77,
+      "learning_rate": 5.125961538461538e-05,
+      "loss": 0.4016,
+      "step": 9131
+    },
+    {
+      "epoch": 83.78,
+      "learning_rate": 5.123076923076922e-05,
+      "loss": 0.4891,
+      "step": 9132
+    },
+    {
+      "epoch": 83.79,
+      "learning_rate": 5.120192307692307e-05,
+      "loss": 0.487,
+      "step": 9133
+    },
+    {
+      "epoch": 83.8,
+      "learning_rate": 5.1173076923076916e-05,
+      "loss": 0.5901,
+      "step": 9134
+    },
+    {
+      "epoch": 83.81,
+      "learning_rate": 5.114423076923077e-05,
+      "loss": 0.5367,
+      "step": 9135
+    },
+    {
+      "epoch": 83.82,
+      "learning_rate": 5.111538461538461e-05,
+      "loss": 0.5997,
+      "step": 9136
+    },
+    {
+      "epoch": 83.83,
+      "learning_rate": 5.108653846153845e-05,
+      "loss": 0.5642,
+      "step": 9137
+    },
+    {
+      "epoch": 83.83,
+      "learning_rate": 5.10576923076923e-05,
+      "loss": 0.5457,
+      "step": 9138
+    },
+    {
+      "epoch": 83.84,
+      "learning_rate": 5.102884615384615e-05,
+      "loss": 0.471,
+      "step": 9139
+    },
+    {
+      "epoch": 83.85,
+      "learning_rate": 5.1e-05,
+      "loss": 0.5453,
+      "step": 9140
+    },
+    {
+      "epoch": 83.86,
+      "learning_rate": 5.0971153846153845e-05,
+      "loss": 0.6849,
+      "step": 9141
+    },
+    {
+      "epoch": 83.87,
+      "learning_rate": 5.094230769230768e-05,
+      "loss": 0.6593,
+      "step": 9142
+    },
+    {
+      "epoch": 83.88,
+      "learning_rate": 5.0913461538461534e-05,
+      "loss": 0.5216,
+      "step": 9143
+    },
+    {
+      "epoch": 83.89,
+      "learning_rate": 5.088461538461538e-05,
+      "loss": 0.5939,
+      "step": 9144
+    },
+    {
+      "epoch": 83.9,
+      "learning_rate": 5.0855769230769224e-05,
+      "loss": 0.57,
+      "step": 9145
+    },
+    {
+      "epoch": 83.91,
+      "learning_rate": 5.0826923076923076e-05,
+      "loss": 0.5209,
+      "step": 9146
+    },
+    {
+      "epoch": 83.92,
+      "learning_rate": 5.0798076923076914e-05,
+      "loss": 0.5664,
+      "step": 9147
+    },
+    {
+      "epoch": 83.93,
+      "learning_rate": 5.0769230769230766e-05,
+      "loss": 0.5867,
+      "step": 9148
+    },
+    {
+      "epoch": 83.94,
+      "learning_rate": 5.074038461538461e-05,
+      "loss": 0.4647,
+      "step": 9149
+    },
+    {
+      "epoch": 83.94,
+      "learning_rate": 5.0711538461538456e-05,
+      "loss": 0.4501,
+      "step": 9150
+    },
+    {
+      "epoch": 83.95,
+      "learning_rate": 5.068269230769231e-05,
+      "loss": 0.616,
+      "step": 9151
+    },
+    {
+      "epoch": 83.96,
+      "learning_rate": 5.0653846153846146e-05,
+      "loss": 0.4869,
+      "step": 9152
+    },
+    {
+      "epoch": 83.97,
+      "learning_rate": 5.0625e-05,
+      "loss": 0.5469,
+      "step": 9153
+    },
+    {
+      "epoch": 83.98,
+      "learning_rate": 5.059615384615384e-05,
+      "loss": 0.423,
+      "step": 9154
+    },
+    {
+      "epoch": 83.99,
+      "learning_rate": 5.056730769230769e-05,
+      "loss": 0.5744,
+      "step": 9155
+    },
+    {
+      "epoch": 84.0,
+      "learning_rate": 5.053846153846154e-05,
+      "loss": 0.5875,
+      "step": 9156
+    },
+    {
+      "epoch": 84.01,
+      "learning_rate": 5.050961538461538e-05,
+      "loss": 0.5723,
+      "step": 9157
+    },
+    {
+      "epoch": 84.02,
+      "learning_rate": 5.048076923076923e-05,
+      "loss": 0.5452,
+      "step": 9158
+    },
+    {
+      "epoch": 84.03,
+      "learning_rate": 5.0451923076923074e-05,
+      "loss": 0.4854,
+      "step": 9159
+    },
+    {
+      "epoch": 84.04,
+      "learning_rate": 5.042307692307692e-05,
+      "loss": 0.5835,
+      "step": 9160
+    },
+    {
+      "epoch": 84.05,
+      "learning_rate": 5.039423076923077e-05,
+      "loss": 0.7142,
+      "step": 9161
+    },
+    {
+      "epoch": 84.06,
+      "learning_rate": 5.036538461538461e-05,
+      "loss": 0.4625,
+      "step": 9162
+    },
+    {
+      "epoch": 84.06,
+      "learning_rate": 5.033653846153845e-05,
+      "loss": 0.587,
+      "step": 9163
+    },
+    {
+      "epoch": 84.07,
+      "learning_rate": 5.0307692307692305e-05,
+      "loss": 0.6176,
+      "step": 9164
+    },
+    {
+      "epoch": 84.08,
+      "learning_rate": 5.027884615384615e-05,
+      "loss": 0.4835,
+      "step": 9165
+    },
+    {
+      "epoch": 84.09,
+      "learning_rate": 5.025e-05,
+      "loss": 0.479,
+      "step": 9166
+    },
+    {
+      "epoch": 84.1,
+      "learning_rate": 5.022115384615384e-05,
+      "loss": 0.4606,
+      "step": 9167
+    },
+    {
+      "epoch": 84.11,
+      "learning_rate": 5.0192307692307685e-05,
+      "loss": 0.593,
+      "step": 9168
+    },
+    {
+      "epoch": 84.12,
+      "learning_rate": 5.0163461538461536e-05,
+      "loss": 0.6489,
+      "step": 9169
+    },
+    {
+      "epoch": 84.13,
+      "learning_rate": 5.013461538461538e-05,
+      "loss": 0.6311,
+      "step": 9170
+    },
+    {
+      "epoch": 84.14,
+      "learning_rate": 5.010576923076923e-05,
+      "loss": 0.5133,
+      "step": 9171
+    },
+    {
+      "epoch": 84.15,
+      "learning_rate": 5.007692307692307e-05,
+      "loss": 0.511,
+      "step": 9172
+    },
+    {
+      "epoch": 84.16,
+      "learning_rate": 5.0048076923076916e-05,
+      "loss": 0.5394,
+      "step": 9173
+    },
+    {
+      "epoch": 84.17,
+      "learning_rate": 5.001923076923077e-05,
+      "loss": 0.4992,
+      "step": 9174
+    },
+    {
+      "epoch": 84.17,
+      "learning_rate": 4.999038461538461e-05,
+      "loss": 0.5437,
+      "step": 9175
+    },
+    {
+      "epoch": 84.18,
+      "learning_rate": 4.9961538461538465e-05,
+      "loss": 0.5264,
+      "step": 9176
+    },
+    {
+      "epoch": 84.19,
+      "learning_rate": 4.99326923076923e-05,
+      "loss": 0.5801,
+      "step": 9177
+    },
+    {
+      "epoch": 84.2,
+      "learning_rate": 4.990384615384615e-05,
+      "loss": 0.7169,
+      "step": 9178
+    },
+    {
+      "epoch": 84.21,
+      "learning_rate": 4.9875e-05,
+      "loss": 0.4813,
+      "step": 9179
+    },
+    {
+      "epoch": 84.22,
+      "learning_rate": 4.9846153846153844e-05,
+      "loss": 0.5893,
+      "step": 9180
+    },
+    {
+      "epoch": 84.23,
+      "learning_rate": 4.981730769230768e-05,
+      "loss": 0.3554,
+      "step": 9181
+    },
+    {
+      "epoch": 84.24,
+      "learning_rate": 4.9788461538461534e-05,
+      "loss": 0.5425,
+      "step": 9182
+    },
+    {
+      "epoch": 84.25,
+      "learning_rate": 4.975961538461538e-05,
+      "loss": 0.5779,
+      "step": 9183
+    },
+    {
+      "epoch": 84.26,
+      "learning_rate": 4.973076923076923e-05,
+      "loss": 0.4263,
+      "step": 9184
+    },
+    {
+      "epoch": 84.27,
+      "learning_rate": 4.9701923076923076e-05,
+      "loss": 0.5452,
+      "step": 9185
+    },
+    {
+      "epoch": 84.28,
+      "learning_rate": 4.9673076923076914e-05,
+      "loss": 0.5474,
+      "step": 9186
+    },
+    {
+      "epoch": 84.28,
+      "learning_rate": 4.9644230769230766e-05,
+      "loss": 0.4919,
+      "step": 9187
+    },
+    {
+      "epoch": 84.29,
+      "learning_rate": 4.961538461538461e-05,
+      "loss": 0.5339,
+      "step": 9188
+    },
+    {
+      "epoch": 84.3,
+      "learning_rate": 4.958653846153846e-05,
+      "loss": 0.5677,
+      "step": 9189
+    },
+    {
+      "epoch": 84.31,
+      "learning_rate": 4.955769230769231e-05,
+      "loss": 0.6365,
+      "step": 9190
+    },
+    {
+      "epoch": 84.32,
+      "learning_rate": 4.9528846153846145e-05,
+      "loss": 0.6009,
+      "step": 9191
+    },
+    {
+      "epoch": 84.33,
+      "learning_rate": 4.95e-05,
+      "loss": 0.4832,
+      "step": 9192
+    },
+    {
+      "epoch": 84.34,
+      "learning_rate": 4.947115384615384e-05,
+      "loss": 0.546,
+      "step": 9193
+    },
+    {
+      "epoch": 84.35,
+      "learning_rate": 4.944230769230769e-05,
+      "loss": 0.5408,
+      "step": 9194
+    },
+    {
+      "epoch": 84.36,
+      "learning_rate": 4.941346153846153e-05,
+      "loss": 0.5804,
+      "step": 9195
+    },
+    {
+      "epoch": 84.37,
+      "learning_rate": 4.938461538461538e-05,
+      "loss": 0.48,
+      "step": 9196
+    },
+    {
+      "epoch": 84.38,
+      "learning_rate": 4.935576923076923e-05,
+      "loss": 0.5871,
+      "step": 9197
+    },
+    {
+      "epoch": 84.39,
+      "learning_rate": 4.932692307692307e-05,
+      "loss": 0.5494,
+      "step": 9198
+    },
+    {
+      "epoch": 84.39,
+      "learning_rate": 4.929807692307692e-05,
+      "loss": 0.4737,
+      "step": 9199
+    },
+    {
+      "epoch": 84.4,
+      "learning_rate": 4.926923076923076e-05,
+      "loss": 0.5627,
+      "step": 9200
+    },
+    {
+      "epoch": 84.41,
+      "learning_rate": 4.924038461538461e-05,
+      "loss": 0.4801,
+      "step": 9201
+    },
+    {
+      "epoch": 84.42,
+      "learning_rate": 4.921153846153846e-05,
+      "loss": 0.5188,
+      "step": 9202
+    },
+    {
+      "epoch": 84.43,
+      "learning_rate": 4.9182692307692305e-05,
+      "loss": 0.5668,
+      "step": 9203
+    },
+    {
+      "epoch": 84.44,
+      "learning_rate": 4.915384615384615e-05,
+      "loss": 0.6637,
+      "step": 9204
+    },
+    {
+      "epoch": 84.45,
+      "learning_rate": 4.9124999999999995e-05,
+      "loss": 0.6761,
+      "step": 9205
+    },
+    {
+      "epoch": 84.46,
+      "learning_rate": 4.909615384615384e-05,
+      "loss": 0.6587,
+      "step": 9206
+    },
+    {
+      "epoch": 84.47,
+      "learning_rate": 4.906730769230769e-05,
+      "loss": 0.6163,
+      "step": 9207
+    },
+    {
+      "epoch": 84.48,
+      "learning_rate": 4.9038461538461536e-05,
+      "loss": 0.5142,
+      "step": 9208
+    },
+    {
+      "epoch": 84.49,
+      "learning_rate": 4.9009615384615374e-05,
+      "loss": 0.5809,
+      "step": 9209
+    },
+    {
+      "epoch": 84.5,
+      "learning_rate": 4.8980769230769226e-05,
+      "loss": 0.6901,
+      "step": 9210
+    },
+    {
+      "epoch": 84.5,
+      "learning_rate": 4.895192307692307e-05,
+      "loss": 0.5366,
+      "step": 9211
+    },
+    {
+      "epoch": 84.51,
+      "learning_rate": 4.8923076923076916e-05,
+      "loss": 0.6051,
+      "step": 9212
+    },
+    {
+      "epoch": 84.52,
+      "learning_rate": 4.889423076923077e-05,
+      "loss": 0.5045,
+      "step": 9213
+    },
+    {
+      "epoch": 84.53,
+      "learning_rate": 4.8865384615384606e-05,
+      "loss": 0.5048,
+      "step": 9214
+    },
+    {
+      "epoch": 84.54,
+      "learning_rate": 4.883653846153846e-05,
+      "loss": 0.5568,
+      "step": 9215
+    },
+    {
+      "epoch": 84.55,
+      "learning_rate": 4.88076923076923e-05,
+      "loss": 0.5786,
+      "step": 9216
+    },
+    {
+      "epoch": 84.56,
+      "learning_rate": 4.877884615384615e-05,
+      "loss": 0.5165,
+      "step": 9217
+    },
+    {
+      "epoch": 84.57,
+      "learning_rate": 4.875e-05,
+      "loss": 0.5949,
+      "step": 9218
+    },
+    {
+      "epoch": 84.58,
+      "learning_rate": 4.872115384615384e-05,
+      "loss": 0.5802,
+      "step": 9219
+    },
+    {
+      "epoch": 84.59,
+      "learning_rate": 4.869230769230769e-05,
+      "loss": 0.6861,
+      "step": 9220
+    },
+    {
+      "epoch": 84.6,
+      "learning_rate": 4.8663461538461534e-05,
+      "loss": 0.5418,
+      "step": 9221
+    },
+    {
+      "epoch": 84.61,
+      "learning_rate": 4.863461538461538e-05,
+      "loss": 0.5413,
+      "step": 9222
+    },
+    {
+      "epoch": 84.61,
+      "learning_rate": 4.860576923076923e-05,
+      "loss": 0.6161,
+      "step": 9223
+    },
+    {
+      "epoch": 84.62,
+      "learning_rate": 4.857692307692307e-05,
+      "loss": 0.5728,
+      "step": 9224
+    },
+    {
+      "epoch": 84.63,
+      "learning_rate": 4.854807692307692e-05,
+      "loss": 0.7381,
+      "step": 9225
+    },
+    {
+      "epoch": 84.64,
+      "learning_rate": 4.8519230769230765e-05,
+      "loss": 0.4472,
+      "step": 9226
+    },
+    {
+      "epoch": 84.65,
+      "learning_rate": 4.849038461538461e-05,
+      "loss": 0.4734,
+      "step": 9227
+    },
+    {
+      "epoch": 84.66,
+      "learning_rate": 4.846153846153846e-05,
+      "loss": 0.502,
+      "step": 9228
+    },
+    {
+      "epoch": 84.67,
+      "learning_rate": 4.84326923076923e-05,
+      "loss": 0.6664,
+      "step": 9229
+    },
+    {
+      "epoch": 84.68,
+      "learning_rate": 4.8403846153846145e-05,
+      "loss": 0.4897,
+      "step": 9230
+    },
+    {
+      "epoch": 84.69,
+      "learning_rate": 4.8375e-05,
+      "loss": 0.5491,
+      "step": 9231
+    },
+    {
+      "epoch": 84.7,
+      "learning_rate": 4.834615384615384e-05,
+      "loss": 0.5884,
+      "step": 9232
+    },
+    {
+      "epoch": 84.71,
+      "learning_rate": 4.8317307692307693e-05,
+      "loss": 0.6373,
+      "step": 9233
+    },
+    {
+      "epoch": 84.72,
+      "learning_rate": 4.828846153846153e-05,
+      "loss": 0.4974,
+      "step": 9234
+    },
+    {
+      "epoch": 84.72,
+      "learning_rate": 4.8259615384615377e-05,
+      "loss": 0.4784,
+      "step": 9235
+    },
+    {
+      "epoch": 84.73,
+      "learning_rate": 4.823076923076923e-05,
+      "loss": 0.4507,
+      "step": 9236
+    },
+    {
+      "epoch": 84.74,
+      "learning_rate": 4.820192307692307e-05,
+      "loss": 0.7605,
+      "step": 9237
+    },
+    {
+      "epoch": 84.75,
+      "learning_rate": 4.8173076923076925e-05,
+      "loss": 0.5467,
+      "step": 9238
+    },
+    {
+      "epoch": 84.76,
+      "learning_rate": 4.814423076923076e-05,
+      "loss": 0.4678,
+      "step": 9239
+    },
+    {
+      "epoch": 84.77,
+      "learning_rate": 4.811538461538461e-05,
+      "loss": 0.5247,
+      "step": 9240
+    },
+    {
+      "epoch": 84.78,
+      "learning_rate": 4.808653846153846e-05,
+      "loss": 0.5767,
+      "step": 9241
+    },
+    {
+      "epoch": 84.79,
+      "learning_rate": 4.8057692307692305e-05,
+      "loss": 0.6357,
+      "step": 9242
+    },
+    {
+      "epoch": 84.8,
+      "learning_rate": 4.802884615384614e-05,
+      "loss": 0.4294,
+      "step": 9243
+    },
+    {
+      "epoch": 84.81,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 0.4648,
+      "step": 9244
+    },
+    {
+      "epoch": 84.82,
+      "learning_rate": 4.797115384615384e-05,
+      "loss": 0.5179,
+      "step": 9245
+    },
+    {
+      "epoch": 84.83,
+      "learning_rate": 4.794230769230769e-05,
+      "loss": 0.5454,
+      "step": 9246
+    },
+    {
+      "epoch": 84.83,
+      "learning_rate": 4.7913461538461536e-05,
+      "loss": 0.4643,
+      "step": 9247
+    },
+    {
+      "epoch": 84.84,
+      "learning_rate": 4.7884615384615374e-05,
+      "loss": 0.6675,
+      "step": 9248
+    },
+    {
+      "epoch": 84.85,
+      "learning_rate": 4.7855769230769226e-05,
+      "loss": 0.6976,
+      "step": 9249
+    },
+    {
+      "epoch": 84.86,
+      "learning_rate": 4.782692307692307e-05,
+      "loss": 0.4238,
+      "step": 9250
+    },
+    {
+      "epoch": 84.87,
+      "learning_rate": 4.779807692307692e-05,
+      "loss": 0.601,
+      "step": 9251
+    },
+    {
+      "epoch": 84.88,
+      "learning_rate": 4.776923076923077e-05,
+      "loss": 0.5785,
+      "step": 9252
+    },
+    {
+      "epoch": 84.89,
+      "learning_rate": 4.7740384615384606e-05,
+      "loss": 0.476,
+      "step": 9253
+    },
+    {
+      "epoch": 84.9,
+      "learning_rate": 4.771153846153846e-05,
+      "loss": 0.6794,
+      "step": 9254
+    },
+    {
+      "epoch": 84.91,
+      "learning_rate": 4.76826923076923e-05,
+      "loss": 0.5549,
+      "step": 9255
+    },
+    {
+      "epoch": 84.92,
+      "learning_rate": 4.7653846153846154e-05,
+      "loss": 0.4361,
+      "step": 9256
+    },
+    {
+      "epoch": 84.93,
+      "learning_rate": 4.7625e-05,
+      "loss": 0.6338,
+      "step": 9257
+    },
+    {
+      "epoch": 84.94,
+      "learning_rate": 4.759615384615384e-05,
+      "loss": 0.5812,
+      "step": 9258
+    },
+    {
+      "epoch": 84.94,
+      "learning_rate": 4.756730769230769e-05,
+      "loss": 0.6766,
+      "step": 9259
+    },
+    {
+      "epoch": 84.95,
+      "learning_rate": 4.7538461538461534e-05,
+      "loss": 0.5393,
+      "step": 9260
+    },
+    {
+      "epoch": 84.96,
+      "learning_rate": 4.750961538461538e-05,
+      "loss": 0.4949,
+      "step": 9261
+    },
+    {
+      "epoch": 84.97,
+      "learning_rate": 4.748076923076923e-05,
+      "loss": 0.5363,
+      "step": 9262
+    },
+    {
+      "epoch": 84.98,
+      "learning_rate": 4.745192307692307e-05,
+      "loss": 0.4521,
+      "step": 9263
+    },
+    {
+      "epoch": 84.99,
+      "learning_rate": 4.742307692307692e-05,
+      "loss": 0.6553,
+      "step": 9264
+    },
+    {
+      "epoch": 85.0,
+      "learning_rate": 4.7394230769230765e-05,
+      "loss": 0.5566,
+      "step": 9265
+    },
+    {
+      "epoch": 85.01,
+      "learning_rate": 4.736538461538461e-05,
+      "loss": 0.6206,
+      "step": 9266
+    },
+    {
+      "epoch": 85.02,
+      "learning_rate": 4.733653846153846e-05,
+      "loss": 0.5158,
+      "step": 9267
+    },
+    {
+      "epoch": 85.03,
+      "learning_rate": 4.73076923076923e-05,
+      "loss": 0.487,
+      "step": 9268
+    },
+    {
+      "epoch": 85.04,
+      "learning_rate": 4.727884615384615e-05,
+      "loss": 0.4611,
+      "step": 9269
+    },
+    {
+      "epoch": 85.05,
+      "learning_rate": 4.7249999999999997e-05,
+      "loss": 0.6049,
+      "step": 9270
+    },
+    {
+      "epoch": 85.06,
+      "learning_rate": 4.722115384615384e-05,
+      "loss": 0.3929,
+      "step": 9271
+    },
+    {
+      "epoch": 85.06,
+      "learning_rate": 4.719230769230769e-05,
+      "loss": 0.6151,
+      "step": 9272
+    },
+    {
+      "epoch": 85.07,
+      "learning_rate": 4.716346153846153e-05,
+      "loss": 0.5507,
+      "step": 9273
+    },
+    {
+      "epoch": 85.08,
+      "learning_rate": 4.713461538461538e-05,
+      "loss": 0.4838,
+      "step": 9274
+    },
+    {
+      "epoch": 85.09,
+      "learning_rate": 4.710576923076923e-05,
+      "loss": 0.476,
+      "step": 9275
+    },
+    {
+      "epoch": 85.1,
+      "learning_rate": 4.707692307692307e-05,
+      "loss": 0.541,
+      "step": 9276
+    },
+    {
+      "epoch": 85.11,
+      "learning_rate": 4.7048076923076925e-05,
+      "loss": 0.6323,
+      "step": 9277
+    },
+    {
+      "epoch": 85.12,
+      "learning_rate": 4.701923076923076e-05,
+      "loss": 0.4707,
+      "step": 9278
+    },
+    {
+      "epoch": 85.13,
+      "learning_rate": 4.699038461538461e-05,
+      "loss": 0.4701,
+      "step": 9279
+    },
+    {
+      "epoch": 85.14,
+      "learning_rate": 4.696153846153846e-05,
+      "loss": 0.3799,
+      "step": 9280
+    },
+    {
+      "epoch": 85.15,
+      "learning_rate": 4.6932692307692304e-05,
+      "loss": 0.4323,
+      "step": 9281
+    },
+    {
+      "epoch": 85.16,
+      "learning_rate": 4.6903846153846156e-05,
+      "loss": 0.4866,
+      "step": 9282
+    },
+    {
+      "epoch": 85.17,
+      "learning_rate": 4.6874999999999994e-05,
+      "loss": 0.4377,
+      "step": 9283
+    },
+    {
+      "epoch": 85.17,
+      "learning_rate": 4.684615384615384e-05,
+      "loss": 0.472,
+      "step": 9284
+    },
+    {
+      "epoch": 85.18,
+      "learning_rate": 4.681730769230769e-05,
+      "loss": 0.4949,
+      "step": 9285
+    },
+    {
+      "epoch": 85.19,
+      "learning_rate": 4.6788461538461536e-05,
+      "loss": 0.431,
+      "step": 9286
+    },
+    {
+      "epoch": 85.2,
+      "learning_rate": 4.675961538461539e-05,
+      "loss": 0.5126,
+      "step": 9287
+    },
+    {
+      "epoch": 85.21,
+      "learning_rate": 4.6730769230769226e-05,
+      "loss": 0.487,
+      "step": 9288
+    },
+    {
+      "epoch": 85.22,
+      "learning_rate": 4.670192307692307e-05,
+      "loss": 0.4878,
+      "step": 9289
+    },
+    {
+      "epoch": 85.23,
+      "learning_rate": 4.667307692307692e-05,
+      "loss": 0.453,
+      "step": 9290
+    },
+    {
+      "epoch": 85.24,
+      "learning_rate": 4.664423076923077e-05,
+      "loss": 0.5082,
+      "step": 9291
+    },
+    {
+      "epoch": 85.25,
+      "learning_rate": 4.6615384615384605e-05,
+      "loss": 0.5831,
+      "step": 9292
+    },
+    {
+      "epoch": 85.26,
+      "learning_rate": 4.658653846153846e-05,
+      "loss": 0.5173,
+      "step": 9293
+    },
+    {
+      "epoch": 85.27,
+      "learning_rate": 4.65576923076923e-05,
+      "loss": 0.494,
+      "step": 9294
+    },
+    {
+      "epoch": 85.28,
+      "learning_rate": 4.6528846153846154e-05,
+      "loss": 0.597,
+      "step": 9295
+    },
+    {
+      "epoch": 85.28,
+      "learning_rate": 4.65e-05,
+      "loss": 0.4719,
+      "step": 9296
+    },
+    {
+      "epoch": 85.29,
+      "learning_rate": 4.647115384615384e-05,
+      "loss": 0.5548,
+      "step": 9297
+    },
+    {
+      "epoch": 85.3,
+      "learning_rate": 4.644230769230769e-05,
+      "loss": 0.5719,
+      "step": 9298
+    },
+    {
+      "epoch": 85.31,
+      "learning_rate": 4.6413461538461533e-05,
+      "loss": 0.4842,
+      "step": 9299
+    },
+    {
+      "epoch": 85.32,
+      "learning_rate": 4.6384615384615385e-05,
+      "loss": 0.4506,
+      "step": 9300
+    },
+    {
+      "epoch": 85.33,
+      "learning_rate": 4.635576923076923e-05,
+      "loss": 0.5629,
+      "step": 9301
+    },
+    {
+      "epoch": 85.34,
+      "learning_rate": 4.632692307692307e-05,
+      "loss": 0.6406,
+      "step": 9302
+    },
+    {
+      "epoch": 85.35,
+      "learning_rate": 4.629807692307692e-05,
+      "loss": 0.4961,
+      "step": 9303
+    },
+    {
+      "epoch": 85.36,
+      "learning_rate": 4.6269230769230765e-05,
+      "loss": 0.6179,
+      "step": 9304
+    },
+    {
+      "epoch": 85.37,
+      "learning_rate": 4.6240384615384617e-05,
+      "loss": 0.5625,
+      "step": 9305
+    },
+    {
+      "epoch": 85.38,
+      "learning_rate": 4.621153846153846e-05,
+      "loss": 0.5949,
+      "step": 9306
+    },
+    {
+      "epoch": 85.39,
+      "learning_rate": 4.61826923076923e-05,
+      "loss": 0.5377,
+      "step": 9307
+    },
+    {
+      "epoch": 85.39,
+      "learning_rate": 4.615384615384615e-05,
+      "loss": 0.5413,
+      "step": 9308
+    },
+    {
+      "epoch": 85.4,
+      "learning_rate": 4.6124999999999996e-05,
+      "loss": 0.4317,
+      "step": 9309
+    },
+    {
+      "epoch": 85.41,
+      "learning_rate": 4.609615384615384e-05,
+      "loss": 0.6547,
+      "step": 9310
+    },
+    {
+      "epoch": 85.42,
+      "learning_rate": 4.6067307692307686e-05,
+      "loss": 0.6023,
+      "step": 9311
+    },
+    {
+      "epoch": 85.43,
+      "learning_rate": 4.603846153846153e-05,
+      "loss": 0.5867,
+      "step": 9312
+    },
+    {
+      "epoch": 85.44,
+      "learning_rate": 4.600961538461538e-05,
+      "loss": 0.4534,
+      "step": 9313
+    },
+    {
+      "epoch": 85.45,
+      "learning_rate": 4.598076923076923e-05,
+      "loss": 0.5559,
+      "step": 9314
+    },
+    {
+      "epoch": 85.46,
+      "learning_rate": 4.595192307692307e-05,
+      "loss": 0.4422,
+      "step": 9315
+    },
+    {
+      "epoch": 85.47,
+      "learning_rate": 4.592307692307692e-05,
+      "loss": 0.4991,
+      "step": 9316
+    },
+    {
+      "epoch": 85.48,
+      "learning_rate": 4.589423076923076e-05,
+      "loss": 0.4519,
+      "step": 9317
+    },
+    {
+      "epoch": 85.49,
+      "learning_rate": 4.5865384615384614e-05,
+      "loss": 0.4646,
+      "step": 9318
+    },
+    {
+      "epoch": 85.5,
+      "learning_rate": 4.583653846153846e-05,
+      "loss": 0.4582,
+      "step": 9319
+    },
+    {
+      "epoch": 85.5,
+      "learning_rate": 4.5807692307692304e-05,
+      "loss": 0.5697,
+      "step": 9320
+    },
+    {
+      "epoch": 85.51,
+      "learning_rate": 4.577884615384615e-05,
+      "loss": 0.6535,
+      "step": 9321
+    },
+    {
+      "epoch": 85.52,
+      "learning_rate": 4.5749999999999994e-05,
+      "loss": 0.5218,
+      "step": 9322
+    },
+    {
+      "epoch": 85.53,
+      "learning_rate": 4.5721153846153846e-05,
+      "loss": 0.5679,
+      "step": 9323
+    },
+    {
+      "epoch": 85.54,
+      "learning_rate": 4.569230769230769e-05,
+      "loss": 0.6061,
+      "step": 9324
+    },
+    {
+      "epoch": 85.55,
+      "learning_rate": 4.566346153846153e-05,
+      "loss": 0.6069,
+      "step": 9325
+    },
+    {
+      "epoch": 85.56,
+      "learning_rate": 4.563461538461538e-05,
+      "loss": 0.347,
+      "step": 9326
+    },
+    {
+      "epoch": 85.57,
+      "learning_rate": 4.5605769230769225e-05,
+      "loss": 0.5081,
+      "step": 9327
+    },
+    {
+      "epoch": 85.58,
+      "learning_rate": 4.557692307692307e-05,
+      "loss": 0.4942,
+      "step": 9328
+    },
+    {
+      "epoch": 85.59,
+      "learning_rate": 4.554807692307692e-05,
+      "loss": 0.5051,
+      "step": 9329
+    },
+    {
+      "epoch": 85.6,
+      "learning_rate": 4.551923076923076e-05,
+      "loss": 0.6278,
+      "step": 9330
+    },
+    {
+      "epoch": 85.61,
+      "learning_rate": 4.549038461538461e-05,
+      "loss": 0.5504,
+      "step": 9331
+    },
+    {
+      "epoch": 85.61,
+      "learning_rate": 4.546153846153846e-05,
+      "loss": 0.5311,
+      "step": 9332
+    },
+    {
+      "epoch": 85.62,
+      "learning_rate": 4.54326923076923e-05,
+      "loss": 0.5114,
+      "step": 9333
+    },
+    {
+      "epoch": 85.63,
+      "learning_rate": 4.5403846153846153e-05,
+      "loss": 0.5838,
+      "step": 9334
+    },
+    {
+      "epoch": 85.64,
+      "learning_rate": 4.537499999999999e-05,
+      "loss": 0.5369,
+      "step": 9335
+    },
+    {
+      "epoch": 85.65,
+      "learning_rate": 4.534615384615384e-05,
+      "loss": 0.5044,
+      "step": 9336
+    },
+    {
+      "epoch": 85.66,
+      "learning_rate": 4.531730769230769e-05,
+      "loss": 0.5694,
+      "step": 9337
+    },
+    {
+      "epoch": 85.67,
+      "learning_rate": 4.528846153846153e-05,
+      "loss": 0.4431,
+      "step": 9338
+    },
+    {
+      "epoch": 85.68,
+      "learning_rate": 4.5259615384615385e-05,
+      "loss": 0.5864,
+      "step": 9339
+    },
+    {
+      "epoch": 85.69,
+      "learning_rate": 4.523076923076922e-05,
+      "loss": 0.501,
+      "step": 9340
+    },
+    {
+      "epoch": 85.7,
+      "learning_rate": 4.520192307692307e-05,
+      "loss": 0.584,
+      "step": 9341
+    },
+    {
+      "epoch": 85.71,
+      "learning_rate": 4.517307692307692e-05,
+      "loss": 0.5464,
+      "step": 9342
+    },
+    {
+      "epoch": 85.72,
+      "learning_rate": 4.5144230769230765e-05,
+      "loss": 0.5706,
+      "step": 9343
+    },
+    {
+      "epoch": 85.72,
+      "learning_rate": 4.5115384615384616e-05,
+      "loss": 0.5049,
+      "step": 9344
+    },
+    {
+      "epoch": 85.73,
+      "learning_rate": 4.5086538461538454e-05,
+      "loss": 0.4844,
+      "step": 9345
+    },
+    {
+      "epoch": 85.74,
+      "learning_rate": 4.50576923076923e-05,
+      "loss": 0.5662,
+      "step": 9346
+    },
+    {
+      "epoch": 85.75,
+      "learning_rate": 4.502884615384615e-05,
+      "loss": 0.5249,
+      "step": 9347
+    },
+    {
+      "epoch": 85.76,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 0.5715,
+      "step": 9348
+    },
+    {
+      "epoch": 85.77,
+      "learning_rate": 4.497115384615385e-05,
+      "loss": 0.582,
+      "step": 9349
+    },
+    {
+      "epoch": 85.78,
+      "learning_rate": 4.4942307692307686e-05,
+      "loss": 0.5834,
+      "step": 9350
+    },
+    {
+      "epoch": 85.79,
+      "learning_rate": 4.491346153846153e-05,
+      "loss": 0.5043,
+      "step": 9351
+    },
+    {
+      "epoch": 85.8,
+      "learning_rate": 4.488461538461538e-05,
+      "loss": 0.5392,
+      "step": 9352
+    },
+    {
+      "epoch": 85.81,
+      "learning_rate": 4.485576923076923e-05,
+      "loss": 0.5667,
+      "step": 9353
+    },
+    {
+      "epoch": 85.82,
+      "learning_rate": 4.482692307692308e-05,
+      "loss": 0.5643,
+      "step": 9354
+    },
+    {
+      "epoch": 85.83,
+      "learning_rate": 4.479807692307692e-05,
+      "loss": 0.533,
+      "step": 9355
+    },
+    {
+      "epoch": 85.83,
+      "learning_rate": 4.476923076923076e-05,
+      "loss": 0.509,
+      "step": 9356
+    },
+    {
+      "epoch": 85.84,
+      "learning_rate": 4.4740384615384614e-05,
+      "loss": 0.5606,
+      "step": 9357
+    },
+    {
+      "epoch": 85.85,
+      "learning_rate": 4.471153846153846e-05,
+      "loss": 0.5605,
+      "step": 9358
+    },
+    {
+      "epoch": 85.86,
+      "learning_rate": 4.46826923076923e-05,
+      "loss": 0.5006,
+      "step": 9359
+    },
+    {
+      "epoch": 85.87,
+      "learning_rate": 4.465384615384615e-05,
+      "loss": 0.5221,
+      "step": 9360
+    },
+    {
+      "epoch": 85.88,
+      "learning_rate": 4.4624999999999994e-05,
+      "loss": 0.5519,
+      "step": 9361
+    },
+    {
+      "epoch": 85.89,
+      "learning_rate": 4.4596153846153845e-05,
+      "loss": 0.4757,
+      "step": 9362
+    },
+    {
+      "epoch": 85.9,
+      "learning_rate": 4.456730769230769e-05,
+      "loss": 0.6342,
+      "step": 9363
+    },
+    {
+      "epoch": 85.91,
+      "learning_rate": 4.453846153846153e-05,
+      "loss": 0.7296,
+      "step": 9364
+    },
+    {
+      "epoch": 85.92,
+      "learning_rate": 4.450961538461538e-05,
+      "loss": 0.4755,
+      "step": 9365
+    },
+    {
+      "epoch": 85.93,
+      "learning_rate": 4.4480769230769225e-05,
+      "loss": 0.5631,
+      "step": 9366
+    },
+    {
+      "epoch": 85.94,
+      "learning_rate": 4.445192307692308e-05,
+      "loss": 0.4282,
+      "step": 9367
+    },
+    {
+      "epoch": 85.94,
+      "learning_rate": 4.442307692307692e-05,
+      "loss": 0.48,
+      "step": 9368
+    },
+    {
+      "epoch": 85.95,
+      "learning_rate": 4.439423076923076e-05,
+      "loss": 0.4453,
+      "step": 9369
+    },
+    {
+      "epoch": 85.96,
+      "learning_rate": 4.436538461538461e-05,
+      "loss": 0.4986,
+      "step": 9370
+    },
+    {
+      "epoch": 85.97,
+      "learning_rate": 4.4336538461538457e-05,
+      "loss": 0.5828,
+      "step": 9371
+    },
+    {
+      "epoch": 85.98,
+      "learning_rate": 4.430769230769231e-05,
+      "loss": 0.5045,
+      "step": 9372
+    },
+    {
+      "epoch": 85.99,
+      "learning_rate": 4.427884615384615e-05,
+      "loss": 0.5185,
+      "step": 9373
+    },
+    {
+      "epoch": 86.0,
+      "learning_rate": 4.424999999999999e-05,
+      "loss": 0.5583,
+      "step": 9374
+    },
+    {
+      "epoch": 86.01,
+      "learning_rate": 4.422115384615384e-05,
+      "loss": 0.4983,
+      "step": 9375
+    },
+    {
+      "epoch": 86.02,
+      "learning_rate": 4.419230769230769e-05,
+      "loss": 0.4201,
+      "step": 9376
+    },
+    {
+      "epoch": 86.03,
+      "learning_rate": 4.416346153846153e-05,
+      "loss": 0.4704,
+      "step": 9377
+    },
+    {
+      "epoch": 86.04,
+      "learning_rate": 4.4134615384615385e-05,
+      "loss": 0.5355,
+      "step": 9378
+    },
+    {
+      "epoch": 86.05,
+      "learning_rate": 4.410576923076922e-05,
+      "loss": 0.4606,
+      "step": 9379
+    },
+    {
+      "epoch": 86.06,
+      "learning_rate": 4.4076923076923075e-05,
+      "loss": 0.4266,
+      "step": 9380
+    },
+    {
+      "epoch": 86.06,
+      "learning_rate": 4.404807692307692e-05,
+      "loss": 0.532,
+      "step": 9381
+    },
+    {
+      "epoch": 86.07,
+      "learning_rate": 4.4019230769230764e-05,
+      "loss": 0.5509,
+      "step": 9382
+    },
+    {
+      "epoch": 86.08,
+      "learning_rate": 4.3990384615384616e-05,
+      "loss": 0.4563,
+      "step": 9383
+    },
+    {
+      "epoch": 86.09,
+      "learning_rate": 4.3961538461538454e-05,
+      "loss": 0.5588,
+      "step": 9384
+    },
+    {
+      "epoch": 86.1,
+      "learning_rate": 4.3932692307692306e-05,
+      "loss": 0.5635,
+      "step": 9385
+    },
+    {
+      "epoch": 86.11,
+      "learning_rate": 4.390384615384615e-05,
+      "loss": 0.5757,
+      "step": 9386
+    },
+    {
+      "epoch": 86.12,
+      "learning_rate": 4.3874999999999996e-05,
+      "loss": 0.5142,
+      "step": 9387
+    },
+    {
+      "epoch": 86.13,
+      "learning_rate": 4.384615384615385e-05,
+      "loss": 0.5932,
+      "step": 9388
+    },
+    {
+      "epoch": 86.14,
+      "learning_rate": 4.3817307692307686e-05,
+      "loss": 0.4965,
+      "step": 9389
+    },
+    {
+      "epoch": 86.15,
+      "learning_rate": 4.378846153846153e-05,
+      "loss": 0.591,
+      "step": 9390
+    },
+    {
+      "epoch": 86.16,
+      "learning_rate": 4.375961538461538e-05,
+      "loss": 0.6709,
+      "step": 9391
+    },
+    {
+      "epoch": 86.17,
+      "learning_rate": 4.373076923076923e-05,
+      "loss": 0.5426,
+      "step": 9392
+    },
+    {
+      "epoch": 86.17,
+      "learning_rate": 4.370192307692308e-05,
+      "loss": 0.5368,
+      "step": 9393
+    },
+    {
+      "epoch": 86.18,
+      "learning_rate": 4.367307692307692e-05,
+      "loss": 0.533,
+      "step": 9394
+    },
+    {
+      "epoch": 86.19,
+      "learning_rate": 4.364423076923076e-05,
+      "loss": 0.4834,
+      "step": 9395
+    },
+    {
+      "epoch": 86.2,
+      "learning_rate": 4.3615384615384614e-05,
+      "loss": 0.6036,
+      "step": 9396
+    },
+    {
+      "epoch": 86.21,
+      "learning_rate": 4.358653846153846e-05,
+      "loss": 0.5623,
+      "step": 9397
+    },
+    {
+      "epoch": 86.22,
+      "learning_rate": 4.355769230769231e-05,
+      "loss": 0.4567,
+      "step": 9398
+    },
+    {
+      "epoch": 86.23,
+      "learning_rate": 4.352884615384615e-05,
+      "loss": 0.4393,
+      "step": 9399
+    },
+    {
+      "epoch": 86.24,
+      "learning_rate": 4.3499999999999993e-05,
+      "loss": 0.5892,
+      "step": 9400
+    },
+    {
+      "epoch": 86.25,
+      "learning_rate": 4.3471153846153845e-05,
+      "loss": 0.5198,
+      "step": 9401
+    },
+    {
+      "epoch": 86.26,
+      "learning_rate": 4.344230769230769e-05,
+      "loss": 0.5496,
+      "step": 9402
+    },
+    {
+      "epoch": 86.27,
+      "learning_rate": 4.341346153846154e-05,
+      "loss": 0.5738,
+      "step": 9403
+    },
+    {
+      "epoch": 86.28,
+      "learning_rate": 4.338461538461538e-05,
+      "loss": 0.5307,
+      "step": 9404
+    },
+    {
+      "epoch": 86.28,
+      "learning_rate": 4.3355769230769225e-05,
+      "loss": 0.473,
+      "step": 9405
+    },
+    {
+      "epoch": 86.29,
+      "learning_rate": 4.332692307692308e-05,
+      "loss": 0.573,
+      "step": 9406
+    },
+    {
+      "epoch": 86.3,
+      "learning_rate": 4.329807692307692e-05,
+      "loss": 0.5301,
+      "step": 9407
+    },
+    {
+      "epoch": 86.31,
+      "learning_rate": 4.326923076923076e-05,
+      "loss": 0.5091,
+      "step": 9408
+    },
+    {
+      "epoch": 86.32,
+      "learning_rate": 4.324038461538461e-05,
+      "loss": 0.4898,
+      "step": 9409
+    },
+    {
+      "epoch": 86.33,
+      "learning_rate": 4.3211538461538456e-05,
+      "loss": 0.5351,
+      "step": 9410
+    },
+    {
+      "epoch": 86.34,
+      "learning_rate": 4.318269230769231e-05,
+      "loss": 0.4879,
+      "step": 9411
+    },
+    {
+      "epoch": 86.35,
+      "learning_rate": 4.315384615384615e-05,
+      "loss": 0.4497,
+      "step": 9412
+    },
+    {
+      "epoch": 86.36,
+      "learning_rate": 4.312499999999999e-05,
+      "loss": 0.584,
+      "step": 9413
+    },
+    {
+      "epoch": 86.37,
+      "learning_rate": 4.309615384615384e-05,
+      "loss": 0.5433,
+      "step": 9414
+    },
+    {
+      "epoch": 86.38,
+      "learning_rate": 4.306730769230769e-05,
+      "loss": 0.58,
+      "step": 9415
+    },
+    {
+      "epoch": 86.39,
+      "learning_rate": 4.303846153846154e-05,
+      "loss": 0.4817,
+      "step": 9416
+    },
+    {
+      "epoch": 86.39,
+      "learning_rate": 4.3009615384615384e-05,
+      "loss": 0.6456,
+      "step": 9417
+    },
+    {
+      "epoch": 86.4,
+      "learning_rate": 4.298076923076922e-05,
+      "loss": 0.4369,
+      "step": 9418
+    },
+    {
+      "epoch": 86.41,
+      "learning_rate": 4.2951923076923074e-05,
+      "loss": 0.4924,
+      "step": 9419
+    },
+    {
+      "epoch": 86.42,
+      "learning_rate": 4.292307692307692e-05,
+      "loss": 0.5849,
+      "step": 9420
+    },
+    {
+      "epoch": 86.43,
+      "learning_rate": 4.289423076923077e-05,
+      "loss": 0.5151,
+      "step": 9421
+    },
+    {
+      "epoch": 86.44,
+      "learning_rate": 4.2865384615384616e-05,
+      "loss": 0.5271,
+      "step": 9422
+    },
+    {
+      "epoch": 86.45,
+      "learning_rate": 4.2836538461538454e-05,
+      "loss": 0.5714,
+      "step": 9423
+    },
+    {
+      "epoch": 86.46,
+      "learning_rate": 4.2807692307692306e-05,
+      "loss": 0.4924,
+      "step": 9424
+    },
+    {
+      "epoch": 86.47,
+      "learning_rate": 4.277884615384615e-05,
+      "loss": 0.5645,
+      "step": 9425
+    },
+    {
+      "epoch": 86.48,
+      "learning_rate": 4.2749999999999996e-05,
+      "loss": 0.519,
+      "step": 9426
+    },
+    {
+      "epoch": 86.49,
+      "learning_rate": 4.272115384615384e-05,
+      "loss": 0.5788,
+      "step": 9427
+    },
+    {
+      "epoch": 86.5,
+      "learning_rate": 4.2692307692307685e-05,
+      "loss": 0.572,
+      "step": 9428
+    },
+    {
+      "epoch": 86.5,
+      "learning_rate": 4.266346153846154e-05,
+      "loss": 0.5369,
+      "step": 9429
+    },
+    {
+      "epoch": 86.51,
+      "learning_rate": 4.263461538461538e-05,
+      "loss": 0.493,
+      "step": 9430
+    },
+    {
+      "epoch": 86.52,
+      "learning_rate": 4.260576923076923e-05,
+      "loss": 0.5529,
+      "step": 9431
+    },
+    {
+      "epoch": 86.53,
+      "learning_rate": 4.257692307692307e-05,
+      "loss": 0.5717,
+      "step": 9432
+    },
+    {
+      "epoch": 86.54,
+      "learning_rate": 4.254807692307692e-05,
+      "loss": 0.5691,
+      "step": 9433
+    },
+    {
+      "epoch": 86.55,
+      "learning_rate": 4.251923076923077e-05,
+      "loss": 0.6605,
+      "step": 9434
+    },
+    {
+      "epoch": 86.56,
+      "learning_rate": 4.2490384615384614e-05,
+      "loss": 0.6371,
+      "step": 9435
+    },
+    {
+      "epoch": 86.57,
+      "learning_rate": 4.246153846153846e-05,
+      "loss": 0.5326,
+      "step": 9436
+    },
+    {
+      "epoch": 86.58,
+      "learning_rate": 4.24326923076923e-05,
+      "loss": 0.49,
+      "step": 9437
+    },
+    {
+      "epoch": 86.59,
+      "learning_rate": 4.240384615384615e-05,
+      "loss": 0.605,
+      "step": 9438
+    },
+    {
+      "epoch": 86.6,
+      "learning_rate": 4.237499999999999e-05,
+      "loss": 0.5053,
+      "step": 9439
+    },
+    {
+      "epoch": 86.61,
+      "learning_rate": 4.2346153846153845e-05,
+      "loss": 0.5396,
+      "step": 9440
+    },
+    {
+      "epoch": 86.61,
+      "learning_rate": 4.231730769230768e-05,
+      "loss": 0.6128,
+      "step": 9441
+    },
+    {
+      "epoch": 86.62,
+      "learning_rate": 4.2288461538461535e-05,
+      "loss": 0.5968,
+      "step": 9442
+    },
+    {
+      "epoch": 86.63,
+      "learning_rate": 4.225961538461538e-05,
+      "loss": 0.5116,
+      "step": 9443
+    },
+    {
+      "epoch": 86.64,
+      "learning_rate": 4.2230769230769225e-05,
+      "loss": 0.5359,
+      "step": 9444
+    },
+    {
+      "epoch": 86.65,
+      "learning_rate": 4.2201923076923076e-05,
+      "loss": 0.4431,
+      "step": 9445
+    },
+    {
+      "epoch": 86.66,
+      "learning_rate": 4.2173076923076915e-05,
+      "loss": 0.5075,
+      "step": 9446
+    },
+    {
+      "epoch": 86.67,
+      "learning_rate": 4.2144230769230766e-05,
+      "loss": 0.5722,
+      "step": 9447
+    },
+    {
+      "epoch": 86.68,
+      "learning_rate": 4.211538461538461e-05,
+      "loss": 0.4845,
+      "step": 9448
+    },
+    {
+      "epoch": 86.69,
+      "learning_rate": 4.2086538461538456e-05,
+      "loss": 0.3635,
+      "step": 9449
+    },
+    {
+      "epoch": 86.7,
+      "learning_rate": 4.205769230769231e-05,
+      "loss": 0.4941,
+      "step": 9450
+    },
+    {
+      "epoch": 86.71,
+      "learning_rate": 4.2028846153846146e-05,
+      "loss": 0.5096,
+      "step": 9451
+    },
+    {
+      "epoch": 86.72,
+      "learning_rate": 4.2e-05,
+      "loss": 0.5396,
+      "step": 9452
+    },
+    {
+      "epoch": 86.72,
+      "learning_rate": 4.197115384615384e-05,
+      "loss": 0.5922,
+      "step": 9453
+    },
+    {
+      "epoch": 86.73,
+      "learning_rate": 4.194230769230769e-05,
+      "loss": 0.55,
+      "step": 9454
+    },
+    {
+      "epoch": 86.74,
+      "learning_rate": 4.191346153846154e-05,
+      "loss": 0.5192,
+      "step": 9455
+    },
+    {
+      "epoch": 86.75,
+      "learning_rate": 4.188461538461538e-05,
+      "loss": 0.5895,
+      "step": 9456
+    },
+    {
+      "epoch": 86.76,
+      "learning_rate": 4.185576923076922e-05,
+      "loss": 0.5089,
+      "step": 9457
+    },
+    {
+      "epoch": 86.77,
+      "learning_rate": 4.1826923076923074e-05,
+      "loss": 0.4143,
+      "step": 9458
+    },
+    {
+      "epoch": 86.78,
+      "learning_rate": 4.179807692307692e-05,
+      "loss": 0.4494,
+      "step": 9459
+    },
+    {
+      "epoch": 86.79,
+      "learning_rate": 4.176923076923077e-05,
+      "loss": 0.5712,
+      "step": 9460
+    },
+    {
+      "epoch": 86.8,
+      "learning_rate": 4.174038461538461e-05,
+      "loss": 0.645,
+      "step": 9461
+    },
+    {
+      "epoch": 86.81,
+      "learning_rate": 4.1711538461538454e-05,
+      "loss": 0.4835,
+      "step": 9462
+    },
+    {
+      "epoch": 86.82,
+      "learning_rate": 4.1682692307692305e-05,
+      "loss": 0.4865,
+      "step": 9463
+    },
+    {
+      "epoch": 86.83,
+      "learning_rate": 4.165384615384615e-05,
+      "loss": 0.4463,
+      "step": 9464
+    },
+    {
+      "epoch": 86.83,
+      "learning_rate": 4.1625e-05,
+      "loss": 0.53,
+      "step": 9465
+    },
+    {
+      "epoch": 86.84,
+      "learning_rate": 4.159615384615384e-05,
+      "loss": 0.648,
+      "step": 9466
+    },
+    {
+      "epoch": 86.85,
+      "learning_rate": 4.1567307692307685e-05,
+      "loss": 0.544,
+      "step": 9467
+    },
+    {
+      "epoch": 86.86,
+      "learning_rate": 4.153846153846154e-05,
+      "loss": 0.4957,
+      "step": 9468
+    },
+    {
+      "epoch": 86.87,
+      "learning_rate": 4.150961538461538e-05,
+      "loss": 0.5451,
+      "step": 9469
+    },
+    {
+      "epoch": 86.88,
+      "learning_rate": 4.148076923076922e-05,
+      "loss": 0.4499,
+      "step": 9470
+    },
+    {
+      "epoch": 86.89,
+      "learning_rate": 4.145192307692307e-05,
+      "loss": 0.5904,
+      "step": 9471
+    },
+    {
+      "epoch": 86.9,
+      "learning_rate": 4.142307692307692e-05,
+      "loss": 0.5154,
+      "step": 9472
+    },
+    {
+      "epoch": 86.91,
+      "learning_rate": 4.139423076923077e-05,
+      "loss": 0.5029,
+      "step": 9473
+    },
+    {
+      "epoch": 86.92,
+      "learning_rate": 4.136538461538461e-05,
+      "loss": 0.3316,
+      "step": 9474
+    },
+    {
+      "epoch": 86.93,
+      "learning_rate": 4.133653846153845e-05,
+      "loss": 0.5324,
+      "step": 9475
+    },
+    {
+      "epoch": 86.94,
+      "learning_rate": 4.13076923076923e-05,
+      "loss": 0.5387,
+      "step": 9476
+    },
+    {
+      "epoch": 86.94,
+      "learning_rate": 4.127884615384615e-05,
+      "loss": 0.4375,
+      "step": 9477
+    },
+    {
+      "epoch": 86.95,
+      "learning_rate": 4.125e-05,
+      "loss": 0.5634,
+      "step": 9478
+    },
+    {
+      "epoch": 86.96,
+      "learning_rate": 4.1221153846153845e-05,
+      "loss": 0.45,
+      "step": 9479
+    },
+    {
+      "epoch": 86.97,
+      "learning_rate": 4.119230769230768e-05,
+      "loss": 0.5424,
+      "step": 9480
+    },
+    {
+      "epoch": 86.98,
+      "learning_rate": 4.1163461538461535e-05,
+      "loss": 0.4522,
+      "step": 9481
+    },
+    {
+      "epoch": 86.99,
+      "learning_rate": 4.113461538461538e-05,
+      "loss": 0.6325,
+      "step": 9482
+    },
+    {
+      "epoch": 87.0,
+      "learning_rate": 4.110576923076923e-05,
+      "loss": 0.4704,
+      "step": 9483
+    },
+    {
+      "epoch": 87.01,
+      "learning_rate": 4.1076923076923076e-05,
+      "loss": 0.6499,
+      "step": 9484
+    },
+    {
+      "epoch": 87.02,
+      "learning_rate": 4.1048076923076914e-05,
+      "loss": 0.4847,
+      "step": 9485
+    },
+    {
+      "epoch": 87.03,
+      "learning_rate": 4.1019230769230766e-05,
+      "loss": 0.4872,
+      "step": 9486
+    },
+    {
+      "epoch": 87.04,
+      "learning_rate": 4.099038461538461e-05,
+      "loss": 0.5995,
+      "step": 9487
+    },
+    {
+      "epoch": 87.05,
+      "learning_rate": 4.0961538461538456e-05,
+      "loss": 0.5773,
+      "step": 9488
+    },
+    {
+      "epoch": 87.06,
+      "learning_rate": 4.093269230769231e-05,
+      "loss": 0.6042,
+      "step": 9489
+    },
+    {
+      "epoch": 87.06,
+      "learning_rate": 4.0903846153846146e-05,
+      "loss": 0.4988,
+      "step": 9490
+    },
+    {
+      "epoch": 87.07,
+      "learning_rate": 4.0875e-05,
+      "loss": 0.4975,
+      "step": 9491
+    },
+    {
+      "epoch": 87.08,
+      "learning_rate": 4.084615384615384e-05,
+      "loss": 0.5946,
+      "step": 9492
+    },
+    {
+      "epoch": 87.09,
+      "learning_rate": 4.081730769230769e-05,
+      "loss": 0.5851,
+      "step": 9493
+    },
+    {
+      "epoch": 87.1,
+      "learning_rate": 4.078846153846154e-05,
+      "loss": 0.4886,
+      "step": 9494
+    },
+    {
+      "epoch": 87.11,
+      "learning_rate": 4.075961538461538e-05,
+      "loss": 0.5965,
+      "step": 9495
+    },
+    {
+      "epoch": 87.12,
+      "learning_rate": 4.073076923076923e-05,
+      "loss": 0.422,
+      "step": 9496
+    },
+    {
+      "epoch": 87.13,
+      "learning_rate": 4.0701923076923074e-05,
+      "loss": 0.6155,
+      "step": 9497
+    },
+    {
+      "epoch": 87.14,
+      "learning_rate": 4.067307692307692e-05,
+      "loss": 0.5098,
+      "step": 9498
+    },
+    {
+      "epoch": 87.15,
+      "learning_rate": 4.064423076923077e-05,
+      "loss": 0.5375,
+      "step": 9499
+    },
+    {
+      "epoch": 87.16,
+      "learning_rate": 4.061538461538461e-05,
+      "loss": 0.4763,
+      "step": 9500
+    },
+    {
+      "epoch": 87.16,
+      "eval_cer": 0.08400425793490157,
+      "eval_loss": 0.36268001794815063,
+      "eval_runtime": 13.945,
+      "eval_samples_per_second": 118.107,
+      "eval_steps_per_second": 1.864,
+      "eval_wer": 0.3509345317127975,
+      "step": 9500
+    },
+    {
+      "epoch": 87.17,
+      "learning_rate": 4.058653846153846e-05,
+      "loss": 0.4159,
+      "step": 9501
+    },
+    {
+      "epoch": 87.17,
+      "learning_rate": 4.0557692307692305e-05,
+      "loss": 0.5184,
+      "step": 9502
+    },
+    {
+      "epoch": 87.18,
+      "learning_rate": 4.052884615384615e-05,
+      "loss": 0.5483,
+      "step": 9503
+    },
+    {
+      "epoch": 87.19,
+      "learning_rate": 4.05e-05,
+      "loss": 0.5306,
+      "step": 9504
+    },
+    {
+      "epoch": 87.2,
+      "learning_rate": 4.047115384615384e-05,
+      "loss": 0.5611,
+      "step": 9505
+    },
+    {
+      "epoch": 87.21,
+      "learning_rate": 4.0442307692307685e-05,
+      "loss": 0.6939,
+      "step": 9506
+    },
+    {
+      "epoch": 87.22,
+      "learning_rate": 4.041346153846154e-05,
+      "loss": 0.5255,
+      "step": 9507
+    },
+    {
+      "epoch": 87.23,
+      "learning_rate": 4.038461538461538e-05,
+      "loss": 0.6116,
+      "step": 9508
+    },
+    {
+      "epoch": 87.24,
+      "learning_rate": 4.035576923076923e-05,
+      "loss": 0.5088,
+      "step": 9509
+    },
+    {
+      "epoch": 87.25,
+      "learning_rate": 4.032692307692307e-05,
+      "loss": 0.6227,
+      "step": 9510
+    },
+    {
+      "epoch": 87.26,
+      "learning_rate": 4.0298076923076916e-05,
+      "loss": 0.5418,
+      "step": 9511
+    },
+    {
+      "epoch": 87.27,
+      "learning_rate": 4.026923076923077e-05,
+      "loss": 0.373,
+      "step": 9512
+    },
+    {
+      "epoch": 87.28,
+      "learning_rate": 4.024038461538461e-05,
+      "loss": 0.3755,
+      "step": 9513
+    },
+    {
+      "epoch": 87.28,
+      "learning_rate": 4.0211538461538465e-05,
+      "loss": 0.4867,
+      "step": 9514
+    },
+    {
+      "epoch": 87.29,
+      "learning_rate": 4.01826923076923e-05,
+      "loss": 0.4775,
+      "step": 9515
+    },
+    {
+      "epoch": 87.3,
+      "learning_rate": 4.015384615384615e-05,
+      "loss": 0.439,
+      "step": 9516
+    },
+    {
+      "epoch": 87.31,
+      "learning_rate": 4.0125e-05,
+      "loss": 0.3941,
+      "step": 9517
+    },
+    {
+      "epoch": 87.32,
+      "learning_rate": 4.0096153846153844e-05,
+      "loss": 0.463,
+      "step": 9518
+    },
+    {
+      "epoch": 87.33,
+      "learning_rate": 4.006730769230768e-05,
+      "loss": 0.4667,
+      "step": 9519
+    },
+    {
+      "epoch": 87.34,
+      "learning_rate": 4.0038461538461534e-05,
+      "loss": 0.6043,
+      "step": 9520
+    },
+    {
+      "epoch": 87.35,
+      "learning_rate": 4.000961538461538e-05,
+      "loss": 0.5088,
+      "step": 9521
+    },
+    {
+      "epoch": 87.36,
+      "learning_rate": 3.998076923076923e-05,
+      "loss": 0.5592,
+      "step": 9522
+    },
+    {
+      "epoch": 87.37,
+      "learning_rate": 3.9951923076923076e-05,
+      "loss": 0.5082,
+      "step": 9523
+    },
+    {
+      "epoch": 87.38,
+      "learning_rate": 3.9923076923076914e-05,
+      "loss": 0.5808,
+      "step": 9524
+    },
+    {
+      "epoch": 87.39,
+      "learning_rate": 3.9894230769230766e-05,
+      "loss": 0.6711,
+      "step": 9525
+    },
+    {
+      "epoch": 87.39,
+      "learning_rate": 3.986538461538461e-05,
+      "loss": 0.5634,
+      "step": 9526
+    },
+    {
+      "epoch": 87.4,
+      "learning_rate": 3.983653846153846e-05,
+      "loss": 0.5517,
+      "step": 9527
+    },
+    {
+      "epoch": 87.41,
+      "learning_rate": 3.980769230769231e-05,
+      "loss": 0.6343,
+      "step": 9528
+    },
+    {
+      "epoch": 87.42,
+      "learning_rate": 3.9778846153846145e-05,
+      "loss": 0.6007,
+      "step": 9529
+    },
+    {
+      "epoch": 87.43,
+      "learning_rate": 3.975e-05,
+      "loss": 0.4261,
+      "step": 9530
+    },
+    {
+      "epoch": 87.44,
+      "learning_rate": 3.972115384615384e-05,
+      "loss": 0.5376,
+      "step": 9531
+    },
+    {
+      "epoch": 87.45,
+      "learning_rate": 3.9692307692307694e-05,
+      "loss": 0.5967,
+      "step": 9532
+    },
+    {
+      "epoch": 87.46,
+      "learning_rate": 3.966346153846154e-05,
+      "loss": 0.4339,
+      "step": 9533
+    },
+    {
+      "epoch": 87.47,
+      "learning_rate": 3.963461538461538e-05,
+      "loss": 0.5877,
+      "step": 9534
+    },
+    {
+      "epoch": 87.48,
+      "learning_rate": 3.960576923076923e-05,
+      "loss": 0.4909,
+      "step": 9535
+    },
+    {
+      "epoch": 87.49,
+      "learning_rate": 3.9576923076923074e-05,
+      "loss": 0.5678,
+      "step": 9536
+    },
+    {
+      "epoch": 87.5,
+      "learning_rate": 3.954807692307692e-05,
+      "loss": 0.5793,
+      "step": 9537
+    },
+    {
+      "epoch": 87.5,
+      "learning_rate": 3.951923076923077e-05,
+      "loss": 0.5103,
+      "step": 9538
+    },
+    {
+      "epoch": 87.51,
+      "learning_rate": 3.949038461538461e-05,
+      "loss": 0.4424,
+      "step": 9539
+    },
+    {
+      "epoch": 87.52,
+      "learning_rate": 3.946153846153846e-05,
+      "loss": 0.5348,
+      "step": 9540
+    },
+    {
+      "epoch": 87.53,
+      "learning_rate": 3.9432692307692305e-05,
+      "loss": 0.4851,
+      "step": 9541
+    },
+    {
+      "epoch": 87.54,
+      "learning_rate": 3.940384615384615e-05,
+      "loss": 0.5052,
+      "step": 9542
+    },
+    {
+      "epoch": 87.55,
+      "learning_rate": 3.9374999999999995e-05,
+      "loss": 0.6386,
+      "step": 9543
+    },
+    {
+      "epoch": 87.56,
+      "learning_rate": 3.934615384615384e-05,
+      "loss": 0.5596,
+      "step": 9544
+    },
+    {
+      "epoch": 87.57,
+      "learning_rate": 3.931730769230769e-05,
+      "loss": 0.6312,
+      "step": 9545
+    },
+    {
+      "epoch": 87.58,
+      "learning_rate": 3.9288461538461536e-05,
+      "loss": 0.4221,
+      "step": 9546
+    },
+    {
+      "epoch": 87.59,
+      "learning_rate": 3.925961538461538e-05,
+      "loss": 0.6041,
+      "step": 9547
+    },
+    {
+      "epoch": 87.6,
+      "learning_rate": 3.9230769230769226e-05,
+      "loss": 0.5759,
+      "step": 9548
+    },
+    {
+      "epoch": 87.61,
+      "learning_rate": 3.920192307692307e-05,
+      "loss": 0.5959,
+      "step": 9549
+    },
+    {
+      "epoch": 87.61,
+      "learning_rate": 3.917307692307692e-05,
+      "loss": 0.4873,
+      "step": 9550
+    },
+    {
+      "epoch": 87.62,
+      "learning_rate": 3.914423076923077e-05,
+      "loss": 0.5868,
+      "step": 9551
+    },
+    {
+      "epoch": 87.63,
+      "learning_rate": 3.911538461538461e-05,
+      "loss": 0.4642,
+      "step": 9552
+    },
+    {
+      "epoch": 87.64,
+      "learning_rate": 3.908653846153846e-05,
+      "loss": 0.6097,
+      "step": 9553
+    },
+    {
+      "epoch": 87.65,
+      "learning_rate": 3.90576923076923e-05,
+      "loss": 0.5283,
+      "step": 9554
+    },
+    {
+      "epoch": 87.66,
+      "learning_rate": 3.902884615384615e-05,
+      "loss": 0.5414,
+      "step": 9555
+    },
+    {
+      "epoch": 87.67,
+      "learning_rate": 3.9e-05,
+      "loss": 0.4713,
+      "step": 9556
+    },
+    {
+      "epoch": 87.68,
+      "learning_rate": 3.8971153846153844e-05,
+      "loss": 0.5573,
+      "step": 9557
+    },
+    {
+      "epoch": 87.69,
+      "learning_rate": 3.894230769230769e-05,
+      "loss": 0.443,
+      "step": 9558
+    },
+    {
+      "epoch": 87.7,
+      "learning_rate": 3.8913461538461534e-05,
+      "loss": 0.4417,
+      "step": 9559
+    },
+    {
+      "epoch": 87.71,
+      "learning_rate": 3.888461538461538e-05,
+      "loss": 0.5535,
+      "step": 9560
+    },
+    {
+      "epoch": 87.72,
+      "learning_rate": 3.885576923076923e-05,
+      "loss": 0.6068,
+      "step": 9561
+    },
+    {
+      "epoch": 87.72,
+      "learning_rate": 3.882692307692307e-05,
+      "loss": 0.4215,
+      "step": 9562
+    },
+    {
+      "epoch": 87.73,
+      "learning_rate": 3.879807692307692e-05,
+      "loss": 0.5309,
+      "step": 9563
+    },
+    {
+      "epoch": 87.74,
+      "learning_rate": 3.8769230769230766e-05,
+      "loss": 0.6112,
+      "step": 9564
+    },
+    {
+      "epoch": 87.75,
+      "learning_rate": 3.874038461538461e-05,
+      "loss": 0.5884,
+      "step": 9565
+    },
+    {
+      "epoch": 87.76,
+      "learning_rate": 3.871153846153846e-05,
+      "loss": 0.5773,
+      "step": 9566
+    },
+    {
+      "epoch": 87.77,
+      "learning_rate": 3.86826923076923e-05,
+      "loss": 0.5246,
+      "step": 9567
+    },
+    {
+      "epoch": 87.78,
+      "learning_rate": 3.8653846153846145e-05,
+      "loss": 0.5494,
+      "step": 9568
+    },
+    {
+      "epoch": 87.79,
+      "learning_rate": 3.8625e-05,
+      "loss": 0.5444,
+      "step": 9569
+    },
+    {
+      "epoch": 87.8,
+      "learning_rate": 3.859615384615384e-05,
+      "loss": 0.636,
+      "step": 9570
+    },
+    {
+      "epoch": 87.81,
+      "learning_rate": 3.8567307692307694e-05,
+      "loss": 0.4976,
+      "step": 9571
+    },
+    {
+      "epoch": 87.82,
+      "learning_rate": 3.853846153846153e-05,
+      "loss": 0.483,
+      "step": 9572
+    },
+    {
+      "epoch": 87.83,
+      "learning_rate": 3.850961538461538e-05,
+      "loss": 0.4805,
+      "step": 9573
+    },
+    {
+      "epoch": 87.83,
+      "learning_rate": 3.848076923076923e-05,
+      "loss": 0.4391,
+      "step": 9574
+    },
+    {
+      "epoch": 87.84,
+      "learning_rate": 3.845192307692307e-05,
+      "loss": 0.5505,
+      "step": 9575
+    },
+    {
+      "epoch": 87.85,
+      "learning_rate": 3.8423076923076925e-05,
+      "loss": 0.5309,
+      "step": 9576
+    },
+    {
+      "epoch": 87.86,
+      "learning_rate": 3.839423076923076e-05,
+      "loss": 0.5607,
+      "step": 9577
+    },
+    {
+      "epoch": 87.87,
+      "learning_rate": 3.836538461538461e-05,
+      "loss": 0.5025,
+      "step": 9578
+    },
+    {
+      "epoch": 87.88,
+      "learning_rate": 3.833653846153846e-05,
+      "loss": 0.4434,
+      "step": 9579
+    },
+    {
+      "epoch": 87.89,
+      "learning_rate": 3.8307692307692305e-05,
+      "loss": 0.4644,
+      "step": 9580
+    },
+    {
+      "epoch": 87.9,
+      "learning_rate": 3.8278846153846156e-05,
+      "loss": 0.5614,
+      "step": 9581
+    },
+    {
+      "epoch": 87.91,
+      "learning_rate": 3.8249999999999995e-05,
+      "loss": 0.5515,
+      "step": 9582
+    },
+    {
+      "epoch": 87.92,
+      "learning_rate": 3.822115384615384e-05,
+      "loss": 0.4754,
+      "step": 9583
+    },
+    {
+      "epoch": 87.93,
+      "learning_rate": 3.819230769230769e-05,
+      "loss": 0.5133,
+      "step": 9584
+    },
+    {
+      "epoch": 87.94,
+      "learning_rate": 3.8163461538461536e-05,
+      "loss": 0.571,
+      "step": 9585
+    },
+    {
+      "epoch": 87.94,
+      "learning_rate": 3.8134615384615374e-05,
+      "loss": 0.5108,
+      "step": 9586
+    },
+    {
+      "epoch": 87.95,
+      "learning_rate": 3.8105769230769226e-05,
+      "loss": 0.5975,
+      "step": 9587
+    },
+    {
+      "epoch": 87.96,
+      "learning_rate": 3.807692307692307e-05,
+      "loss": 0.5178,
+      "step": 9588
+    },
+    {
+      "epoch": 87.97,
+      "learning_rate": 3.804807692307692e-05,
+      "loss": 0.4963,
+      "step": 9589
+    },
+    {
+      "epoch": 87.98,
+      "learning_rate": 3.801923076923077e-05,
+      "loss": 0.5221,
+      "step": 9590
+    },
+    {
+      "epoch": 87.99,
+      "learning_rate": 3.7990384615384606e-05,
+      "loss": 0.4992,
+      "step": 9591
+    },
+    {
+      "epoch": 88.0,
+      "learning_rate": 3.796153846153846e-05,
+      "loss": 0.5859,
+      "step": 9592
+    },
+    {
+      "epoch": 88.01,
+      "learning_rate": 3.79326923076923e-05,
+      "loss": 0.5891,
+      "step": 9593
+    },
+    {
+      "epoch": 88.02,
+      "learning_rate": 3.7903846153846154e-05,
+      "loss": 0.5364,
+      "step": 9594
+    },
+    {
+      "epoch": 88.03,
+      "learning_rate": 3.7875e-05,
+      "loss": 0.5035,
+      "step": 9595
+    },
+    {
+      "epoch": 88.04,
+      "learning_rate": 3.784615384615384e-05,
+      "loss": 0.6467,
+      "step": 9596
+    },
+    {
+      "epoch": 88.05,
+      "learning_rate": 3.781730769230769e-05,
+      "loss": 0.5718,
+      "step": 9597
+    },
+    {
+      "epoch": 88.06,
+      "learning_rate": 3.7788461538461534e-05,
+      "loss": 0.5379,
+      "step": 9598
+    },
+    {
+      "epoch": 88.06,
+      "learning_rate": 3.7759615384615386e-05,
+      "loss": 0.4793,
+      "step": 9599
+    },
+    {
+      "epoch": 88.07,
+      "learning_rate": 3.773076923076923e-05,
+      "loss": 0.6771,
+      "step": 9600
+    },
+    {
+      "epoch": 88.08,
+      "learning_rate": 3.770192307692307e-05,
+      "loss": 0.5851,
+      "step": 9601
+    },
+    {
+      "epoch": 88.09,
+      "learning_rate": 3.767307692307692e-05,
+      "loss": 0.4875,
+      "step": 9602
+    },
+    {
+      "epoch": 88.1,
+      "learning_rate": 3.7644230769230765e-05,
+      "loss": 0.5595,
+      "step": 9603
+    },
+    {
+      "epoch": 88.11,
+      "learning_rate": 3.761538461538461e-05,
+      "loss": 0.5256,
+      "step": 9604
+    },
+    {
+      "epoch": 88.12,
+      "learning_rate": 3.758653846153846e-05,
+      "loss": 0.5388,
+      "step": 9605
+    },
+    {
+      "epoch": 88.13,
+      "learning_rate": 3.75576923076923e-05,
+      "loss": 0.4754,
+      "step": 9606
+    },
+    {
+      "epoch": 88.14,
+      "learning_rate": 3.752884615384615e-05,
+      "loss": 0.4555,
+      "step": 9607
+    },
+    {
+      "epoch": 88.15,
+      "learning_rate": 3.75e-05,
+      "loss": 0.5026,
+      "step": 9608
+    },
+    {
+      "epoch": 88.16,
+      "learning_rate": 3.747115384615384e-05,
+      "loss": 0.4405,
+      "step": 9609
+    },
+    {
+      "epoch": 88.17,
+      "learning_rate": 3.744230769230769e-05,
+      "loss": 0.4507,
+      "step": 9610
+    },
+    {
+      "epoch": 88.17,
+      "learning_rate": 3.741346153846153e-05,
+      "loss": 0.5189,
+      "step": 9611
+    },
+    {
+      "epoch": 88.18,
+      "learning_rate": 3.738461538461538e-05,
+      "loss": 0.5879,
+      "step": 9612
+    },
+    {
+      "epoch": 88.19,
+      "learning_rate": 3.735576923076923e-05,
+      "loss": 0.5225,
+      "step": 9613
+    },
+    {
+      "epoch": 88.2,
+      "learning_rate": 3.732692307692307e-05,
+      "loss": 0.4853,
+      "step": 9614
+    },
+    {
+      "epoch": 88.21,
+      "learning_rate": 3.729807692307692e-05,
+      "loss": 0.5036,
+      "step": 9615
+    },
+    {
+      "epoch": 88.22,
+      "learning_rate": 3.726923076923076e-05,
+      "loss": 0.6257,
+      "step": 9616
+    },
+    {
+      "epoch": 88.23,
+      "learning_rate": 3.7240384615384615e-05,
+      "loss": 0.5751,
+      "step": 9617
+    },
+    {
+      "epoch": 88.24,
+      "learning_rate": 3.721153846153846e-05,
+      "loss": 0.5357,
+      "step": 9618
+    },
+    {
+      "epoch": 88.25,
+      "learning_rate": 3.7182692307692305e-05,
+      "loss": 0.5701,
+      "step": 9619
+    },
+    {
+      "epoch": 88.26,
+      "learning_rate": 3.715384615384615e-05,
+      "loss": 0.542,
+      "step": 9620
+    },
+    {
+      "epoch": 88.27,
+      "learning_rate": 3.7124999999999994e-05,
+      "loss": 0.6166,
+      "step": 9621
+    },
+    {
+      "epoch": 88.28,
+      "learning_rate": 3.7096153846153846e-05,
+      "loss": 0.6385,
+      "step": 9622
+    },
+    {
+      "epoch": 88.28,
+      "learning_rate": 3.706730769230769e-05,
+      "loss": 0.5014,
+      "step": 9623
+    },
+    {
+      "epoch": 88.29,
+      "learning_rate": 3.7038461538461536e-05,
+      "loss": 0.5889,
+      "step": 9624
+    },
+    {
+      "epoch": 88.3,
+      "learning_rate": 3.700961538461538e-05,
+      "loss": 0.5647,
+      "step": 9625
+    },
+    {
+      "epoch": 88.31,
+      "learning_rate": 3.6980769230769226e-05,
+      "loss": 0.5251,
+      "step": 9626
+    },
+    {
+      "epoch": 88.32,
+      "learning_rate": 3.695192307692308e-05,
+      "loss": 0.4505,
+      "step": 9627
+    },
+    {
+      "epoch": 88.33,
+      "learning_rate": 3.692307692307692e-05,
+      "loss": 0.4581,
+      "step": 9628
+    },
+    {
+      "epoch": 88.34,
+      "learning_rate": 3.689423076923077e-05,
+      "loss": 0.4238,
+      "step": 9629
+    },
+    {
+      "epoch": 88.35,
+      "learning_rate": 3.686538461538461e-05,
+      "loss": 0.5833,
+      "step": 9630
+    },
+    {
+      "epoch": 88.36,
+      "learning_rate": 3.683653846153846e-05,
+      "loss": 0.502,
+      "step": 9631
+    },
+    {
+      "epoch": 88.37,
+      "learning_rate": 3.680769230769231e-05,
+      "loss": 0.5473,
+      "step": 9632
+    },
+    {
+      "epoch": 88.38,
+      "learning_rate": 3.677884615384615e-05,
+      "loss": 0.4242,
+      "step": 9633
+    },
+    {
+      "epoch": 88.39,
+      "learning_rate": 3.675e-05,
+      "loss": 0.5308,
+      "step": 9634
+    },
+    {
+      "epoch": 88.39,
+      "learning_rate": 3.6721153846153844e-05,
+      "loss": 0.4873,
+      "step": 9635
+    },
+    {
+      "epoch": 88.4,
+      "learning_rate": 3.669230769230769e-05,
+      "loss": 0.641,
+      "step": 9636
+    },
+    {
+      "epoch": 88.41,
+      "learning_rate": 3.6663461538461534e-05,
+      "loss": 0.5936,
+      "step": 9637
+    },
+    {
+      "epoch": 88.42,
+      "learning_rate": 3.663461538461538e-05,
+      "loss": 0.4974,
+      "step": 9638
+    },
+    {
+      "epoch": 88.43,
+      "learning_rate": 3.660576923076923e-05,
+      "loss": 0.4659,
+      "step": 9639
+    },
+    {
+      "epoch": 88.44,
+      "learning_rate": 3.6576923076923075e-05,
+      "loss": 0.4981,
+      "step": 9640
+    },
+    {
+      "epoch": 88.45,
+      "learning_rate": 3.654807692307692e-05,
+      "loss": 0.6319,
+      "step": 9641
+    },
+    {
+      "epoch": 88.46,
+      "learning_rate": 3.6519230769230765e-05,
+      "loss": 0.6756,
+      "step": 9642
+    },
+    {
+      "epoch": 88.47,
+      "learning_rate": 3.649038461538461e-05,
+      "loss": 0.5674,
+      "step": 9643
+    },
+    {
+      "epoch": 88.48,
+      "learning_rate": 3.646153846153846e-05,
+      "loss": 0.5398,
+      "step": 9644
+    },
+    {
+      "epoch": 88.49,
+      "learning_rate": 3.6432692307692307e-05,
+      "loss": 0.4594,
+      "step": 9645
+    },
+    {
+      "epoch": 88.5,
+      "learning_rate": 3.640384615384615e-05,
+      "loss": 0.5689,
+      "step": 9646
+    },
+    {
+      "epoch": 88.5,
+      "learning_rate": 3.6374999999999996e-05,
+      "loss": 0.4981,
+      "step": 9647
+    },
+    {
+      "epoch": 88.51,
+      "learning_rate": 3.634615384615384e-05,
+      "loss": 0.5023,
+      "step": 9648
+    },
+    {
+      "epoch": 88.52,
+      "learning_rate": 3.631730769230769e-05,
+      "loss": 0.6406,
+      "step": 9649
+    },
+    {
+      "epoch": 88.53,
+      "learning_rate": 3.628846153846154e-05,
+      "loss": 0.5541,
+      "step": 9650
+    },
+    {
+      "epoch": 88.54,
+      "learning_rate": 3.625961538461538e-05,
+      "loss": 0.5314,
+      "step": 9651
+    },
+    {
+      "epoch": 88.55,
+      "learning_rate": 3.623076923076923e-05,
+      "loss": 0.5517,
+      "step": 9652
+    },
+    {
+      "epoch": 88.56,
+      "learning_rate": 3.620192307692307e-05,
+      "loss": 0.5111,
+      "step": 9653
+    },
+    {
+      "epoch": 88.57,
+      "learning_rate": 3.6173076923076925e-05,
+      "loss": 0.559,
+      "step": 9654
+    },
+    {
+      "epoch": 88.58,
+      "learning_rate": 3.614423076923076e-05,
+      "loss": 0.5108,
+      "step": 9655
+    },
+    {
+      "epoch": 88.59,
+      "learning_rate": 3.6115384615384614e-05,
+      "loss": 0.5652,
+      "step": 9656
+    },
+    {
+      "epoch": 88.6,
+      "learning_rate": 3.608653846153846e-05,
+      "loss": 0.6073,
+      "step": 9657
+    },
+    {
+      "epoch": 88.61,
+      "learning_rate": 3.6057692307692304e-05,
+      "loss": 0.5458,
+      "step": 9658
+    },
+    {
+      "epoch": 88.61,
+      "learning_rate": 3.602884615384615e-05,
+      "loss": 0.5901,
+      "step": 9659
+    },
+    {
+      "epoch": 88.62,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 0.5232,
+      "step": 9660
+    },
+    {
+      "epoch": 88.63,
+      "learning_rate": 3.5971153846153846e-05,
+      "loss": 0.5009,
+      "step": 9661
+    },
+    {
+      "epoch": 88.64,
+      "learning_rate": 3.594230769230769e-05,
+      "loss": 0.5482,
+      "step": 9662
+    },
+    {
+      "epoch": 88.65,
+      "learning_rate": 3.5913461538461536e-05,
+      "loss": 0.5742,
+      "step": 9663
+    },
+    {
+      "epoch": 88.66,
+      "learning_rate": 3.588461538461538e-05,
+      "loss": 0.6927,
+      "step": 9664
+    },
+    {
+      "epoch": 88.67,
+      "learning_rate": 3.5855769230769226e-05,
+      "loss": 0.5224,
+      "step": 9665
+    },
+    {
+      "epoch": 88.68,
+      "learning_rate": 3.582692307692308e-05,
+      "loss": 0.5744,
+      "step": 9666
+    },
+    {
+      "epoch": 88.69,
+      "learning_rate": 3.579807692307692e-05,
+      "loss": 0.4629,
+      "step": 9667
+    },
+    {
+      "epoch": 88.7,
+      "learning_rate": 3.576923076923077e-05,
+      "loss": 0.6104,
+      "step": 9668
+    },
+    {
+      "epoch": 88.71,
+      "learning_rate": 3.574038461538461e-05,
+      "loss": 0.4433,
+      "step": 9669
+    },
+    {
+      "epoch": 88.72,
+      "learning_rate": 3.571153846153846e-05,
+      "loss": 0.5632,
+      "step": 9670
+    },
+    {
+      "epoch": 88.72,
+      "learning_rate": 3.56826923076923e-05,
+      "loss": 0.5155,
+      "step": 9671
+    },
+    {
+      "epoch": 88.73,
+      "learning_rate": 3.5653846153846154e-05,
+      "loss": 0.5153,
+      "step": 9672
+    },
+    {
+      "epoch": 88.74,
+      "learning_rate": 3.5625e-05,
+      "loss": 0.5445,
+      "step": 9673
+    },
+    {
+      "epoch": 88.75,
+      "learning_rate": 3.5596153846153844e-05,
+      "loss": 0.4775,
+      "step": 9674
+    },
+    {
+      "epoch": 88.76,
+      "learning_rate": 3.556730769230769e-05,
+      "loss": 0.4663,
+      "step": 9675
+    },
+    {
+      "epoch": 88.77,
+      "learning_rate": 3.553846153846153e-05,
+      "loss": 0.4543,
+      "step": 9676
+    },
+    {
+      "epoch": 88.78,
+      "learning_rate": 3.5509615384615385e-05,
+      "loss": 0.4812,
+      "step": 9677
+    },
+    {
+      "epoch": 88.79,
+      "learning_rate": 3.548076923076922e-05,
+      "loss": 0.5533,
+      "step": 9678
+    },
+    {
+      "epoch": 88.8,
+      "learning_rate": 3.5451923076923075e-05,
+      "loss": 0.5006,
+      "step": 9679
+    },
+    {
+      "epoch": 88.81,
+      "learning_rate": 3.542307692307692e-05,
+      "loss": 0.4286,
+      "step": 9680
+    },
+    {
+      "epoch": 88.82,
+      "learning_rate": 3.5394230769230765e-05,
+      "loss": 0.5048,
+      "step": 9681
+    },
+    {
+      "epoch": 88.83,
+      "learning_rate": 3.536538461538461e-05,
+      "loss": 0.4003,
+      "step": 9682
+    },
+    {
+      "epoch": 88.83,
+      "learning_rate": 3.5336538461538455e-05,
+      "loss": 0.4987,
+      "step": 9683
+    },
+    {
+      "epoch": 88.84,
+      "learning_rate": 3.5307692307692306e-05,
+      "loss": 0.4862,
+      "step": 9684
+    },
+    {
+      "epoch": 88.85,
+      "learning_rate": 3.527884615384615e-05,
+      "loss": 0.3961,
+      "step": 9685
+    },
+    {
+      "epoch": 88.86,
+      "learning_rate": 3.5249999999999996e-05,
+      "loss": 0.4511,
+      "step": 9686
+    },
+    {
+      "epoch": 88.87,
+      "learning_rate": 3.522115384615384e-05,
+      "loss": 0.5424,
+      "step": 9687
+    },
+    {
+      "epoch": 88.88,
+      "learning_rate": 3.5192307692307686e-05,
+      "loss": 0.6107,
+      "step": 9688
+    },
+    {
+      "epoch": 88.89,
+      "learning_rate": 3.516346153846154e-05,
+      "loss": 0.5652,
+      "step": 9689
+    },
+    {
+      "epoch": 88.9,
+      "learning_rate": 3.513461538461538e-05,
+      "loss": 0.5262,
+      "step": 9690
+    },
+    {
+      "epoch": 88.91,
+      "learning_rate": 3.510576923076923e-05,
+      "loss": 0.5429,
+      "step": 9691
+    },
+    {
+      "epoch": 88.92,
+      "learning_rate": 3.507692307692307e-05,
+      "loss": 0.6542,
+      "step": 9692
+    },
+    {
+      "epoch": 88.93,
+      "learning_rate": 3.504807692307692e-05,
+      "loss": 0.4559,
+      "step": 9693
+    },
+    {
+      "epoch": 88.94,
+      "learning_rate": 3.501923076923077e-05,
+      "loss": 0.5193,
+      "step": 9694
+    },
+    {
+      "epoch": 88.94,
+      "learning_rate": 3.499038461538461e-05,
+      "loss": 0.4405,
+      "step": 9695
+    },
+    {
+      "epoch": 88.95,
+      "learning_rate": 3.496153846153846e-05,
+      "loss": 0.52,
+      "step": 9696
+    },
+    {
+      "epoch": 88.96,
+      "learning_rate": 3.4932692307692304e-05,
+      "loss": 0.5198,
+      "step": 9697
+    },
+    {
+      "epoch": 88.97,
+      "learning_rate": 3.490384615384615e-05,
+      "loss": 0.4501,
+      "step": 9698
+    },
+    {
+      "epoch": 88.98,
+      "learning_rate": 3.4875e-05,
+      "loss": 0.4646,
+      "step": 9699
+    },
+    {
+      "epoch": 88.99,
+      "learning_rate": 3.484615384615384e-05,
+      "loss": 0.534,
+      "step": 9700
+    },
+    {
+      "epoch": 89.0,
+      "learning_rate": 3.481730769230769e-05,
+      "loss": 0.4617,
+      "step": 9701
+    },
+    {
+      "epoch": 89.01,
+      "learning_rate": 3.4788461538461535e-05,
+      "loss": 0.5055,
+      "step": 9702
+    },
+    {
+      "epoch": 89.02,
+      "learning_rate": 3.475961538461538e-05,
+      "loss": 0.5173,
+      "step": 9703
+    },
+    {
+      "epoch": 89.03,
+      "learning_rate": 3.4730769230769225e-05,
+      "loss": 0.4953,
+      "step": 9704
+    },
+    {
+      "epoch": 89.04,
+      "learning_rate": 3.470192307692307e-05,
+      "loss": 0.5476,
+      "step": 9705
+    },
+    {
+      "epoch": 89.05,
+      "learning_rate": 3.467307692307692e-05,
+      "loss": 0.5066,
+      "step": 9706
+    },
+    {
+      "epoch": 89.06,
+      "learning_rate": 3.464423076923077e-05,
+      "loss": 0.4948,
+      "step": 9707
+    },
+    {
+      "epoch": 89.06,
+      "learning_rate": 3.461538461538461e-05,
+      "loss": 0.5016,
+      "step": 9708
+    },
+    {
+      "epoch": 89.07,
+      "learning_rate": 3.458653846153846e-05,
+      "loss": 0.499,
+      "step": 9709
+    },
+    {
+      "epoch": 89.08,
+      "learning_rate": 3.45576923076923e-05,
+      "loss": 0.5402,
+      "step": 9710
+    },
+    {
+      "epoch": 89.09,
+      "learning_rate": 3.4528846153846153e-05,
+      "loss": 0.5609,
+      "step": 9711
+    },
+    {
+      "epoch": 89.1,
+      "learning_rate": 3.45e-05,
+      "loss": 0.5338,
+      "step": 9712
+    },
+    {
+      "epoch": 89.11,
+      "learning_rate": 3.447115384615384e-05,
+      "loss": 0.4525,
+      "step": 9713
+    },
+    {
+      "epoch": 89.12,
+      "learning_rate": 3.444230769230769e-05,
+      "loss": 0.4606,
+      "step": 9714
+    },
+    {
+      "epoch": 89.13,
+      "learning_rate": 3.441346153846153e-05,
+      "loss": 0.4922,
+      "step": 9715
+    },
+    {
+      "epoch": 89.14,
+      "learning_rate": 3.4384615384615385e-05,
+      "loss": 0.6006,
+      "step": 9716
+    },
+    {
+      "epoch": 89.15,
+      "learning_rate": 3.435576923076923e-05,
+      "loss": 0.575,
+      "step": 9717
+    },
+    {
+      "epoch": 89.16,
+      "learning_rate": 3.4326923076923075e-05,
+      "loss": 0.5818,
+      "step": 9718
+    },
+    {
+      "epoch": 89.17,
+      "learning_rate": 3.429807692307692e-05,
+      "loss": 0.4764,
+      "step": 9719
+    },
+    {
+      "epoch": 89.17,
+      "learning_rate": 3.4269230769230765e-05,
+      "loss": 0.368,
+      "step": 9720
+    },
+    {
+      "epoch": 89.18,
+      "learning_rate": 3.4240384615384616e-05,
+      "loss": 0.5802,
+      "step": 9721
+    },
+    {
+      "epoch": 89.19,
+      "learning_rate": 3.4211538461538454e-05,
+      "loss": 0.5158,
+      "step": 9722
+    },
+    {
+      "epoch": 89.2,
+      "learning_rate": 3.4182692307692306e-05,
+      "loss": 0.5649,
+      "step": 9723
+    },
+    {
+      "epoch": 89.21,
+      "learning_rate": 3.415384615384615e-05,
+      "loss": 0.5743,
+      "step": 9724
+    },
+    {
+      "epoch": 89.22,
+      "learning_rate": 3.4124999999999996e-05,
+      "loss": 0.572,
+      "step": 9725
+    },
+    {
+      "epoch": 89.23,
+      "learning_rate": 3.409615384615384e-05,
+      "loss": 0.4825,
+      "step": 9726
+    },
+    {
+      "epoch": 89.24,
+      "learning_rate": 3.4067307692307686e-05,
+      "loss": 0.5925,
+      "step": 9727
+    },
+    {
+      "epoch": 89.25,
+      "learning_rate": 3.403846153846154e-05,
+      "loss": 0.5488,
+      "step": 9728
+    },
+    {
+      "epoch": 89.26,
+      "learning_rate": 3.400961538461538e-05,
+      "loss": 0.6049,
+      "step": 9729
+    },
+    {
+      "epoch": 89.27,
+      "learning_rate": 3.398076923076923e-05,
+      "loss": 0.5378,
+      "step": 9730
+    },
+    {
+      "epoch": 89.28,
+      "learning_rate": 3.395192307692307e-05,
+      "loss": 0.5991,
+      "step": 9731
+    },
+    {
+      "epoch": 89.28,
+      "learning_rate": 3.392307692307692e-05,
+      "loss": 0.4877,
+      "step": 9732
+    },
+    {
+      "epoch": 89.29,
+      "learning_rate": 3.389423076923077e-05,
+      "loss": 0.4006,
+      "step": 9733
+    },
+    {
+      "epoch": 89.3,
+      "learning_rate": 3.3865384615384614e-05,
+      "loss": 0.4453,
+      "step": 9734
+    },
+    {
+      "epoch": 89.31,
+      "learning_rate": 3.383653846153846e-05,
+      "loss": 0.4925,
+      "step": 9735
+    },
+    {
+      "epoch": 89.32,
+      "learning_rate": 3.3807692307692304e-05,
+      "loss": 0.598,
+      "step": 9736
+    },
+    {
+      "epoch": 89.33,
+      "learning_rate": 3.377884615384615e-05,
+      "loss": 0.4545,
+      "step": 9737
+    },
+    {
+      "epoch": 89.34,
+      "learning_rate": 3.375e-05,
+      "loss": 0.4056,
+      "step": 9738
+    },
+    {
+      "epoch": 89.35,
+      "learning_rate": 3.3721153846153845e-05,
+      "loss": 0.6402,
+      "step": 9739
+    },
+    {
+      "epoch": 89.36,
+      "learning_rate": 3.369230769230769e-05,
+      "loss": 0.454,
+      "step": 9740
+    },
+    {
+      "epoch": 89.37,
+      "learning_rate": 3.3663461538461535e-05,
+      "loss": 0.4567,
+      "step": 9741
+    },
+    {
+      "epoch": 89.38,
+      "learning_rate": 3.363461538461538e-05,
+      "loss": 0.5355,
+      "step": 9742
+    },
+    {
+      "epoch": 89.39,
+      "learning_rate": 3.360576923076923e-05,
+      "loss": 0.5426,
+      "step": 9743
+    },
+    {
+      "epoch": 89.39,
+      "learning_rate": 3.357692307692307e-05,
+      "loss": 0.5101,
+      "step": 9744
+    },
+    {
+      "epoch": 89.4,
+      "learning_rate": 3.354807692307692e-05,
+      "loss": 0.6327,
+      "step": 9745
+    },
+    {
+      "epoch": 89.41,
+      "learning_rate": 3.351923076923077e-05,
+      "loss": 0.4927,
+      "step": 9746
+    },
+    {
+      "epoch": 89.42,
+      "learning_rate": 3.349038461538461e-05,
+      "loss": 0.582,
+      "step": 9747
+    },
+    {
+      "epoch": 89.43,
+      "learning_rate": 3.346153846153846e-05,
+      "loss": 0.5008,
+      "step": 9748
+    },
+    {
+      "epoch": 89.44,
+      "learning_rate": 3.34326923076923e-05,
+      "loss": 0.5,
+      "step": 9749
+    },
+    {
+      "epoch": 89.45,
+      "learning_rate": 3.340384615384615e-05,
+      "loss": 0.512,
+      "step": 9750
+    },
+    {
+      "epoch": 89.46,
+      "learning_rate": 3.3375e-05,
+      "loss": 0.4489,
+      "step": 9751
+    },
+    {
+      "epoch": 89.47,
+      "learning_rate": 3.334615384615384e-05,
+      "loss": 0.6177,
+      "step": 9752
+    },
+    {
+      "epoch": 89.48,
+      "learning_rate": 3.331730769230769e-05,
+      "loss": 0.4588,
+      "step": 9753
+    },
+    {
+      "epoch": 89.49,
+      "learning_rate": 3.328846153846153e-05,
+      "loss": 0.5394,
+      "step": 9754
+    },
+    {
+      "epoch": 89.5,
+      "learning_rate": 3.3259615384615385e-05,
+      "loss": 0.5318,
+      "step": 9755
+    },
+    {
+      "epoch": 89.5,
+      "learning_rate": 3.323076923076923e-05,
+      "loss": 0.4983,
+      "step": 9756
+    },
+    {
+      "epoch": 89.51,
+      "learning_rate": 3.3201923076923074e-05,
+      "loss": 0.5071,
+      "step": 9757
+    },
+    {
+      "epoch": 89.52,
+      "learning_rate": 3.317307692307692e-05,
+      "loss": 0.6059,
+      "step": 9758
+    },
+    {
+      "epoch": 89.53,
+      "learning_rate": 3.3144230769230764e-05,
+      "loss": 0.5475,
+      "step": 9759
+    },
+    {
+      "epoch": 89.54,
+      "learning_rate": 3.3115384615384616e-05,
+      "loss": 0.4348,
+      "step": 9760
+    },
+    {
+      "epoch": 89.55,
+      "learning_rate": 3.308653846153846e-05,
+      "loss": 0.5719,
+      "step": 9761
+    },
+    {
+      "epoch": 89.56,
+      "learning_rate": 3.3057692307692306e-05,
+      "loss": 0.516,
+      "step": 9762
+    },
+    {
+      "epoch": 89.57,
+      "learning_rate": 3.302884615384615e-05,
+      "loss": 0.5464,
+      "step": 9763
+    },
+    {
+      "epoch": 89.58,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 0.485,
+      "step": 9764
+    },
+    {
+      "epoch": 89.59,
+      "learning_rate": 3.297115384615385e-05,
+      "loss": 0.4925,
+      "step": 9765
+    },
+    {
+      "epoch": 89.6,
+      "learning_rate": 3.294230769230769e-05,
+      "loss": 0.4933,
+      "step": 9766
+    },
+    {
+      "epoch": 89.61,
+      "learning_rate": 3.291346153846154e-05,
+      "loss": 0.4999,
+      "step": 9767
+    },
+    {
+      "epoch": 89.61,
+      "learning_rate": 3.288461538461538e-05,
+      "loss": 0.5292,
+      "step": 9768
+    },
+    {
+      "epoch": 89.62,
+      "learning_rate": 3.285576923076923e-05,
+      "loss": 0.5658,
+      "step": 9769
+    },
+    {
+      "epoch": 89.63,
+      "learning_rate": 3.282692307692308e-05,
+      "loss": 0.5512,
+      "step": 9770
+    },
+    {
+      "epoch": 89.64,
+      "learning_rate": 3.279807692307692e-05,
+      "loss": 0.5657,
+      "step": 9771
+    },
+    {
+      "epoch": 89.65,
+      "learning_rate": 3.276923076923077e-05,
+      "loss": 0.4385,
+      "step": 9772
+    },
+    {
+      "epoch": 89.66,
+      "learning_rate": 3.2740384615384614e-05,
+      "loss": 0.5393,
+      "step": 9773
+    },
+    {
+      "epoch": 89.67,
+      "learning_rate": 3.271153846153846e-05,
+      "loss": 0.4815,
+      "step": 9774
+    },
+    {
+      "epoch": 89.68,
+      "learning_rate": 3.2682692307692304e-05,
+      "loss": 0.562,
+      "step": 9775
+    },
+    {
+      "epoch": 89.69,
+      "learning_rate": 3.265384615384615e-05,
+      "loss": 0.3516,
+      "step": 9776
+    },
+    {
+      "epoch": 89.7,
+      "learning_rate": 3.2625e-05,
+      "loss": 0.5253,
+      "step": 9777
+    },
+    {
+      "epoch": 89.71,
+      "learning_rate": 3.2596153846153845e-05,
+      "loss": 0.5449,
+      "step": 9778
+    },
+    {
+      "epoch": 89.72,
+      "learning_rate": 3.256730769230769e-05,
+      "loss": 0.5429,
+      "step": 9779
+    },
+    {
+      "epoch": 89.72,
+      "learning_rate": 3.2538461538461535e-05,
+      "loss": 0.5814,
+      "step": 9780
+    },
+    {
+      "epoch": 89.73,
+      "learning_rate": 3.250961538461538e-05,
+      "loss": 0.4661,
+      "step": 9781
+    },
+    {
+      "epoch": 89.74,
+      "learning_rate": 3.248076923076923e-05,
+      "loss": 0.5667,
+      "step": 9782
+    },
+    {
+      "epoch": 89.75,
+      "learning_rate": 3.2451923076923077e-05,
+      "loss": 0.4898,
+      "step": 9783
+    },
+    {
+      "epoch": 89.76,
+      "learning_rate": 3.242307692307692e-05,
+      "loss": 0.5321,
+      "step": 9784
+    },
+    {
+      "epoch": 89.77,
+      "learning_rate": 3.2394230769230766e-05,
+      "loss": 0.5331,
+      "step": 9785
+    },
+    {
+      "epoch": 89.78,
+      "learning_rate": 3.236538461538461e-05,
+      "loss": 0.5664,
+      "step": 9786
+    },
+    {
+      "epoch": 89.79,
+      "learning_rate": 3.2336538461538456e-05,
+      "loss": 0.559,
+      "step": 9787
+    },
+    {
+      "epoch": 89.8,
+      "learning_rate": 3.230769230769231e-05,
+      "loss": 0.4119,
+      "step": 9788
+    },
+    {
+      "epoch": 89.81,
+      "learning_rate": 3.227884615384615e-05,
+      "loss": 0.5586,
+      "step": 9789
+    },
+    {
+      "epoch": 89.82,
+      "learning_rate": 3.225e-05,
+      "loss": 0.5497,
+      "step": 9790
+    },
+    {
+      "epoch": 89.83,
+      "learning_rate": 3.222115384615384e-05,
+      "loss": 0.5824,
+      "step": 9791
+    },
+    {
+      "epoch": 89.83,
+      "learning_rate": 3.219230769230769e-05,
+      "loss": 0.6702,
+      "step": 9792
+    },
+    {
+      "epoch": 89.84,
+      "learning_rate": 3.216346153846153e-05,
+      "loss": 0.5826,
+      "step": 9793
+    },
+    {
+      "epoch": 89.85,
+      "learning_rate": 3.213461538461538e-05,
+      "loss": 0.4853,
+      "step": 9794
+    },
+    {
+      "epoch": 89.86,
+      "learning_rate": 3.210576923076923e-05,
+      "loss": 0.5469,
+      "step": 9795
+    },
+    {
+      "epoch": 89.87,
+      "learning_rate": 3.2076923076923074e-05,
+      "loss": 0.5001,
+      "step": 9796
+    },
+    {
+      "epoch": 89.88,
+      "learning_rate": 3.204807692307692e-05,
+      "loss": 0.4933,
+      "step": 9797
+    },
+    {
+      "epoch": 89.89,
+      "learning_rate": 3.2019230769230764e-05,
+      "loss": 0.5405,
+      "step": 9798
+    },
+    {
+      "epoch": 89.9,
+      "learning_rate": 3.199038461538461e-05,
+      "loss": 0.5273,
+      "step": 9799
+    },
+    {
+      "epoch": 89.91,
+      "learning_rate": 3.196153846153846e-05,
+      "loss": 0.547,
+      "step": 9800
+    },
+    {
+      "epoch": 89.92,
+      "learning_rate": 3.1932692307692306e-05,
+      "loss": 0.4513,
+      "step": 9801
+    },
+    {
+      "epoch": 89.93,
+      "learning_rate": 3.190384615384615e-05,
+      "loss": 0.6073,
+      "step": 9802
+    },
+    {
+      "epoch": 89.94,
+      "learning_rate": 3.1874999999999996e-05,
+      "loss": 0.4552,
+      "step": 9803
+    },
+    {
+      "epoch": 89.94,
+      "learning_rate": 3.184615384615384e-05,
+      "loss": 0.4927,
+      "step": 9804
+    },
+    {
+      "epoch": 89.95,
+      "learning_rate": 3.181730769230769e-05,
+      "loss": 0.5958,
+      "step": 9805
+    },
+    {
+      "epoch": 89.96,
+      "learning_rate": 3.178846153846154e-05,
+      "loss": 0.5112,
+      "step": 9806
+    },
+    {
+      "epoch": 89.97,
+      "learning_rate": 3.175961538461538e-05,
+      "loss": 0.4944,
+      "step": 9807
+    },
+    {
+      "epoch": 89.98,
+      "learning_rate": 3.173076923076923e-05,
+      "loss": 0.4482,
+      "step": 9808
+    },
+    {
+      "epoch": 89.99,
+      "learning_rate": 3.170192307692307e-05,
+      "loss": 0.5369,
+      "step": 9809
+    },
+    {
+      "epoch": 90.0,
+      "learning_rate": 3.1673076923076924e-05,
+      "loss": 0.4699,
+      "step": 9810
+    },
+    {
+      "epoch": 90.01,
+      "learning_rate": 3.164423076923076e-05,
+      "loss": 0.629,
+      "step": 9811
+    },
+    {
+      "epoch": 90.02,
+      "learning_rate": 3.1615384615384613e-05,
+      "loss": 0.5189,
+      "step": 9812
+    },
+    {
+      "epoch": 90.03,
+      "learning_rate": 3.158653846153846e-05,
+      "loss": 0.4932,
+      "step": 9813
+    },
+    {
+      "epoch": 90.04,
+      "learning_rate": 3.15576923076923e-05,
+      "loss": 0.4712,
+      "step": 9814
+    },
+    {
+      "epoch": 90.05,
+      "learning_rate": 3.1528846153846155e-05,
+      "loss": 0.5493,
+      "step": 9815
+    },
+    {
+      "epoch": 90.06,
+      "learning_rate": 3.149999999999999e-05,
+      "loss": 0.5682,
+      "step": 9816
+    },
+    {
+      "epoch": 90.06,
+      "learning_rate": 3.1471153846153845e-05,
+      "loss": 0.4857,
+      "step": 9817
+    },
+    {
+      "epoch": 90.07,
+      "learning_rate": 3.144230769230769e-05,
+      "loss": 0.4363,
+      "step": 9818
+    },
+    {
+      "epoch": 90.08,
+      "learning_rate": 3.1413461538461535e-05,
+      "loss": 0.3946,
+      "step": 9819
+    },
+    {
+      "epoch": 90.09,
+      "learning_rate": 3.138461538461538e-05,
+      "loss": 0.5097,
+      "step": 9820
+    },
+    {
+      "epoch": 90.1,
+      "learning_rate": 3.1355769230769225e-05,
+      "loss": 0.5549,
+      "step": 9821
+    },
+    {
+      "epoch": 90.11,
+      "learning_rate": 3.1326923076923076e-05,
+      "loss": 0.6007,
+      "step": 9822
+    },
+    {
+      "epoch": 90.12,
+      "learning_rate": 3.129807692307692e-05,
+      "loss": 0.5437,
+      "step": 9823
+    },
+    {
+      "epoch": 90.13,
+      "learning_rate": 3.1269230769230766e-05,
+      "loss": 0.5367,
+      "step": 9824
+    },
+    {
+      "epoch": 90.14,
+      "learning_rate": 3.124038461538461e-05,
+      "loss": 0.4221,
+      "step": 9825
+    },
+    {
+      "epoch": 90.15,
+      "learning_rate": 3.1211538461538456e-05,
+      "loss": 0.4925,
+      "step": 9826
+    },
+    {
+      "epoch": 90.16,
+      "learning_rate": 3.118269230769231e-05,
+      "loss": 0.509,
+      "step": 9827
+    },
+    {
+      "epoch": 90.17,
+      "learning_rate": 3.115384615384615e-05,
+      "loss": 0.4971,
+      "step": 9828
+    },
+    {
+      "epoch": 90.17,
+      "learning_rate": 3.1125e-05,
+      "loss": 0.4856,
+      "step": 9829
+    },
+    {
+      "epoch": 90.18,
+      "learning_rate": 3.109615384615384e-05,
+      "loss": 0.4814,
+      "step": 9830
+    },
+    {
+      "epoch": 90.19,
+      "learning_rate": 3.106730769230769e-05,
+      "loss": 0.4641,
+      "step": 9831
+    },
+    {
+      "epoch": 90.2,
+      "learning_rate": 3.103846153846154e-05,
+      "loss": 0.4175,
+      "step": 9832
+    },
+    {
+      "epoch": 90.21,
+      "learning_rate": 3.100961538461538e-05,
+      "loss": 0.5011,
+      "step": 9833
+    },
+    {
+      "epoch": 90.22,
+      "learning_rate": 3.098076923076923e-05,
+      "loss": 0.5196,
+      "step": 9834
+    },
+    {
+      "epoch": 90.23,
+      "learning_rate": 3.0951923076923074e-05,
+      "loss": 0.3837,
+      "step": 9835
+    },
+    {
+      "epoch": 90.24,
+      "learning_rate": 3.092307692307692e-05,
+      "loss": 0.5463,
+      "step": 9836
+    },
+    {
+      "epoch": 90.25,
+      "learning_rate": 3.089423076923077e-05,
+      "loss": 0.4689,
+      "step": 9837
+    },
+    {
+      "epoch": 90.26,
+      "learning_rate": 3.086538461538461e-05,
+      "loss": 0.462,
+      "step": 9838
+    },
+    {
+      "epoch": 90.27,
+      "learning_rate": 3.083653846153846e-05,
+      "loss": 0.4451,
+      "step": 9839
+    },
+    {
+      "epoch": 90.28,
+      "learning_rate": 3.0807692307692305e-05,
+      "loss": 0.4569,
+      "step": 9840
+    },
+    {
+      "epoch": 90.28,
+      "learning_rate": 3.077884615384615e-05,
+      "loss": 0.4632,
+      "step": 9841
+    },
+    {
+      "epoch": 90.29,
+      "learning_rate": 3.0749999999999995e-05,
+      "loss": 0.5576,
+      "step": 9842
+    },
+    {
+      "epoch": 90.3,
+      "learning_rate": 3.072115384615384e-05,
+      "loss": 0.5077,
+      "step": 9843
+    },
+    {
+      "epoch": 90.31,
+      "learning_rate": 3.069230769230769e-05,
+      "loss": 0.4909,
+      "step": 9844
+    },
+    {
+      "epoch": 90.32,
+      "learning_rate": 3.066346153846154e-05,
+      "loss": 0.5441,
+      "step": 9845
+    },
+    {
+      "epoch": 90.33,
+      "learning_rate": 3.063461538461538e-05,
+      "loss": 0.5001,
+      "step": 9846
+    },
+    {
+      "epoch": 90.34,
+      "learning_rate": 3.060576923076923e-05,
+      "loss": 0.4971,
+      "step": 9847
+    },
+    {
+      "epoch": 90.35,
+      "learning_rate": 3.057692307692307e-05,
+      "loss": 0.4287,
+      "step": 9848
+    },
+    {
+      "epoch": 90.36,
+      "learning_rate": 3.054807692307692e-05,
+      "loss": 0.4566,
+      "step": 9849
+    },
+    {
+      "epoch": 90.37,
+      "learning_rate": 3.051923076923077e-05,
+      "loss": 0.487,
+      "step": 9850
+    },
+    {
+      "epoch": 90.38,
+      "learning_rate": 3.049038461538461e-05,
+      "loss": 0.4752,
+      "step": 9851
+    },
+    {
+      "epoch": 90.39,
+      "learning_rate": 3.0461538461538458e-05,
+      "loss": 0.4109,
+      "step": 9852
+    },
+    {
+      "epoch": 90.39,
+      "learning_rate": 3.0432692307692306e-05,
+      "loss": 0.5339,
+      "step": 9853
+    },
+    {
+      "epoch": 90.4,
+      "learning_rate": 3.040384615384615e-05,
+      "loss": 0.5614,
+      "step": 9854
+    },
+    {
+      "epoch": 90.41,
+      "learning_rate": 3.0375e-05,
+      "loss": 0.4866,
+      "step": 9855
+    },
+    {
+      "epoch": 90.42,
+      "learning_rate": 3.034615384615384e-05,
+      "loss": 0.5137,
+      "step": 9856
+    },
+    {
+      "epoch": 90.43,
+      "learning_rate": 3.031730769230769e-05,
+      "loss": 0.4057,
+      "step": 9857
+    },
+    {
+      "epoch": 90.44,
+      "learning_rate": 3.0288461538461538e-05,
+      "loss": 0.5446,
+      "step": 9858
+    },
+    {
+      "epoch": 90.45,
+      "learning_rate": 3.0259615384615383e-05,
+      "loss": 0.5904,
+      "step": 9859
+    },
+    {
+      "epoch": 90.46,
+      "learning_rate": 3.0230769230769228e-05,
+      "loss": 0.4836,
+      "step": 9860
+    },
+    {
+      "epoch": 90.47,
+      "learning_rate": 3.0201923076923073e-05,
+      "loss": 0.6764,
+      "step": 9861
+    },
+    {
+      "epoch": 90.48,
+      "learning_rate": 3.017307692307692e-05,
+      "loss": 0.5117,
+      "step": 9862
+    },
+    {
+      "epoch": 90.49,
+      "learning_rate": 3.014423076923077e-05,
+      "loss": 0.5504,
+      "step": 9863
+    },
+    {
+      "epoch": 90.5,
+      "learning_rate": 3.011538461538461e-05,
+      "loss": 0.5075,
+      "step": 9864
+    },
+    {
+      "epoch": 90.5,
+      "learning_rate": 3.008653846153846e-05,
+      "loss": 0.5597,
+      "step": 9865
+    },
+    {
+      "epoch": 90.51,
+      "learning_rate": 3.0057692307692304e-05,
+      "loss": 0.4278,
+      "step": 9866
+    },
+    {
+      "epoch": 90.52,
+      "learning_rate": 3.0028846153846152e-05,
+      "loss": 0.547,
+      "step": 9867
+    },
+    {
+      "epoch": 90.53,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 0.5221,
+      "step": 9868
+    },
+    {
+      "epoch": 90.54,
+      "learning_rate": 2.9971153846153842e-05,
+      "loss": 0.6526,
+      "step": 9869
+    },
+    {
+      "epoch": 90.55,
+      "learning_rate": 2.994230769230769e-05,
+      "loss": 0.5851,
+      "step": 9870
+    },
+    {
+      "epoch": 90.56,
+      "learning_rate": 2.9913461538461536e-05,
+      "loss": 0.6377,
+      "step": 9871
+    },
+    {
+      "epoch": 90.57,
+      "learning_rate": 2.9884615384615384e-05,
+      "loss": 0.5079,
+      "step": 9872
+    },
+    {
+      "epoch": 90.58,
+      "learning_rate": 2.9855769230769225e-05,
+      "loss": 0.6133,
+      "step": 9873
+    },
+    {
+      "epoch": 90.59,
+      "learning_rate": 2.9826923076923074e-05,
+      "loss": 0.5786,
+      "step": 9874
+    },
+    {
+      "epoch": 90.6,
+      "learning_rate": 2.979807692307692e-05,
+      "loss": 0.5328,
+      "step": 9875
+    },
+    {
+      "epoch": 90.61,
+      "learning_rate": 2.9769230769230767e-05,
+      "loss": 0.5243,
+      "step": 9876
+    },
+    {
+      "epoch": 90.61,
+      "learning_rate": 2.9740384615384615e-05,
+      "loss": 0.5586,
+      "step": 9877
+    },
+    {
+      "epoch": 90.62,
+      "learning_rate": 2.9711538461538457e-05,
+      "loss": 0.4964,
+      "step": 9878
+    },
+    {
+      "epoch": 90.63,
+      "learning_rate": 2.9682692307692305e-05,
+      "loss": 0.4565,
+      "step": 9879
+    },
+    {
+      "epoch": 90.64,
+      "learning_rate": 2.965384615384615e-05,
+      "loss": 0.5102,
+      "step": 9880
+    },
+    {
+      "epoch": 90.65,
+      "learning_rate": 2.9625e-05,
+      "loss": 0.4036,
+      "step": 9881
+    },
+    {
+      "epoch": 90.66,
+      "learning_rate": 2.9596153846153843e-05,
+      "loss": 0.5274,
+      "step": 9882
+    },
+    {
+      "epoch": 90.67,
+      "learning_rate": 2.9567307692307688e-05,
+      "loss": 0.5659,
+      "step": 9883
+    },
+    {
+      "epoch": 90.68,
+      "learning_rate": 2.9538461538461537e-05,
+      "loss": 0.4054,
+      "step": 9884
+    },
+    {
+      "epoch": 90.69,
+      "learning_rate": 2.950961538461538e-05,
+      "loss": 0.5925,
+      "step": 9885
+    },
+    {
+      "epoch": 90.7,
+      "learning_rate": 2.948076923076923e-05,
+      "loss": 0.4998,
+      "step": 9886
+    },
+    {
+      "epoch": 90.71,
+      "learning_rate": 2.945192307692307e-05,
+      "loss": 0.5643,
+      "step": 9887
+    },
+    {
+      "epoch": 90.72,
+      "learning_rate": 2.942307692307692e-05,
+      "loss": 0.5277,
+      "step": 9888
+    },
+    {
+      "epoch": 90.72,
+      "learning_rate": 2.9394230769230768e-05,
+      "loss": 0.5059,
+      "step": 9889
+    },
+    {
+      "epoch": 90.73,
+      "learning_rate": 2.9365384615384613e-05,
+      "loss": 0.5242,
+      "step": 9890
+    },
+    {
+      "epoch": 90.74,
+      "learning_rate": 2.9336538461538458e-05,
+      "loss": 0.6365,
+      "step": 9891
+    },
+    {
+      "epoch": 90.75,
+      "learning_rate": 2.9307692307692303e-05,
+      "loss": 0.4806,
+      "step": 9892
+    },
+    {
+      "epoch": 90.76,
+      "learning_rate": 2.927884615384615e-05,
+      "loss": 0.492,
+      "step": 9893
+    },
+    {
+      "epoch": 90.77,
+      "learning_rate": 2.925e-05,
+      "loss": 0.4436,
+      "step": 9894
+    },
+    {
+      "epoch": 90.78,
+      "learning_rate": 2.9221153846153844e-05,
+      "loss": 0.5818,
+      "step": 9895
+    },
+    {
+      "epoch": 90.79,
+      "learning_rate": 2.919230769230769e-05,
+      "loss": 0.5601,
+      "step": 9896
+    },
+    {
+      "epoch": 90.8,
+      "learning_rate": 2.9163461538461534e-05,
+      "loss": 0.4669,
+      "step": 9897
+    },
+    {
+      "epoch": 90.81,
+      "learning_rate": 2.9134615384615383e-05,
+      "loss": 0.5508,
+      "step": 9898
+    },
+    {
+      "epoch": 90.82,
+      "learning_rate": 2.910576923076923e-05,
+      "loss": 0.5789,
+      "step": 9899
+    },
+    {
+      "epoch": 90.83,
+      "learning_rate": 2.9076923076923072e-05,
+      "loss": 0.5901,
+      "step": 9900
+    },
+    {
+      "epoch": 90.83,
+      "learning_rate": 2.904807692307692e-05,
+      "loss": 0.5752,
+      "step": 9901
+    },
+    {
+      "epoch": 90.84,
+      "learning_rate": 2.9019230769230766e-05,
+      "loss": 0.5099,
+      "step": 9902
+    },
+    {
+      "epoch": 90.85,
+      "learning_rate": 2.8990384615384614e-05,
+      "loss": 0.4198,
+      "step": 9903
+    },
+    {
+      "epoch": 90.86,
+      "learning_rate": 2.8961538461538462e-05,
+      "loss": 0.4417,
+      "step": 9904
+    },
+    {
+      "epoch": 90.87,
+      "learning_rate": 2.8932692307692304e-05,
+      "loss": 0.4854,
+      "step": 9905
+    },
+    {
+      "epoch": 90.88,
+      "learning_rate": 2.8903846153846152e-05,
+      "loss": 0.5978,
+      "step": 9906
+    },
+    {
+      "epoch": 90.89,
+      "learning_rate": 2.8874999999999997e-05,
+      "loss": 0.562,
+      "step": 9907
+    },
+    {
+      "epoch": 90.9,
+      "learning_rate": 2.8846153846153845e-05,
+      "loss": 0.4841,
+      "step": 9908
+    },
+    {
+      "epoch": 90.91,
+      "learning_rate": 2.8817307692307687e-05,
+      "loss": 0.5224,
+      "step": 9909
+    },
+    {
+      "epoch": 90.92,
+      "learning_rate": 2.8788461538461535e-05,
+      "loss": 0.4105,
+      "step": 9910
+    },
+    {
+      "epoch": 90.93,
+      "learning_rate": 2.8759615384615384e-05,
+      "loss": 0.4424,
+      "step": 9911
+    },
+    {
+      "epoch": 90.94,
+      "learning_rate": 2.873076923076923e-05,
+      "loss": 0.4572,
+      "step": 9912
+    },
+    {
+      "epoch": 90.94,
+      "learning_rate": 2.8701923076923073e-05,
+      "loss": 0.5274,
+      "step": 9913
+    },
+    {
+      "epoch": 90.95,
+      "learning_rate": 2.867307692307692e-05,
+      "loss": 0.5847,
+      "step": 9914
+    },
+    {
+      "epoch": 90.96,
+      "learning_rate": 2.8644230769230767e-05,
+      "loss": 0.4694,
+      "step": 9915
+    },
+    {
+      "epoch": 90.97,
+      "learning_rate": 2.8615384615384615e-05,
+      "loss": 0.5413,
+      "step": 9916
+    },
+    {
+      "epoch": 90.98,
+      "learning_rate": 2.858653846153846e-05,
+      "loss": 0.4234,
+      "step": 9917
+    },
+    {
+      "epoch": 90.99,
+      "learning_rate": 2.8557692307692305e-05,
+      "loss": 0.5706,
+      "step": 9918
+    },
+    {
+      "epoch": 91.0,
+      "learning_rate": 2.852884615384615e-05,
+      "loss": 0.4771,
+      "step": 9919
+    },
+    {
+      "epoch": 91.01,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 0.5609,
+      "step": 9920
+    },
+    {
+      "epoch": 91.02,
+      "learning_rate": 2.8471153846153847e-05,
+      "loss": 0.4928,
+      "step": 9921
+    },
+    {
+      "epoch": 91.03,
+      "learning_rate": 2.8442307692307688e-05,
+      "loss": 0.4309,
+      "step": 9922
+    },
+    {
+      "epoch": 91.04,
+      "learning_rate": 2.8413461538461536e-05,
+      "loss": 0.5798,
+      "step": 9923
+    },
+    {
+      "epoch": 91.05,
+      "learning_rate": 2.838461538461538e-05,
+      "loss": 0.5447,
+      "step": 9924
+    },
+    {
+      "epoch": 91.06,
+      "learning_rate": 2.835576923076923e-05,
+      "loss": 0.4764,
+      "step": 9925
+    },
+    {
+      "epoch": 91.06,
+      "learning_rate": 2.8326923076923075e-05,
+      "loss": 0.5054,
+      "step": 9926
+    },
+    {
+      "epoch": 91.07,
+      "learning_rate": 2.829807692307692e-05,
+      "loss": 0.5724,
+      "step": 9927
+    },
+    {
+      "epoch": 91.08,
+      "learning_rate": 2.8269230769230768e-05,
+      "loss": 0.5111,
+      "step": 9928
+    },
+    {
+      "epoch": 91.09,
+      "learning_rate": 2.8240384615384613e-05,
+      "loss": 0.656,
+      "step": 9929
+    },
+    {
+      "epoch": 91.1,
+      "learning_rate": 2.821153846153846e-05,
+      "loss": 0.5639,
+      "step": 9930
+    },
+    {
+      "epoch": 91.11,
+      "learning_rate": 2.8182692307692303e-05,
+      "loss": 0.4972,
+      "step": 9931
+    },
+    {
+      "epoch": 91.12,
+      "learning_rate": 2.815384615384615e-05,
+      "loss": 0.5235,
+      "step": 9932
+    },
+    {
+      "epoch": 91.13,
+      "learning_rate": 2.8125e-05,
+      "loss": 0.4466,
+      "step": 9933
+    },
+    {
+      "epoch": 91.14,
+      "learning_rate": 2.8096153846153844e-05,
+      "loss": 0.455,
+      "step": 9934
+    },
+    {
+      "epoch": 91.15,
+      "learning_rate": 2.8067307692307692e-05,
+      "loss": 0.5569,
+      "step": 9935
+    },
+    {
+      "epoch": 91.16,
+      "learning_rate": 2.8038461538461534e-05,
+      "loss": 0.5075,
+      "step": 9936
+    },
+    {
+      "epoch": 91.17,
+      "learning_rate": 2.8009615384615382e-05,
+      "loss": 0.5712,
+      "step": 9937
+    },
+    {
+      "epoch": 91.17,
+      "learning_rate": 2.7980769230769227e-05,
+      "loss": 0.5146,
+      "step": 9938
+    },
+    {
+      "epoch": 91.18,
+      "learning_rate": 2.7951923076923076e-05,
+      "loss": 0.6184,
+      "step": 9939
+    },
+    {
+      "epoch": 91.19,
+      "learning_rate": 2.792307692307692e-05,
+      "loss": 0.4745,
+      "step": 9940
+    },
+    {
+      "epoch": 91.2,
+      "learning_rate": 2.7894230769230765e-05,
+      "loss": 0.5312,
+      "step": 9941
+    },
+    {
+      "epoch": 91.21,
+      "learning_rate": 2.7865384615384614e-05,
+      "loss": 0.6312,
+      "step": 9942
+    },
+    {
+      "epoch": 91.22,
+      "learning_rate": 2.783653846153846e-05,
+      "loss": 0.4769,
+      "step": 9943
+    },
+    {
+      "epoch": 91.23,
+      "learning_rate": 2.7807692307692307e-05,
+      "loss": 0.5258,
+      "step": 9944
+    },
+    {
+      "epoch": 91.24,
+      "learning_rate": 2.777884615384615e-05,
+      "loss": 0.5038,
+      "step": 9945
+    },
+    {
+      "epoch": 91.25,
+      "learning_rate": 2.7749999999999997e-05,
+      "loss": 0.4106,
+      "step": 9946
+    },
+    {
+      "epoch": 91.26,
+      "learning_rate": 2.7721153846153845e-05,
+      "loss": 0.5272,
+      "step": 9947
+    },
+    {
+      "epoch": 91.27,
+      "learning_rate": 2.769230769230769e-05,
+      "loss": 0.5201,
+      "step": 9948
+    },
+    {
+      "epoch": 91.28,
+      "learning_rate": 2.7663461538461535e-05,
+      "loss": 0.5881,
+      "step": 9949
+    },
+    {
+      "epoch": 91.28,
+      "learning_rate": 2.763461538461538e-05,
+      "loss": 0.5213,
+      "step": 9950
+    },
+    {
+      "epoch": 91.29,
+      "learning_rate": 2.760576923076923e-05,
+      "loss": 0.472,
+      "step": 9951
+    },
+    {
+      "epoch": 91.3,
+      "learning_rate": 2.7576923076923077e-05,
+      "loss": 0.5769,
+      "step": 9952
+    },
+    {
+      "epoch": 91.31,
+      "learning_rate": 2.754807692307692e-05,
+      "loss": 0.5168,
+      "step": 9953
+    },
+    {
+      "epoch": 91.32,
+      "learning_rate": 2.7519230769230767e-05,
+      "loss": 0.5099,
+      "step": 9954
+    },
+    {
+      "epoch": 91.33,
+      "learning_rate": 2.749038461538461e-05,
+      "loss": 0.4948,
+      "step": 9955
+    },
+    {
+      "epoch": 91.34,
+      "learning_rate": 2.746153846153846e-05,
+      "loss": 0.5092,
+      "step": 9956
+    },
+    {
+      "epoch": 91.35,
+      "learning_rate": 2.7432692307692308e-05,
+      "loss": 0.5441,
+      "step": 9957
+    },
+    {
+      "epoch": 91.36,
+      "learning_rate": 2.740384615384615e-05,
+      "loss": 0.5731,
+      "step": 9958
+    },
+    {
+      "epoch": 91.37,
+      "learning_rate": 2.7374999999999998e-05,
+      "loss": 0.4685,
+      "step": 9959
+    },
+    {
+      "epoch": 91.38,
+      "learning_rate": 2.7346153846153843e-05,
+      "loss": 0.544,
+      "step": 9960
+    },
+    {
+      "epoch": 91.39,
+      "learning_rate": 2.731730769230769e-05,
+      "loss": 0.4963,
+      "step": 9961
+    },
+    {
+      "epoch": 91.39,
+      "learning_rate": 2.7288461538461533e-05,
+      "loss": 0.5533,
+      "step": 9962
+    },
+    {
+      "epoch": 91.4,
+      "learning_rate": 2.725961538461538e-05,
+      "loss": 0.6058,
+      "step": 9963
+    },
+    {
+      "epoch": 91.41,
+      "learning_rate": 2.723076923076923e-05,
+      "loss": 0.405,
+      "step": 9964
+    },
+    {
+      "epoch": 91.42,
+      "learning_rate": 2.7201923076923074e-05,
+      "loss": 0.5265,
+      "step": 9965
+    },
+    {
+      "epoch": 91.43,
+      "learning_rate": 2.7173076923076923e-05,
+      "loss": 0.5213,
+      "step": 9966
+    },
+    {
+      "epoch": 91.44,
+      "learning_rate": 2.7144230769230764e-05,
+      "loss": 0.4401,
+      "step": 9967
+    },
+    {
+      "epoch": 91.45,
+      "learning_rate": 2.7115384615384612e-05,
+      "loss": 0.4267,
+      "step": 9968
+    },
+    {
+      "epoch": 91.46,
+      "learning_rate": 2.708653846153846e-05,
+      "loss": 0.3502,
+      "step": 9969
+    },
+    {
+      "epoch": 91.47,
+      "learning_rate": 2.7057692307692306e-05,
+      "loss": 0.6128,
+      "step": 9970
+    },
+    {
+      "epoch": 91.48,
+      "learning_rate": 2.702884615384615e-05,
+      "loss": 0.5688,
+      "step": 9971
+    },
+    {
+      "epoch": 91.49,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 0.5907,
+      "step": 9972
+    },
+    {
+      "epoch": 91.5,
+      "learning_rate": 2.6971153846153844e-05,
+      "loss": 0.5213,
+      "step": 9973
+    },
+    {
+      "epoch": 91.5,
+      "learning_rate": 2.6942307692307692e-05,
+      "loss": 0.4939,
+      "step": 9974
+    },
+    {
+      "epoch": 91.51,
+      "learning_rate": 2.6913461538461537e-05,
+      "loss": 0.4875,
+      "step": 9975
+    },
+    {
+      "epoch": 91.52,
+      "learning_rate": 2.6884615384615382e-05,
+      "loss": 0.5623,
+      "step": 9976
+    },
+    {
+      "epoch": 91.53,
+      "learning_rate": 2.6855769230769227e-05,
+      "loss": 0.5581,
+      "step": 9977
+    },
+    {
+      "epoch": 91.54,
+      "learning_rate": 2.6826923076923075e-05,
+      "loss": 0.5758,
+      "step": 9978
+    },
+    {
+      "epoch": 91.55,
+      "learning_rate": 2.6798076923076924e-05,
+      "loss": 0.517,
+      "step": 9979
+    },
+    {
+      "epoch": 91.56,
+      "learning_rate": 2.6769230769230765e-05,
+      "loss": 0.5697,
+      "step": 9980
+    },
+    {
+      "epoch": 91.57,
+      "learning_rate": 2.6740384615384614e-05,
+      "loss": 0.528,
+      "step": 9981
+    },
+    {
+      "epoch": 91.58,
+      "learning_rate": 2.671153846153846e-05,
+      "loss": 0.4431,
+      "step": 9982
+    },
+    {
+      "epoch": 91.59,
+      "learning_rate": 2.6682692307692307e-05,
+      "loss": 0.4607,
+      "step": 9983
+    },
+    {
+      "epoch": 91.6,
+      "learning_rate": 2.6653846153846152e-05,
+      "loss": 0.5042,
+      "step": 9984
+    },
+    {
+      "epoch": 91.61,
+      "learning_rate": 2.6624999999999997e-05,
+      "loss": 0.4685,
+      "step": 9985
+    },
+    {
+      "epoch": 91.61,
+      "learning_rate": 2.6596153846153845e-05,
+      "loss": 0.475,
+      "step": 9986
+    },
+    {
+      "epoch": 91.62,
+      "learning_rate": 2.656730769230769e-05,
+      "loss": 0.4436,
+      "step": 9987
+    },
+    {
+      "epoch": 91.63,
+      "learning_rate": 2.6538461538461538e-05,
+      "loss": 0.474,
+      "step": 9988
+    },
+    {
+      "epoch": 91.64,
+      "learning_rate": 2.650961538461538e-05,
+      "loss": 0.5143,
+      "step": 9989
+    },
+    {
+      "epoch": 91.65,
+      "learning_rate": 2.6480769230769228e-05,
+      "loss": 0.4633,
+      "step": 9990
+    },
+    {
+      "epoch": 91.66,
+      "learning_rate": 2.6451923076923076e-05,
+      "loss": 0.5397,
+      "step": 9991
+    },
+    {
+      "epoch": 91.67,
+      "learning_rate": 2.642307692307692e-05,
+      "loss": 0.4742,
+      "step": 9992
+    },
+    {
+      "epoch": 91.68,
+      "learning_rate": 2.639423076923077e-05,
+      "loss": 0.4266,
+      "step": 9993
+    },
+    {
+      "epoch": 91.69,
+      "learning_rate": 2.636538461538461e-05,
+      "loss": 0.4722,
+      "step": 9994
+    },
+    {
+      "epoch": 91.7,
+      "learning_rate": 2.633653846153846e-05,
+      "loss": 0.6405,
+      "step": 9995
+    },
+    {
+      "epoch": 91.71,
+      "learning_rate": 2.6307692307692304e-05,
+      "loss": 0.4998,
+      "step": 9996
+    },
+    {
+      "epoch": 91.72,
+      "learning_rate": 2.6278846153846153e-05,
+      "loss": 0.4402,
+      "step": 9997
+    },
+    {
+      "epoch": 91.72,
+      "learning_rate": 2.6249999999999998e-05,
+      "loss": 0.5542,
+      "step": 9998
+    },
+    {
+      "epoch": 91.73,
+      "learning_rate": 2.6221153846153843e-05,
+      "loss": 0.6076,
+      "step": 9999
+    },
+    {
+      "epoch": 91.74,
+      "learning_rate": 2.619230769230769e-05,
+      "loss": 0.3544,
+      "step": 10000
+    },
+    {
+      "epoch": 91.74,
+      "eval_cer": 0.08341954152236165,
+      "eval_loss": 0.3690292537212372,
+      "eval_runtime": 13.8967,
+      "eval_samples_per_second": 118.517,
+      "eval_steps_per_second": 1.871,
+      "eval_wer": 0.34950464712491064,
+      "step": 10000
+    },
+    {
+      "epoch": 91.75,
+      "learning_rate": 2.6163461538461536e-05,
+      "loss": 0.5176,
+      "step": 10001
+    },
+    {
+      "epoch": 91.76,
+      "learning_rate": 2.6134615384615384e-05,
+      "loss": 0.4773,
+      "step": 10002
+    },
+    {
+      "epoch": 91.77,
+      "learning_rate": 2.6134615384615384e-05,
+      "loss": 0.5721,
+      "step": 10003
+    },
+    {
+      "epoch": 91.78,
+      "learning_rate": 2.6105769230769226e-05,
+      "loss": 0.6251,
+      "step": 10004
+    },
+    {
+      "epoch": 91.79,
+      "learning_rate": 2.6076923076923074e-05,
+      "loss": 0.6071,
+      "step": 10005
+    },
+    {
+      "epoch": 91.8,
+      "learning_rate": 2.6048076923076922e-05,
+      "loss": 0.4817,
+      "step": 10006
+    },
+    {
+      "epoch": 91.81,
+      "learning_rate": 2.6019230769230767e-05,
+      "loss": 0.5224,
+      "step": 10007
+    },
+    {
+      "epoch": 91.82,
+      "learning_rate": 2.5990384615384612e-05,
+      "loss": 0.4964,
+      "step": 10008
+    },
+    {
+      "epoch": 91.83,
+      "learning_rate": 2.5961538461538457e-05,
+      "loss": 0.4894,
+      "step": 10009
+    },
+    {
+      "epoch": 91.83,
+      "learning_rate": 2.5932692307692305e-05,
+      "loss": 0.44,
+      "step": 10010
+    },
+    {
+      "epoch": 91.84,
+      "learning_rate": 2.5903846153846154e-05,
+      "loss": 0.41,
+      "step": 10011
+    },
+    {
+      "epoch": 91.85,
+      "learning_rate": 2.5874999999999995e-05,
+      "loss": 0.5578,
+      "step": 10012
+    },
+    {
+      "epoch": 91.86,
+      "learning_rate": 2.5846153846153844e-05,
+      "loss": 0.405,
+      "step": 10013
+    },
+    {
+      "epoch": 91.87,
+      "learning_rate": 2.581730769230769e-05,
+      "loss": 0.5499,
+      "step": 10014
+    },
+    {
+      "epoch": 91.88,
+      "learning_rate": 2.5788461538461537e-05,
+      "loss": 0.5359,
+      "step": 10015
+    },
+    {
+      "epoch": 91.89,
+      "learning_rate": 2.5759615384615385e-05,
+      "loss": 0.5173,
+      "step": 10016
+    },
+    {
+      "epoch": 91.9,
+      "learning_rate": 2.5730769230769227e-05,
+      "loss": 0.5559,
+      "step": 10017
+    },
+    {
+      "epoch": 91.91,
+      "learning_rate": 2.5701923076923075e-05,
+      "loss": 0.5215,
+      "step": 10018
+    },
+    {
+      "epoch": 91.92,
+      "learning_rate": 2.567307692307692e-05,
+      "loss": 0.3916,
+      "step": 10019
+    },
+    {
+      "epoch": 91.93,
+      "learning_rate": 2.564423076923077e-05,
+      "loss": 0.5391,
+      "step": 10020
+    },
+    {
+      "epoch": 91.94,
+      "learning_rate": 2.561538461538461e-05,
+      "loss": 0.5677,
+      "step": 10021
+    },
+    {
+      "epoch": 91.94,
+      "learning_rate": 2.5586538461538458e-05,
+      "loss": 0.4006,
+      "step": 10022
+    },
+    {
+      "epoch": 91.95,
+      "learning_rate": 2.5557692307692307e-05,
+      "loss": 0.5416,
+      "step": 10023
+    },
+    {
+      "epoch": 91.96,
+      "learning_rate": 2.552884615384615e-05,
+      "loss": 0.4426,
+      "step": 10024
+    },
+    {
+      "epoch": 91.97,
+      "learning_rate": 2.55e-05,
+      "loss": 0.5558,
+      "step": 10025
+    },
+    {
+      "epoch": 91.98,
+      "learning_rate": 2.547115384615384e-05,
+      "loss": 0.441,
+      "step": 10026
+    },
+    {
+      "epoch": 91.99,
+      "learning_rate": 2.544230769230769e-05,
+      "loss": 0.5444,
+      "step": 10027
+    },
+    {
+      "epoch": 92.0,
+      "learning_rate": 2.5413461538461538e-05,
+      "loss": 0.4337,
+      "step": 10028
+    },
+    {
+      "epoch": 92.01,
+      "learning_rate": 2.5384615384615383e-05,
+      "loss": 0.5253,
+      "step": 10029
+    },
+    {
+      "epoch": 92.02,
+      "learning_rate": 2.5355769230769228e-05,
+      "loss": 0.3992,
+      "step": 10030
+    },
+    {
+      "epoch": 92.03,
+      "learning_rate": 2.5326923076923073e-05,
+      "loss": 0.5554,
+      "step": 10031
+    },
+    {
+      "epoch": 92.04,
+      "learning_rate": 2.529807692307692e-05,
+      "loss": 0.4362,
+      "step": 10032
+    },
+    {
+      "epoch": 92.05,
+      "learning_rate": 2.526923076923077e-05,
+      "loss": 0.5828,
+      "step": 10033
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 2.5240384615384614e-05,
+      "loss": 0.5025,
+      "step": 10034
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 2.521153846153846e-05,
+      "loss": 0.4046,
+      "step": 10035
+    },
+    {
+      "epoch": 92.07,
+      "learning_rate": 2.5182692307692304e-05,
+      "loss": 0.4501,
+      "step": 10036
+    },
+    {
+      "epoch": 92.08,
+      "learning_rate": 2.5153846153846153e-05,
+      "loss": 0.4759,
+      "step": 10037
+    },
+    {
+      "epoch": 92.09,
+      "learning_rate": 2.5125e-05,
+      "loss": 0.4586,
+      "step": 10038
+    },
+    {
+      "epoch": 92.1,
+      "learning_rate": 2.5096153846153842e-05,
+      "loss": 0.5938,
+      "step": 10039
+    },
+    {
+      "epoch": 92.11,
+      "learning_rate": 2.506730769230769e-05,
+      "loss": 0.3984,
+      "step": 10040
+    },
+    {
+      "epoch": 92.12,
+      "learning_rate": 2.5038461538461536e-05,
+      "loss": 0.4954,
+      "step": 10041
+    },
+    {
+      "epoch": 92.13,
+      "learning_rate": 2.5009615384615384e-05,
+      "loss": 0.4618,
+      "step": 10042
+    },
+    {
+      "epoch": 92.14,
+      "learning_rate": 2.4980769230769232e-05,
+      "loss": 0.4316,
+      "step": 10043
+    },
+    {
+      "epoch": 92.15,
+      "learning_rate": 2.4951923076923074e-05,
+      "loss": 0.41,
+      "step": 10044
+    },
+    {
+      "epoch": 92.16,
+      "learning_rate": 2.4923076923076922e-05,
+      "loss": 0.4952,
+      "step": 10045
+    },
+    {
+      "epoch": 92.17,
+      "learning_rate": 2.4894230769230767e-05,
+      "loss": 0.588,
+      "step": 10046
+    },
+    {
+      "epoch": 92.17,
+      "learning_rate": 2.4865384615384615e-05,
+      "loss": 0.5823,
+      "step": 10047
+    },
+    {
+      "epoch": 92.18,
+      "learning_rate": 2.4836538461538457e-05,
+      "loss": 0.4808,
+      "step": 10048
+    },
+    {
+      "epoch": 92.19,
+      "learning_rate": 2.4807692307692305e-05,
+      "loss": 0.5606,
+      "step": 10049
+    },
+    {
+      "epoch": 92.2,
+      "learning_rate": 2.4778846153846154e-05,
+      "loss": 0.4579,
+      "step": 10050
+    },
+    {
+      "epoch": 92.21,
+      "learning_rate": 2.475e-05,
+      "loss": 0.5661,
+      "step": 10051
+    },
+    {
+      "epoch": 92.22,
+      "learning_rate": 2.4721153846153843e-05,
+      "loss": 0.4643,
+      "step": 10052
+    },
+    {
+      "epoch": 92.23,
+      "learning_rate": 2.469230769230769e-05,
+      "loss": 0.4829,
+      "step": 10053
+    },
+    {
+      "epoch": 92.24,
+      "learning_rate": 2.4663461538461537e-05,
+      "loss": 0.6202,
+      "step": 10054
+    },
+    {
+      "epoch": 92.25,
+      "learning_rate": 2.463461538461538e-05,
+      "loss": 0.4853,
+      "step": 10055
+    },
+    {
+      "epoch": 92.26,
+      "learning_rate": 2.460576923076923e-05,
+      "loss": 0.49,
+      "step": 10056
+    },
+    {
+      "epoch": 92.27,
+      "learning_rate": 2.4576923076923075e-05,
+      "loss": 0.454,
+      "step": 10057
+    },
+    {
+      "epoch": 92.28,
+      "learning_rate": 2.454807692307692e-05,
+      "loss": 0.4963,
+      "step": 10058
+    },
+    {
+      "epoch": 92.28,
+      "learning_rate": 2.4519230769230768e-05,
+      "loss": 0.5186,
+      "step": 10059
+    },
+    {
+      "epoch": 92.29,
+      "learning_rate": 2.4490384615384613e-05,
+      "loss": 0.5179,
+      "step": 10060
+    },
+    {
+      "epoch": 92.3,
+      "learning_rate": 2.4461538461538458e-05,
+      "loss": 0.4819,
+      "step": 10061
+    },
+    {
+      "epoch": 92.31,
+      "learning_rate": 2.4432692307692303e-05,
+      "loss": 0.5501,
+      "step": 10062
+    },
+    {
+      "epoch": 92.32,
+      "learning_rate": 2.440384615384615e-05,
+      "loss": 0.4471,
+      "step": 10063
+    },
+    {
+      "epoch": 92.33,
+      "learning_rate": 2.4375e-05,
+      "loss": 0.5032,
+      "step": 10064
+    },
+    {
+      "epoch": 92.34,
+      "learning_rate": 2.4346153846153844e-05,
+      "loss": 0.5365,
+      "step": 10065
+    },
+    {
+      "epoch": 92.35,
+      "learning_rate": 2.431730769230769e-05,
+      "loss": 0.4495,
+      "step": 10066
+    },
+    {
+      "epoch": 92.36,
+      "learning_rate": 2.4288461538461534e-05,
+      "loss": 0.553,
+      "step": 10067
+    },
+    {
+      "epoch": 92.37,
+      "learning_rate": 2.4259615384615383e-05,
+      "loss": 0.5144,
+      "step": 10068
+    },
+    {
+      "epoch": 92.38,
+      "learning_rate": 2.423076923076923e-05,
+      "loss": 0.4943,
+      "step": 10069
+    },
+    {
+      "epoch": 92.39,
+      "learning_rate": 2.4201923076923073e-05,
+      "loss": 0.566,
+      "step": 10070
+    },
+    {
+      "epoch": 92.39,
+      "learning_rate": 2.417307692307692e-05,
+      "loss": 0.4514,
+      "step": 10071
+    },
+    {
+      "epoch": 92.4,
+      "learning_rate": 2.4144230769230766e-05,
+      "loss": 0.3776,
+      "step": 10072
+    },
+    {
+      "epoch": 92.41,
+      "learning_rate": 2.4115384615384614e-05,
+      "loss": 0.4926,
+      "step": 10073
+    },
+    {
+      "epoch": 92.42,
+      "learning_rate": 2.4086538461538462e-05,
+      "loss": 0.4547,
+      "step": 10074
+    },
+    {
+      "epoch": 92.43,
+      "learning_rate": 2.4057692307692304e-05,
+      "loss": 0.5037,
+      "step": 10075
+    },
+    {
+      "epoch": 92.44,
+      "learning_rate": 2.4028846153846152e-05,
+      "loss": 0.4676,
+      "step": 10076
+    },
+    {
+      "epoch": 92.45,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 0.5975,
+      "step": 10077
+    },
+    {
+      "epoch": 92.46,
+      "learning_rate": 2.3971153846153846e-05,
+      "loss": 0.4229,
+      "step": 10078
+    },
+    {
+      "epoch": 92.47,
+      "learning_rate": 2.3942307692307687e-05,
+      "loss": 0.5966,
+      "step": 10079
+    },
+    {
+      "epoch": 92.48,
+      "learning_rate": 2.3913461538461535e-05,
+      "loss": 0.4459,
+      "step": 10080
+    },
+    {
+      "epoch": 92.49,
+      "learning_rate": 2.3884615384615384e-05,
+      "loss": 0.3851,
+      "step": 10081
+    },
+    {
+      "epoch": 92.5,
+      "learning_rate": 2.385576923076923e-05,
+      "loss": 0.5814,
+      "step": 10082
+    },
+    {
+      "epoch": 92.5,
+      "learning_rate": 2.3826923076923077e-05,
+      "loss": 0.4867,
+      "step": 10083
+    },
+    {
+      "epoch": 92.51,
+      "learning_rate": 2.379807692307692e-05,
+      "loss": 0.4357,
+      "step": 10084
+    },
+    {
+      "epoch": 92.52,
+      "learning_rate": 2.3769230769230767e-05,
+      "loss": 0.4942,
+      "step": 10085
+    },
+    {
+      "epoch": 92.53,
+      "learning_rate": 2.3740384615384615e-05,
+      "loss": 0.5603,
+      "step": 10086
+    },
+    {
+      "epoch": 92.54,
+      "learning_rate": 2.371153846153846e-05,
+      "loss": 0.5293,
+      "step": 10087
+    },
+    {
+      "epoch": 92.55,
+      "learning_rate": 2.3682692307692305e-05,
+      "loss": 0.6047,
+      "step": 10088
+    },
+    {
+      "epoch": 92.56,
+      "learning_rate": 2.365384615384615e-05,
+      "loss": 0.4443,
+      "step": 10089
+    },
+    {
+      "epoch": 92.57,
+      "learning_rate": 2.3624999999999998e-05,
+      "loss": 0.6818,
+      "step": 10090
+    },
+    {
+      "epoch": 92.58,
+      "learning_rate": 2.3596153846153847e-05,
+      "loss": 0.4804,
+      "step": 10091
+    },
+    {
+      "epoch": 92.59,
+      "learning_rate": 2.356730769230769e-05,
+      "loss": 0.5287,
+      "step": 10092
+    },
+    {
+      "epoch": 92.6,
+      "learning_rate": 2.3538461538461536e-05,
+      "loss": 0.475,
+      "step": 10093
+    },
+    {
+      "epoch": 92.61,
+      "learning_rate": 2.350961538461538e-05,
+      "loss": 0.3871,
+      "step": 10094
+    },
+    {
+      "epoch": 92.61,
+      "learning_rate": 2.348076923076923e-05,
+      "loss": 0.5352,
+      "step": 10095
+    },
+    {
+      "epoch": 92.62,
+      "learning_rate": 2.3451923076923078e-05,
+      "loss": 0.5186,
+      "step": 10096
+    },
+    {
+      "epoch": 92.63,
+      "learning_rate": 2.342307692307692e-05,
+      "loss": 0.4273,
+      "step": 10097
+    },
+    {
+      "epoch": 92.64,
+      "learning_rate": 2.3394230769230768e-05,
+      "loss": 0.5479,
+      "step": 10098
+    },
+    {
+      "epoch": 92.65,
+      "learning_rate": 2.3365384615384613e-05,
+      "loss": 0.4635,
+      "step": 10099
+    },
+    {
+      "epoch": 92.66,
+      "learning_rate": 2.333653846153846e-05,
+      "loss": 0.5216,
+      "step": 10100
+    },
+    {
+      "epoch": 92.67,
+      "learning_rate": 2.3307692307692303e-05,
+      "loss": 0.671,
+      "step": 10101
+    },
+    {
+      "epoch": 92.68,
+      "learning_rate": 2.327884615384615e-05,
+      "loss": 0.5697,
+      "step": 10102
+    },
+    {
+      "epoch": 92.69,
+      "learning_rate": 2.325e-05,
+      "loss": 0.4967,
+      "step": 10103
+    },
+    {
+      "epoch": 92.7,
+      "learning_rate": 2.3221153846153844e-05,
+      "loss": 0.6021,
+      "step": 10104
+    },
+    {
+      "epoch": 92.71,
+      "learning_rate": 2.3192307692307693e-05,
+      "loss": 0.4389,
+      "step": 10105
+    },
+    {
+      "epoch": 92.72,
+      "learning_rate": 2.3163461538461534e-05,
+      "loss": 0.4471,
+      "step": 10106
+    },
+    {
+      "epoch": 92.72,
+      "learning_rate": 2.3134615384615382e-05,
+      "loss": 0.5436,
+      "step": 10107
+    },
+    {
+      "epoch": 92.73,
+      "learning_rate": 2.310576923076923e-05,
+      "loss": 0.4831,
+      "step": 10108
+    },
+    {
+      "epoch": 92.74,
+      "learning_rate": 2.3076923076923076e-05,
+      "loss": 0.5174,
+      "step": 10109
+    },
+    {
+      "epoch": 92.75,
+      "learning_rate": 2.304807692307692e-05,
+      "loss": 0.4321,
+      "step": 10110
+    },
+    {
+      "epoch": 92.76,
+      "learning_rate": 2.3019230769230766e-05,
+      "loss": 0.4459,
+      "step": 10111
+    },
+    {
+      "epoch": 92.77,
+      "learning_rate": 2.2990384615384614e-05,
+      "loss": 0.4632,
+      "step": 10112
+    },
+    {
+      "epoch": 92.78,
+      "learning_rate": 2.296153846153846e-05,
+      "loss": 0.5027,
+      "step": 10113
+    },
+    {
+      "epoch": 92.79,
+      "learning_rate": 2.2932692307692307e-05,
+      "loss": 0.6499,
+      "step": 10114
+    },
+    {
+      "epoch": 92.8,
+      "learning_rate": 2.2903846153846152e-05,
+      "loss": 0.4937,
+      "step": 10115
+    },
+    {
+      "epoch": 92.81,
+      "learning_rate": 2.2874999999999997e-05,
+      "loss": 0.4624,
+      "step": 10116
+    },
+    {
+      "epoch": 92.82,
+      "learning_rate": 2.2846153846153845e-05,
+      "loss": 0.5931,
+      "step": 10117
+    },
+    {
+      "epoch": 92.83,
+      "learning_rate": 2.281730769230769e-05,
+      "loss": 0.6654,
+      "step": 10118
+    },
+    {
+      "epoch": 92.83,
+      "learning_rate": 2.2788461538461535e-05,
+      "loss": 0.503,
+      "step": 10119
+    },
+    {
+      "epoch": 92.84,
+      "learning_rate": 2.275961538461538e-05,
+      "loss": 0.5411,
+      "step": 10120
+    },
+    {
+      "epoch": 92.85,
+      "learning_rate": 2.273076923076923e-05,
+      "loss": 0.4673,
+      "step": 10121
+    },
+    {
+      "epoch": 92.86,
+      "learning_rate": 2.2701923076923077e-05,
+      "loss": 0.4454,
+      "step": 10122
+    },
+    {
+      "epoch": 92.87,
+      "learning_rate": 2.267307692307692e-05,
+      "loss": 0.5331,
+      "step": 10123
+    },
+    {
+      "epoch": 92.88,
+      "learning_rate": 2.2644230769230767e-05,
+      "loss": 0.4617,
+      "step": 10124
+    },
+    {
+      "epoch": 92.89,
+      "learning_rate": 2.261538461538461e-05,
+      "loss": 0.4945,
+      "step": 10125
+    },
+    {
+      "epoch": 92.9,
+      "learning_rate": 2.258653846153846e-05,
+      "loss": 0.4891,
+      "step": 10126
+    },
+    {
+      "epoch": 92.91,
+      "learning_rate": 2.2557692307692308e-05,
+      "loss": 0.4513,
+      "step": 10127
+    },
+    {
+      "epoch": 92.92,
+      "learning_rate": 2.252884615384615e-05,
+      "loss": 0.4122,
+      "step": 10128
+    },
+    {
+      "epoch": 92.93,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 0.5415,
+      "step": 10129
+    },
+    {
+      "epoch": 92.94,
+      "learning_rate": 2.2471153846153843e-05,
+      "loss": 0.5872,
+      "step": 10130
+    },
+    {
+      "epoch": 92.94,
+      "learning_rate": 2.244230769230769e-05,
+      "loss": 0.4821,
+      "step": 10131
+    },
+    {
+      "epoch": 92.95,
+      "learning_rate": 2.241346153846154e-05,
+      "loss": 0.4895,
+      "step": 10132
+    },
+    {
+      "epoch": 92.96,
+      "learning_rate": 2.238461538461538e-05,
+      "loss": 0.5203,
+      "step": 10133
+    },
+    {
+      "epoch": 92.97,
+      "learning_rate": 2.235576923076923e-05,
+      "loss": 0.5635,
+      "step": 10134
+    },
+    {
+      "epoch": 92.98,
+      "learning_rate": 2.2326923076923074e-05,
+      "loss": 0.5363,
+      "step": 10135
+    },
+    {
+      "epoch": 92.99,
+      "learning_rate": 2.2298076923076923e-05,
+      "loss": 0.5943,
+      "step": 10136
+    },
+    {
+      "epoch": 93.0,
+      "learning_rate": 2.2269230769230764e-05,
+      "loss": 0.4203,
+      "step": 10137
+    },
+    {
+      "epoch": 93.01,
+      "learning_rate": 2.2240384615384613e-05,
+      "loss": 0.5649,
+      "step": 10138
+    },
+    {
+      "epoch": 93.02,
+      "learning_rate": 2.221153846153846e-05,
+      "loss": 0.5841,
+      "step": 10139
+    },
+    {
+      "epoch": 93.03,
+      "learning_rate": 2.2182692307692306e-05,
+      "loss": 0.4898,
+      "step": 10140
+    },
+    {
+      "epoch": 93.04,
+      "learning_rate": 2.2153846153846154e-05,
+      "loss": 0.4082,
+      "step": 10141
+    },
+    {
+      "epoch": 93.05,
+      "learning_rate": 2.2124999999999996e-05,
+      "loss": 0.601,
+      "step": 10142
+    },
+    {
+      "epoch": 93.06,
+      "learning_rate": 2.2096153846153844e-05,
+      "loss": 0.4418,
+      "step": 10143
+    },
+    {
+      "epoch": 93.06,
+      "learning_rate": 2.2067307692307692e-05,
+      "loss": 0.5129,
+      "step": 10144
+    },
+    {
+      "epoch": 93.07,
+      "learning_rate": 2.2038461538461537e-05,
+      "loss": 0.4959,
+      "step": 10145
+    },
+    {
+      "epoch": 93.08,
+      "learning_rate": 2.2009615384615382e-05,
+      "loss": 0.4907,
+      "step": 10146
+    },
+    {
+      "epoch": 93.09,
+      "learning_rate": 2.1980769230769227e-05,
+      "loss": 0.4547,
+      "step": 10147
+    },
+    {
+      "epoch": 93.1,
+      "learning_rate": 2.1951923076923075e-05,
+      "loss": 0.4687,
+      "step": 10148
+    },
+    {
+      "epoch": 93.11,
+      "learning_rate": 2.1923076923076924e-05,
+      "loss": 0.4489,
+      "step": 10149
+    },
+    {
+      "epoch": 93.12,
+      "learning_rate": 2.1894230769230765e-05,
+      "loss": 0.5545,
+      "step": 10150
+    },
+    {
+      "epoch": 93.13,
+      "learning_rate": 2.1865384615384614e-05,
+      "loss": 0.6022,
+      "step": 10151
+    },
+    {
+      "epoch": 93.14,
+      "learning_rate": 2.183653846153846e-05,
+      "loss": 0.462,
+      "step": 10152
+    },
+    {
+      "epoch": 93.15,
+      "learning_rate": 2.1807692307692307e-05,
+      "loss": 0.4912,
+      "step": 10153
+    },
+    {
+      "epoch": 93.16,
+      "learning_rate": 2.1778846153846155e-05,
+      "loss": 0.4676,
+      "step": 10154
+    },
+    {
+      "epoch": 93.17,
+      "learning_rate": 2.1749999999999997e-05,
+      "loss": 0.5166,
+      "step": 10155
+    },
+    {
+      "epoch": 93.17,
+      "learning_rate": 2.1721153846153845e-05,
+      "loss": 0.403,
+      "step": 10156
+    },
+    {
+      "epoch": 93.18,
+      "learning_rate": 2.169230769230769e-05,
+      "loss": 0.5219,
+      "step": 10157
+    },
+    {
+      "epoch": 93.19,
+      "learning_rate": 2.166346153846154e-05,
+      "loss": 0.4708,
+      "step": 10158
+    },
+    {
+      "epoch": 93.2,
+      "learning_rate": 2.163461538461538e-05,
+      "loss": 0.4819,
+      "step": 10159
+    },
+    {
+      "epoch": 93.21,
+      "learning_rate": 2.1605769230769228e-05,
+      "loss": 0.4526,
+      "step": 10160
+    },
+    {
+      "epoch": 93.22,
+      "learning_rate": 2.1576923076923076e-05,
+      "loss": 0.5019,
+      "step": 10161
+    },
+    {
+      "epoch": 93.23,
+      "learning_rate": 2.154807692307692e-05,
+      "loss": 0.4363,
+      "step": 10162
+    },
+    {
+      "epoch": 93.24,
+      "learning_rate": 2.151923076923077e-05,
+      "loss": 0.5629,
+      "step": 10163
+    },
+    {
+      "epoch": 93.25,
+      "learning_rate": 2.149038461538461e-05,
+      "loss": 0.5304,
+      "step": 10164
+    },
+    {
+      "epoch": 93.26,
+      "learning_rate": 2.146153846153846e-05,
+      "loss": 0.5425,
+      "step": 10165
+    },
+    {
+      "epoch": 93.27,
+      "learning_rate": 2.1432692307692308e-05,
+      "loss": 0.4932,
+      "step": 10166
+    },
+    {
+      "epoch": 93.28,
+      "learning_rate": 2.1403846153846153e-05,
+      "loss": 0.6319,
+      "step": 10167
+    },
+    {
+      "epoch": 93.28,
+      "learning_rate": 2.1374999999999998e-05,
+      "loss": 0.5889,
+      "step": 10168
+    },
+    {
+      "epoch": 93.29,
+      "learning_rate": 2.1346153846153843e-05,
+      "loss": 0.6135,
+      "step": 10169
+    },
+    {
+      "epoch": 93.3,
+      "learning_rate": 2.131730769230769e-05,
+      "loss": 0.4907,
+      "step": 10170
+    },
+    {
+      "epoch": 93.31,
+      "learning_rate": 2.1288461538461536e-05,
+      "loss": 0.4906,
+      "step": 10171
+    },
+    {
+      "epoch": 93.32,
+      "learning_rate": 2.1259615384615384e-05,
+      "loss": 0.5689,
+      "step": 10172
+    },
+    {
+      "epoch": 93.33,
+      "learning_rate": 2.123076923076923e-05,
+      "loss": 0.498,
+      "step": 10173
+    },
+    {
+      "epoch": 93.34,
+      "learning_rate": 2.1201923076923074e-05,
+      "loss": 0.5115,
+      "step": 10174
+    },
+    {
+      "epoch": 93.35,
+      "learning_rate": 2.1173076923076922e-05,
+      "loss": 0.5023,
+      "step": 10175
+    },
+    {
+      "epoch": 93.36,
+      "learning_rate": 2.1144230769230767e-05,
+      "loss": 0.5092,
+      "step": 10176
+    },
+    {
+      "epoch": 93.37,
+      "learning_rate": 2.1115384615384612e-05,
+      "loss": 0.3918,
+      "step": 10177
+    },
+    {
+      "epoch": 93.38,
+      "learning_rate": 2.1086538461538457e-05,
+      "loss": 0.5353,
+      "step": 10178
+    },
+    {
+      "epoch": 93.39,
+      "learning_rate": 2.1057692307692306e-05,
+      "loss": 0.6214,
+      "step": 10179
+    },
+    {
+      "epoch": 93.39,
+      "learning_rate": 2.1028846153846154e-05,
+      "loss": 0.5299,
+      "step": 10180
+    },
+    {
+      "epoch": 93.4,
+      "learning_rate": 2.1e-05,
+      "loss": 0.468,
+      "step": 10181
+    },
+    {
+      "epoch": 93.41,
+      "learning_rate": 2.0971153846153844e-05,
+      "loss": 0.6188,
+      "step": 10182
+    },
+    {
+      "epoch": 93.42,
+      "learning_rate": 2.094230769230769e-05,
+      "loss": 0.5737,
+      "step": 10183
+    },
+    {
+      "epoch": 93.43,
+      "learning_rate": 2.0913461538461537e-05,
+      "loss": 0.4287,
+      "step": 10184
+    },
+    {
+      "epoch": 93.44,
+      "learning_rate": 2.0884615384615385e-05,
+      "loss": 0.453,
+      "step": 10185
+    },
+    {
+      "epoch": 93.45,
+      "learning_rate": 2.0855769230769227e-05,
+      "loss": 0.5237,
+      "step": 10186
+    },
+    {
+      "epoch": 93.46,
+      "learning_rate": 2.0826923076923075e-05,
+      "loss": 0.4423,
+      "step": 10187
+    },
+    {
+      "epoch": 93.47,
+      "learning_rate": 2.079807692307692e-05,
+      "loss": 0.4689,
+      "step": 10188
+    },
+    {
+      "epoch": 93.48,
+      "learning_rate": 2.076923076923077e-05,
+      "loss": 0.4634,
+      "step": 10189
+    },
+    {
+      "epoch": 93.49,
+      "learning_rate": 2.074038461538461e-05,
+      "loss": 0.7314,
+      "step": 10190
+    },
+    {
+      "epoch": 93.5,
+      "learning_rate": 2.071153846153846e-05,
+      "loss": 0.509,
+      "step": 10191
+    },
+    {
+      "epoch": 93.5,
+      "learning_rate": 2.0682692307692307e-05,
+      "loss": 0.5296,
+      "step": 10192
+    },
+    {
+      "epoch": 93.51,
+      "learning_rate": 2.065384615384615e-05,
+      "loss": 0.517,
+      "step": 10193
+    },
+    {
+      "epoch": 93.52,
+      "learning_rate": 2.0625e-05,
+      "loss": 0.4147,
+      "step": 10194
+    },
+    {
+      "epoch": 93.53,
+      "learning_rate": 2.059615384615384e-05,
+      "loss": 0.4808,
+      "step": 10195
+    },
+    {
+      "epoch": 93.54,
+      "learning_rate": 2.056730769230769e-05,
+      "loss": 0.5597,
+      "step": 10196
+    },
+    {
+      "epoch": 93.55,
+      "learning_rate": 2.0538461538461538e-05,
+      "loss": 0.5554,
+      "step": 10197
+    },
+    {
+      "epoch": 93.56,
+      "learning_rate": 2.0509615384615383e-05,
+      "loss": 0.5193,
+      "step": 10198
+    },
+    {
+      "epoch": 93.57,
+      "learning_rate": 2.0480769230769228e-05,
+      "loss": 0.5136,
+      "step": 10199
+    },
+    {
+      "epoch": 93.58,
+      "learning_rate": 2.0451923076923073e-05,
+      "loss": 0.5152,
+      "step": 10200
+    },
+    {
+      "epoch": 93.59,
+      "learning_rate": 2.042307692307692e-05,
+      "loss": 0.4189,
+      "step": 10201
+    },
+    {
+      "epoch": 93.6,
+      "learning_rate": 2.039423076923077e-05,
+      "loss": 0.4642,
+      "step": 10202
+    },
+    {
+      "epoch": 93.61,
+      "learning_rate": 2.0365384615384614e-05,
+      "loss": 0.4793,
+      "step": 10203
+    },
+    {
+      "epoch": 93.61,
+      "learning_rate": 2.033653846153846e-05,
+      "loss": 0.5372,
+      "step": 10204
+    },
+    {
+      "epoch": 93.62,
+      "learning_rate": 2.0307692307692304e-05,
+      "loss": 0.4925,
+      "step": 10205
+    },
+    {
+      "epoch": 93.63,
+      "learning_rate": 2.0278846153846153e-05,
+      "loss": 0.4778,
+      "step": 10206
+    },
+    {
+      "epoch": 93.64,
+      "learning_rate": 2.025e-05,
+      "loss": 0.4694,
+      "step": 10207
+    },
+    {
+      "epoch": 93.65,
+      "learning_rate": 2.0221153846153842e-05,
+      "loss": 0.4642,
+      "step": 10208
+    },
+    {
+      "epoch": 93.66,
+      "learning_rate": 2.019230769230769e-05,
+      "loss": 0.5516,
+      "step": 10209
+    },
+    {
+      "epoch": 93.67,
+      "learning_rate": 2.0163461538461536e-05,
+      "loss": 0.6058,
+      "step": 10210
+    },
+    {
+      "epoch": 93.68,
+      "learning_rate": 2.0134615384615384e-05,
+      "loss": 0.5463,
+      "step": 10211
+    },
+    {
+      "epoch": 93.69,
+      "learning_rate": 2.0105769230769232e-05,
+      "loss": 0.584,
+      "step": 10212
+    },
+    {
+      "epoch": 93.7,
+      "learning_rate": 2.0076923076923074e-05,
+      "loss": 0.5369,
+      "step": 10213
+    },
+    {
+      "epoch": 93.71,
+      "learning_rate": 2.0048076923076922e-05,
+      "loss": 0.493,
+      "step": 10214
+    },
+    {
+      "epoch": 93.72,
+      "learning_rate": 2.0019230769230767e-05,
+      "loss": 0.5108,
+      "step": 10215
+    },
+    {
+      "epoch": 93.72,
+      "learning_rate": 1.9990384615384615e-05,
+      "loss": 0.5971,
+      "step": 10216
+    },
+    {
+      "epoch": 93.73,
+      "learning_rate": 1.9961538461538457e-05,
+      "loss": 0.4822,
+      "step": 10217
+    },
+    {
+      "epoch": 93.74,
+      "learning_rate": 1.9932692307692305e-05,
+      "loss": 0.4048,
+      "step": 10218
+    },
+    {
+      "epoch": 93.75,
+      "learning_rate": 1.9903846153846154e-05,
+      "loss": 0.5265,
+      "step": 10219
+    },
+    {
+      "epoch": 93.76,
+      "learning_rate": 1.9875e-05,
+      "loss": 0.4613,
+      "step": 10220
+    },
+    {
+      "epoch": 93.77,
+      "learning_rate": 1.9846153846153847e-05,
+      "loss": 0.5983,
+      "step": 10221
+    },
+    {
+      "epoch": 93.78,
+      "learning_rate": 1.981730769230769e-05,
+      "loss": 0.6184,
+      "step": 10222
+    },
+    {
+      "epoch": 93.79,
+      "learning_rate": 1.9788461538461537e-05,
+      "loss": 0.5115,
+      "step": 10223
+    },
+    {
+      "epoch": 93.8,
+      "learning_rate": 1.9759615384615385e-05,
+      "loss": 0.4294,
+      "step": 10224
+    },
+    {
+      "epoch": 93.81,
+      "learning_rate": 1.973076923076923e-05,
+      "loss": 0.432,
+      "step": 10225
+    },
+    {
+      "epoch": 93.82,
+      "learning_rate": 1.9701923076923075e-05,
+      "loss": 0.4944,
+      "step": 10226
+    },
+    {
+      "epoch": 93.83,
+      "learning_rate": 1.967307692307692e-05,
+      "loss": 0.4117,
+      "step": 10227
+    },
+    {
+      "epoch": 93.83,
+      "learning_rate": 1.9644230769230768e-05,
+      "loss": 0.4788,
+      "step": 10228
+    },
+    {
+      "epoch": 93.84,
+      "learning_rate": 1.9615384615384613e-05,
+      "loss": 0.448,
+      "step": 10229
+    },
+    {
+      "epoch": 93.85,
+      "learning_rate": 1.958653846153846e-05,
+      "loss": 0.4035,
+      "step": 10230
+    },
+    {
+      "epoch": 93.86,
+      "learning_rate": 1.9557692307692306e-05,
+      "loss": 0.3676,
+      "step": 10231
+    },
+    {
+      "epoch": 93.87,
+      "learning_rate": 1.952884615384615e-05,
+      "loss": 0.5534,
+      "step": 10232
+    },
+    {
+      "epoch": 93.88,
+      "learning_rate": 1.95e-05,
+      "loss": 0.4795,
+      "step": 10233
+    },
+    {
+      "epoch": 93.89,
+      "learning_rate": 1.9471153846153845e-05,
+      "loss": 0.5564,
+      "step": 10234
+    },
+    {
+      "epoch": 93.9,
+      "learning_rate": 1.944230769230769e-05,
+      "loss": 0.468,
+      "step": 10235
+    },
+    {
+      "epoch": 93.91,
+      "learning_rate": 1.9413461538461534e-05,
+      "loss": 0.6356,
+      "step": 10236
+    },
+    {
+      "epoch": 93.92,
+      "learning_rate": 1.9384615384615383e-05,
+      "loss": 0.4412,
+      "step": 10237
+    },
+    {
+      "epoch": 93.93,
+      "learning_rate": 1.935576923076923e-05,
+      "loss": 0.505,
+      "step": 10238
+    },
+    {
+      "epoch": 93.94,
+      "learning_rate": 1.9326923076923073e-05,
+      "loss": 0.4514,
+      "step": 10239
+    },
+    {
+      "epoch": 93.94,
+      "learning_rate": 1.929807692307692e-05,
+      "loss": 0.428,
+      "step": 10240
+    },
+    {
+      "epoch": 93.95,
+      "learning_rate": 1.9269230769230766e-05,
+      "loss": 0.443,
+      "step": 10241
+    },
+    {
+      "epoch": 93.96,
+      "learning_rate": 1.9240384615384614e-05,
+      "loss": 0.5377,
+      "step": 10242
+    },
+    {
+      "epoch": 93.97,
+      "learning_rate": 1.9211538461538463e-05,
+      "loss": 0.4659,
+      "step": 10243
+    },
+    {
+      "epoch": 93.98,
+      "learning_rate": 1.9182692307692304e-05,
+      "loss": 0.5364,
+      "step": 10244
+    },
+    {
+      "epoch": 93.99,
+      "learning_rate": 1.9153846153846152e-05,
+      "loss": 0.4349,
+      "step": 10245
+    },
+    {
+      "epoch": 94.0,
+      "learning_rate": 1.9124999999999997e-05,
+      "loss": 0.5046,
+      "step": 10246
+    },
+    {
+      "epoch": 94.01,
+      "learning_rate": 1.9096153846153846e-05,
+      "loss": 0.5356,
+      "step": 10247
+    },
+    {
+      "epoch": 94.02,
+      "learning_rate": 1.9067307692307687e-05,
+      "loss": 0.5224,
+      "step": 10248
+    },
+    {
+      "epoch": 94.03,
+      "learning_rate": 1.9038461538461535e-05,
+      "loss": 0.5711,
+      "step": 10249
+    },
+    {
+      "epoch": 94.04,
+      "learning_rate": 1.9009615384615384e-05,
+      "loss": 0.4909,
+      "step": 10250
+    },
+    {
+      "epoch": 94.05,
+      "learning_rate": 1.898076923076923e-05,
+      "loss": 0.3797,
+      "step": 10251
+    },
+    {
+      "epoch": 94.06,
+      "learning_rate": 1.8951923076923077e-05,
+      "loss": 0.4017,
+      "step": 10252
+    },
+    {
+      "epoch": 94.06,
+      "learning_rate": 1.892307692307692e-05,
+      "loss": 0.4771,
+      "step": 10253
+    },
+    {
+      "epoch": 94.07,
+      "learning_rate": 1.8894230769230767e-05,
+      "loss": 0.4401,
+      "step": 10254
+    },
+    {
+      "epoch": 94.08,
+      "learning_rate": 1.8865384615384615e-05,
+      "loss": 0.5889,
+      "step": 10255
+    },
+    {
+      "epoch": 94.09,
+      "learning_rate": 1.883653846153846e-05,
+      "loss": 0.669,
+      "step": 10256
+    },
+    {
+      "epoch": 94.1,
+      "learning_rate": 1.8807692307692305e-05,
+      "loss": 0.509,
+      "step": 10257
+    },
+    {
+      "epoch": 94.11,
+      "learning_rate": 1.877884615384615e-05,
+      "loss": 0.4274,
+      "step": 10258
+    },
+    {
+      "epoch": 94.12,
+      "learning_rate": 1.875e-05,
+      "loss": 0.5718,
+      "step": 10259
+    },
+    {
+      "epoch": 94.13,
+      "learning_rate": 1.8721153846153847e-05,
+      "loss": 0.3719,
+      "step": 10260
+    },
+    {
+      "epoch": 94.14,
+      "learning_rate": 1.869230769230769e-05,
+      "loss": 0.3848,
+      "step": 10261
+    },
+    {
+      "epoch": 94.15,
+      "learning_rate": 1.8663461538461537e-05,
+      "loss": 0.4957,
+      "step": 10262
+    },
+    {
+      "epoch": 94.16,
+      "learning_rate": 1.863461538461538e-05,
+      "loss": 0.5029,
+      "step": 10263
+    },
+    {
+      "epoch": 94.17,
+      "learning_rate": 1.860576923076923e-05,
+      "loss": 0.372,
+      "step": 10264
+    },
+    {
+      "epoch": 94.17,
+      "learning_rate": 1.8576923076923075e-05,
+      "loss": 0.4622,
+      "step": 10265
+    },
+    {
+      "epoch": 94.18,
+      "learning_rate": 1.8548076923076923e-05,
+      "loss": 0.5866,
+      "step": 10266
+    },
+    {
+      "epoch": 94.19,
+      "learning_rate": 1.8519230769230768e-05,
+      "loss": 0.4896,
+      "step": 10267
+    },
+    {
+      "epoch": 94.2,
+      "learning_rate": 1.8490384615384613e-05,
+      "loss": 0.4859,
+      "step": 10268
+    },
+    {
+      "epoch": 94.21,
+      "learning_rate": 1.846153846153846e-05,
+      "loss": 0.5404,
+      "step": 10269
+    },
+    {
+      "epoch": 94.22,
+      "learning_rate": 1.8432692307692306e-05,
+      "loss": 0.4372,
+      "step": 10270
+    },
+    {
+      "epoch": 94.23,
+      "learning_rate": 1.8403846153846154e-05,
+      "loss": 0.5385,
+      "step": 10271
+    },
+    {
+      "epoch": 94.24,
+      "learning_rate": 1.8375e-05,
+      "loss": 0.5995,
+      "step": 10272
+    },
+    {
+      "epoch": 94.25,
+      "learning_rate": 1.8346153846153844e-05,
+      "loss": 0.5548,
+      "step": 10273
+    },
+    {
+      "epoch": 94.26,
+      "learning_rate": 1.831730769230769e-05,
+      "loss": 0.4992,
+      "step": 10274
+    },
+    {
+      "epoch": 94.27,
+      "learning_rate": 1.8288461538461538e-05,
+      "loss": 0.4706,
+      "step": 10275
+    },
+    {
+      "epoch": 94.28,
+      "learning_rate": 1.8259615384615383e-05,
+      "loss": 0.5537,
+      "step": 10276
+    },
+    {
+      "epoch": 94.28,
+      "learning_rate": 1.823076923076923e-05,
+      "loss": 0.6601,
+      "step": 10277
+    },
+    {
+      "epoch": 94.29,
+      "learning_rate": 1.8201923076923076e-05,
+      "loss": 0.5131,
+      "step": 10278
+    },
+    {
+      "epoch": 94.3,
+      "learning_rate": 1.817307692307692e-05,
+      "loss": 0.4256,
+      "step": 10279
+    },
+    {
+      "epoch": 94.31,
+      "learning_rate": 1.814423076923077e-05,
+      "loss": 0.5928,
+      "step": 10280
+    },
+    {
+      "epoch": 94.32,
+      "learning_rate": 1.8115384615384614e-05,
+      "loss": 0.4823,
+      "step": 10281
+    },
+    {
+      "epoch": 94.33,
+      "learning_rate": 1.8086538461538462e-05,
+      "loss": 0.5435,
+      "step": 10282
+    },
+    {
+      "epoch": 94.34,
+      "learning_rate": 1.8057692307692307e-05,
+      "loss": 0.503,
+      "step": 10283
+    },
+    {
+      "epoch": 94.35,
+      "learning_rate": 1.8028846153846152e-05,
+      "loss": 0.4281,
+      "step": 10284
+    },
+    {
+      "epoch": 94.36,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 0.5213,
+      "step": 10285
+    },
+    {
+      "epoch": 94.37,
+      "learning_rate": 1.7971153846153845e-05,
+      "loss": 0.5577,
+      "step": 10286
+    },
+    {
+      "epoch": 94.38,
+      "learning_rate": 1.794230769230769e-05,
+      "loss": 0.4501,
+      "step": 10287
+    },
+    {
+      "epoch": 94.39,
+      "learning_rate": 1.791346153846154e-05,
+      "loss": 0.5165,
+      "step": 10288
+    },
+    {
+      "epoch": 94.39,
+      "learning_rate": 1.7884615384615384e-05,
+      "loss": 0.6849,
+      "step": 10289
+    },
+    {
+      "epoch": 94.4,
+      "learning_rate": 1.785576923076923e-05,
+      "loss": 0.45,
+      "step": 10290
+    },
+    {
+      "epoch": 94.41,
+      "learning_rate": 1.7826923076923077e-05,
+      "loss": 0.4712,
+      "step": 10291
+    },
+    {
+      "epoch": 94.42,
+      "learning_rate": 1.7798076923076922e-05,
+      "loss": 0.5051,
+      "step": 10292
+    },
+    {
+      "epoch": 94.43,
+      "learning_rate": 1.7769230769230767e-05,
+      "loss": 0.3958,
+      "step": 10293
+    },
+    {
+      "epoch": 94.44,
+      "learning_rate": 1.774038461538461e-05,
+      "loss": 0.4814,
+      "step": 10294
+    },
+    {
+      "epoch": 94.45,
+      "learning_rate": 1.771153846153846e-05,
+      "loss": 0.4809,
+      "step": 10295
+    },
+    {
+      "epoch": 94.46,
+      "learning_rate": 1.7682692307692305e-05,
+      "loss": 0.5339,
+      "step": 10296
+    },
+    {
+      "epoch": 94.47,
+      "learning_rate": 1.7653846153846153e-05,
+      "loss": 0.5241,
+      "step": 10297
+    },
+    {
+      "epoch": 94.48,
+      "learning_rate": 1.7624999999999998e-05,
+      "loss": 0.5824,
+      "step": 10298
+    },
+    {
+      "epoch": 94.49,
+      "learning_rate": 1.7596153846153843e-05,
+      "loss": 0.4249,
+      "step": 10299
+    },
+    {
+      "epoch": 94.5,
+      "learning_rate": 1.756730769230769e-05,
+      "loss": 0.5614,
+      "step": 10300
+    },
+    {
+      "epoch": 94.5,
+      "learning_rate": 1.7538461538461536e-05,
+      "loss": 0.4316,
+      "step": 10301
+    },
+    {
+      "epoch": 94.51,
+      "learning_rate": 1.7509615384615385e-05,
+      "loss": 0.4874,
+      "step": 10302
+    },
+    {
+      "epoch": 94.52,
+      "learning_rate": 1.748076923076923e-05,
+      "loss": 0.5693,
+      "step": 10303
+    },
+    {
+      "epoch": 94.53,
+      "learning_rate": 1.7451923076923074e-05,
+      "loss": 0.5389,
+      "step": 10304
+    },
+    {
+      "epoch": 94.54,
+      "learning_rate": 1.742307692307692e-05,
+      "loss": 0.4566,
+      "step": 10305
+    },
+    {
+      "epoch": 94.55,
+      "learning_rate": 1.7394230769230768e-05,
+      "loss": 0.458,
+      "step": 10306
+    },
+    {
+      "epoch": 94.56,
+      "learning_rate": 1.7365384615384613e-05,
+      "loss": 0.5176,
+      "step": 10307
+    },
+    {
+      "epoch": 94.57,
+      "learning_rate": 1.733653846153846e-05,
+      "loss": 0.3417,
+      "step": 10308
+    },
+    {
+      "epoch": 94.58,
+      "learning_rate": 1.7307692307692306e-05,
+      "loss": 0.5135,
+      "step": 10309
+    },
+    {
+      "epoch": 94.59,
+      "learning_rate": 1.727884615384615e-05,
+      "loss": 0.6086,
+      "step": 10310
+    },
+    {
+      "epoch": 94.6,
+      "learning_rate": 1.725e-05,
+      "loss": 0.4663,
+      "step": 10311
+    },
+    {
+      "epoch": 94.61,
+      "learning_rate": 1.7221153846153844e-05,
+      "loss": 0.5868,
+      "step": 10312
+    },
+    {
+      "epoch": 94.61,
+      "learning_rate": 1.7192307692307692e-05,
+      "loss": 0.6314,
+      "step": 10313
+    },
+    {
+      "epoch": 94.62,
+      "learning_rate": 1.7163461538461537e-05,
+      "loss": 0.4777,
+      "step": 10314
+    },
+    {
+      "epoch": 94.63,
+      "learning_rate": 1.7134615384615382e-05,
+      "loss": 0.4082,
+      "step": 10315
+    },
+    {
+      "epoch": 94.64,
+      "learning_rate": 1.7105769230769227e-05,
+      "loss": 0.5343,
+      "step": 10316
+    },
+    {
+      "epoch": 94.65,
+      "learning_rate": 1.7076923076923076e-05,
+      "loss": 0.5111,
+      "step": 10317
+    },
+    {
+      "epoch": 94.66,
+      "learning_rate": 1.704807692307692e-05,
+      "loss": 0.555,
+      "step": 10318
+    },
+    {
+      "epoch": 94.67,
+      "learning_rate": 1.701923076923077e-05,
+      "loss": 0.42,
+      "step": 10319
+    },
+    {
+      "epoch": 94.68,
+      "learning_rate": 1.6990384615384614e-05,
+      "loss": 0.5835,
+      "step": 10320
+    },
+    {
+      "epoch": 94.69,
+      "learning_rate": 1.696153846153846e-05,
+      "loss": 0.4413,
+      "step": 10321
+    },
+    {
+      "epoch": 94.7,
+      "learning_rate": 1.6932692307692307e-05,
+      "loss": 0.543,
+      "step": 10322
+    },
+    {
+      "epoch": 94.71,
+      "learning_rate": 1.6903846153846152e-05,
+      "loss": 0.3799,
+      "step": 10323
+    },
+    {
+      "epoch": 94.72,
+      "learning_rate": 1.6875e-05,
+      "loss": 0.4917,
+      "step": 10324
+    },
+    {
+      "epoch": 94.72,
+      "learning_rate": 1.6846153846153845e-05,
+      "loss": 0.4492,
+      "step": 10325
+    },
+    {
+      "epoch": 94.73,
+      "learning_rate": 1.681730769230769e-05,
+      "loss": 0.4943,
+      "step": 10326
+    },
+    {
+      "epoch": 94.74,
+      "learning_rate": 1.6788461538461535e-05,
+      "loss": 0.3607,
+      "step": 10327
+    },
+    {
+      "epoch": 94.75,
+      "learning_rate": 1.6759615384615383e-05,
+      "loss": 0.4219,
+      "step": 10328
+    },
+    {
+      "epoch": 94.76,
+      "learning_rate": 1.673076923076923e-05,
+      "loss": 0.5151,
+      "step": 10329
+    },
+    {
+      "epoch": 94.77,
+      "learning_rate": 1.6701923076923077e-05,
+      "loss": 0.4301,
+      "step": 10330
+    },
+    {
+      "epoch": 94.78,
+      "learning_rate": 1.667307692307692e-05,
+      "loss": 0.4917,
+      "step": 10331
+    },
+    {
+      "epoch": 94.79,
+      "learning_rate": 1.6644230769230766e-05,
+      "loss": 0.5467,
+      "step": 10332
+    },
+    {
+      "epoch": 94.8,
+      "learning_rate": 1.6615384615384615e-05,
+      "loss": 0.4145,
+      "step": 10333
+    },
+    {
+      "epoch": 94.81,
+      "learning_rate": 1.658653846153846e-05,
+      "loss": 0.5441,
+      "step": 10334
+    },
+    {
+      "epoch": 94.82,
+      "learning_rate": 1.6557692307692308e-05,
+      "loss": 0.5497,
+      "step": 10335
+    },
+    {
+      "epoch": 94.83,
+      "learning_rate": 1.6528846153846153e-05,
+      "loss": 0.423,
+      "step": 10336
+    },
+    {
+      "epoch": 94.83,
+      "learning_rate": 1.6499999999999998e-05,
+      "loss": 0.5512,
+      "step": 10337
+    },
+    {
+      "epoch": 94.84,
+      "learning_rate": 1.6471153846153846e-05,
+      "loss": 0.5173,
+      "step": 10338
+    },
+    {
+      "epoch": 94.85,
+      "learning_rate": 1.644230769230769e-05,
+      "loss": 0.6498,
+      "step": 10339
+    },
+    {
+      "epoch": 94.86,
+      "learning_rate": 1.641346153846154e-05,
+      "loss": 0.4231,
+      "step": 10340
+    },
+    {
+      "epoch": 94.87,
+      "learning_rate": 1.6384615384615384e-05,
+      "loss": 0.4022,
+      "step": 10341
+    },
+    {
+      "epoch": 94.88,
+      "learning_rate": 1.635576923076923e-05,
+      "loss": 0.5196,
+      "step": 10342
+    },
+    {
+      "epoch": 94.89,
+      "learning_rate": 1.6326923076923074e-05,
+      "loss": 0.4197,
+      "step": 10343
+    },
+    {
+      "epoch": 94.9,
+      "learning_rate": 1.6298076923076923e-05,
+      "loss": 0.4642,
+      "step": 10344
+    },
+    {
+      "epoch": 94.91,
+      "learning_rate": 1.6269230769230767e-05,
+      "loss": 0.4012,
+      "step": 10345
+    },
+    {
+      "epoch": 94.92,
+      "learning_rate": 1.6240384615384616e-05,
+      "loss": 0.4509,
+      "step": 10346
+    },
+    {
+      "epoch": 94.93,
+      "learning_rate": 1.621153846153846e-05,
+      "loss": 0.4727,
+      "step": 10347
+    },
+    {
+      "epoch": 94.94,
+      "learning_rate": 1.6182692307692306e-05,
+      "loss": 0.5677,
+      "step": 10348
+    },
+    {
+      "epoch": 94.94,
+      "learning_rate": 1.6153846153846154e-05,
+      "loss": 0.4604,
+      "step": 10349
+    },
+    {
+      "epoch": 94.95,
+      "learning_rate": 1.6125e-05,
+      "loss": 0.5113,
+      "step": 10350
+    },
+    {
+      "epoch": 94.96,
+      "learning_rate": 1.6096153846153844e-05,
+      "loss": 0.4173,
+      "step": 10351
+    },
+    {
+      "epoch": 94.97,
+      "learning_rate": 1.606730769230769e-05,
+      "loss": 0.4972,
+      "step": 10352
+    },
+    {
+      "epoch": 94.98,
+      "learning_rate": 1.6038461538461537e-05,
+      "loss": 0.5197,
+      "step": 10353
+    },
+    {
+      "epoch": 94.99,
+      "learning_rate": 1.6009615384615382e-05,
+      "loss": 0.6155,
+      "step": 10354
+    },
+    {
+      "epoch": 95.0,
+      "learning_rate": 1.598076923076923e-05,
+      "loss": 0.4603,
+      "step": 10355
+    },
+    {
+      "epoch": 95.01,
+      "learning_rate": 1.5951923076923075e-05,
+      "loss": 0.4347,
+      "step": 10356
+    },
+    {
+      "epoch": 95.02,
+      "learning_rate": 1.592307692307692e-05,
+      "loss": 0.4695,
+      "step": 10357
+    },
+    {
+      "epoch": 95.03,
+      "learning_rate": 1.589423076923077e-05,
+      "loss": 0.6166,
+      "step": 10358
+    },
+    {
+      "epoch": 95.04,
+      "learning_rate": 1.5865384615384613e-05,
+      "loss": 0.49,
+      "step": 10359
+    },
+    {
+      "epoch": 95.05,
+      "learning_rate": 1.5836538461538462e-05,
+      "loss": 0.4806,
+      "step": 10360
+    },
+    {
+      "epoch": 95.06,
+      "learning_rate": 1.5807692307692307e-05,
+      "loss": 0.6627,
+      "step": 10361
+    },
+    {
+      "epoch": 95.06,
+      "learning_rate": 1.577884615384615e-05,
+      "loss": 0.4526,
+      "step": 10362
+    },
+    {
+      "epoch": 95.07,
+      "learning_rate": 1.5749999999999997e-05,
+      "loss": 0.5139,
+      "step": 10363
+    },
+    {
+      "epoch": 95.08,
+      "learning_rate": 1.5721153846153845e-05,
+      "loss": 0.5903,
+      "step": 10364
+    },
+    {
+      "epoch": 95.09,
+      "learning_rate": 1.569230769230769e-05,
+      "loss": 0.3888,
+      "step": 10365
+    },
+    {
+      "epoch": 95.1,
+      "learning_rate": 1.5663461538461538e-05,
+      "loss": 0.5849,
+      "step": 10366
+    },
+    {
+      "epoch": 95.11,
+      "learning_rate": 1.5634615384615383e-05,
+      "loss": 0.512,
+      "step": 10367
+    },
+    {
+      "epoch": 95.12,
+      "learning_rate": 1.5605769230769228e-05,
+      "loss": 0.4922,
+      "step": 10368
+    },
+    {
+      "epoch": 95.13,
+      "learning_rate": 1.5576923076923076e-05,
+      "loss": 0.449,
+      "step": 10369
+    },
+    {
+      "epoch": 95.14,
+      "learning_rate": 1.554807692307692e-05,
+      "loss": 0.4927,
+      "step": 10370
+    },
+    {
+      "epoch": 95.15,
+      "learning_rate": 1.551923076923077e-05,
+      "loss": 0.5614,
+      "step": 10371
+    },
+    {
+      "epoch": 95.16,
+      "learning_rate": 1.5490384615384615e-05,
+      "loss": 0.5259,
+      "step": 10372
+    },
+    {
+      "epoch": 95.17,
+      "learning_rate": 1.546153846153846e-05,
+      "loss": 0.4598,
+      "step": 10373
+    },
+    {
+      "epoch": 95.17,
+      "learning_rate": 1.5432692307692304e-05,
+      "loss": 0.4167,
+      "step": 10374
+    },
+    {
+      "epoch": 95.18,
+      "learning_rate": 1.5403846153846153e-05,
+      "loss": 0.4718,
+      "step": 10375
+    },
+    {
+      "epoch": 95.19,
+      "learning_rate": 1.5374999999999998e-05,
+      "loss": 0.5417,
+      "step": 10376
+    },
+    {
+      "epoch": 95.2,
+      "learning_rate": 1.5346153846153846e-05,
+      "loss": 0.4678,
+      "step": 10377
+    },
+    {
+      "epoch": 95.21,
+      "learning_rate": 1.531730769230769e-05,
+      "loss": 0.4674,
+      "step": 10378
+    },
+    {
+      "epoch": 95.22,
+      "learning_rate": 1.5288461538461536e-05,
+      "loss": 0.374,
+      "step": 10379
+    },
+    {
+      "epoch": 95.23,
+      "learning_rate": 1.5259615384615384e-05,
+      "loss": 0.4622,
+      "step": 10380
+    },
+    {
+      "epoch": 95.24,
+      "learning_rate": 1.5230769230769229e-05,
+      "loss": 0.4749,
+      "step": 10381
+    },
+    {
+      "epoch": 95.25,
+      "learning_rate": 1.5201923076923076e-05,
+      "loss": 0.5358,
+      "step": 10382
+    },
+    {
+      "epoch": 95.26,
+      "learning_rate": 1.517307692307692e-05,
+      "loss": 0.4208,
+      "step": 10383
+    },
+    {
+      "epoch": 95.27,
+      "learning_rate": 1.5144230769230769e-05,
+      "loss": 0.4744,
+      "step": 10384
+    },
+    {
+      "epoch": 95.28,
+      "learning_rate": 1.5115384615384614e-05,
+      "loss": 0.5754,
+      "step": 10385
+    },
+    {
+      "epoch": 95.28,
+      "learning_rate": 1.508653846153846e-05,
+      "loss": 0.4923,
+      "step": 10386
+    },
+    {
+      "epoch": 95.29,
+      "learning_rate": 1.5057692307692305e-05,
+      "loss": 0.4965,
+      "step": 10387
+    },
+    {
+      "epoch": 95.3,
+      "learning_rate": 1.5028846153846152e-05,
+      "loss": 0.4215,
+      "step": 10388
+    },
+    {
+      "epoch": 95.31,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 0.4856,
+      "step": 10389
+    },
+    {
+      "epoch": 95.32,
+      "learning_rate": 1.4971153846153845e-05,
+      "loss": 0.4536,
+      "step": 10390
+    },
+    {
+      "epoch": 95.33,
+      "learning_rate": 1.4942307692307692e-05,
+      "loss": 0.4671,
+      "step": 10391
+    },
+    {
+      "epoch": 95.34,
+      "learning_rate": 1.4913461538461537e-05,
+      "loss": 0.6395,
+      "step": 10392
+    },
+    {
+      "epoch": 95.35,
+      "learning_rate": 1.4884615384615383e-05,
+      "loss": 0.5057,
+      "step": 10393
+    },
+    {
+      "epoch": 95.36,
+      "learning_rate": 1.4855769230769228e-05,
+      "loss": 0.486,
+      "step": 10394
+    },
+    {
+      "epoch": 95.37,
+      "learning_rate": 1.4826923076923075e-05,
+      "loss": 0.4376,
+      "step": 10395
+    },
+    {
+      "epoch": 95.38,
+      "learning_rate": 1.4798076923076922e-05,
+      "loss": 0.5343,
+      "step": 10396
+    },
+    {
+      "epoch": 95.39,
+      "learning_rate": 1.4769230769230768e-05,
+      "loss": 0.49,
+      "step": 10397
+    },
+    {
+      "epoch": 95.39,
+      "learning_rate": 1.4740384615384615e-05,
+      "loss": 0.5985,
+      "step": 10398
+    },
+    {
+      "epoch": 95.4,
+      "learning_rate": 1.471153846153846e-05,
+      "loss": 0.4997,
+      "step": 10399
+    },
+    {
+      "epoch": 95.41,
+      "learning_rate": 1.4682692307692306e-05,
+      "loss": 0.5388,
+      "step": 10400
+    },
+    {
+      "epoch": 95.42,
+      "learning_rate": 1.4653846153846151e-05,
+      "loss": 0.5061,
+      "step": 10401
+    },
+    {
+      "epoch": 95.43,
+      "learning_rate": 1.4625e-05,
+      "loss": 0.5248,
+      "step": 10402
+    },
+    {
+      "epoch": 95.44,
+      "learning_rate": 1.4596153846153845e-05,
+      "loss": 0.5129,
+      "step": 10403
+    },
+    {
+      "epoch": 95.45,
+      "learning_rate": 1.4567307692307691e-05,
+      "loss": 0.5422,
+      "step": 10404
+    },
+    {
+      "epoch": 95.46,
+      "learning_rate": 1.4538461538461536e-05,
+      "loss": 0.395,
+      "step": 10405
+    },
+    {
+      "epoch": 95.47,
+      "learning_rate": 1.4509615384615383e-05,
+      "loss": 0.4277,
+      "step": 10406
+    },
+    {
+      "epoch": 95.48,
+      "learning_rate": 1.4480769230769231e-05,
+      "loss": 0.5281,
+      "step": 10407
+    },
+    {
+      "epoch": 95.49,
+      "learning_rate": 1.4451923076923076e-05,
+      "loss": 0.573,
+      "step": 10408
+    },
+    {
+      "epoch": 95.5,
+      "learning_rate": 1.4423076923076923e-05,
+      "loss": 0.4747,
+      "step": 10409
+    },
+    {
+      "epoch": 95.5,
+      "learning_rate": 1.4394230769230768e-05,
+      "loss": 0.5316,
+      "step": 10410
+    },
+    {
+      "epoch": 95.51,
+      "learning_rate": 1.4365384615384614e-05,
+      "loss": 0.4936,
+      "step": 10411
+    },
+    {
+      "epoch": 95.52,
+      "learning_rate": 1.433653846153846e-05,
+      "loss": 0.5074,
+      "step": 10412
+    },
+    {
+      "epoch": 95.53,
+      "learning_rate": 1.4307692307692308e-05,
+      "loss": 0.5332,
+      "step": 10413
+    },
+    {
+      "epoch": 95.54,
+      "learning_rate": 1.4278846153846152e-05,
+      "loss": 0.5044,
+      "step": 10414
+    },
+    {
+      "epoch": 95.55,
+      "learning_rate": 1.4249999999999999e-05,
+      "loss": 0.574,
+      "step": 10415
+    },
+    {
+      "epoch": 95.56,
+      "learning_rate": 1.4221153846153844e-05,
+      "loss": 0.5701,
+      "step": 10416
+    },
+    {
+      "epoch": 95.57,
+      "learning_rate": 1.419230769230769e-05,
+      "loss": 0.6147,
+      "step": 10417
+    },
+    {
+      "epoch": 95.58,
+      "learning_rate": 1.4163461538461537e-05,
+      "loss": 0.4703,
+      "step": 10418
+    },
+    {
+      "epoch": 95.59,
+      "learning_rate": 1.4134615384615384e-05,
+      "loss": 0.54,
+      "step": 10419
+    },
+    {
+      "epoch": 95.6,
+      "learning_rate": 1.410576923076923e-05,
+      "loss": 0.4339,
+      "step": 10420
+    },
+    {
+      "epoch": 95.61,
+      "learning_rate": 1.4076923076923075e-05,
+      "loss": 0.476,
+      "step": 10421
+    },
+    {
+      "epoch": 95.61,
+      "learning_rate": 1.4048076923076922e-05,
+      "loss": 0.5356,
+      "step": 10422
+    },
+    {
+      "epoch": 95.62,
+      "learning_rate": 1.4019230769230767e-05,
+      "loss": 0.46,
+      "step": 10423
+    },
+    {
+      "epoch": 95.63,
+      "learning_rate": 1.3990384615384614e-05,
+      "loss": 0.3782,
+      "step": 10424
+    },
+    {
+      "epoch": 95.64,
+      "learning_rate": 1.396153846153846e-05,
+      "loss": 0.4552,
+      "step": 10425
+    },
+    {
+      "epoch": 95.65,
+      "learning_rate": 1.3932692307692307e-05,
+      "loss": 0.4303,
+      "step": 10426
+    },
+    {
+      "epoch": 95.66,
+      "learning_rate": 1.3903846153846154e-05,
+      "loss": 0.6525,
+      "step": 10427
+    },
+    {
+      "epoch": 95.67,
+      "learning_rate": 1.3874999999999998e-05,
+      "loss": 0.5905,
+      "step": 10428
+    },
+    {
+      "epoch": 95.68,
+      "learning_rate": 1.3846153846153845e-05,
+      "loss": 0.4533,
+      "step": 10429
+    },
+    {
+      "epoch": 95.69,
+      "learning_rate": 1.381730769230769e-05,
+      "loss": 0.4857,
+      "step": 10430
+    },
+    {
+      "epoch": 95.7,
+      "learning_rate": 1.3788461538461538e-05,
+      "loss": 0.5717,
+      "step": 10431
+    },
+    {
+      "epoch": 95.71,
+      "learning_rate": 1.3759615384615383e-05,
+      "loss": 0.5037,
+      "step": 10432
+    },
+    {
+      "epoch": 95.72,
+      "learning_rate": 1.373076923076923e-05,
+      "loss": 0.4059,
+      "step": 10433
+    },
+    {
+      "epoch": 95.72,
+      "learning_rate": 1.3701923076923075e-05,
+      "loss": 0.4854,
+      "step": 10434
+    },
+    {
+      "epoch": 95.73,
+      "learning_rate": 1.3673076923076921e-05,
+      "loss": 0.4411,
+      "step": 10435
+    },
+    {
+      "epoch": 95.74,
+      "learning_rate": 1.3644230769230766e-05,
+      "loss": 0.4931,
+      "step": 10436
+    },
+    {
+      "epoch": 95.75,
+      "learning_rate": 1.3615384615384615e-05,
+      "loss": 0.4542,
+      "step": 10437
+    },
+    {
+      "epoch": 95.76,
+      "learning_rate": 1.3586538461538461e-05,
+      "loss": 0.493,
+      "step": 10438
+    },
+    {
+      "epoch": 95.77,
+      "learning_rate": 1.3557692307692306e-05,
+      "loss": 0.4616,
+      "step": 10439
+    },
+    {
+      "epoch": 95.78,
+      "learning_rate": 1.3528846153846153e-05,
+      "loss": 0.6212,
+      "step": 10440
+    },
+    {
+      "epoch": 95.79,
+      "learning_rate": 1.3499999999999998e-05,
+      "loss": 0.5507,
+      "step": 10441
+    },
+    {
+      "epoch": 95.8,
+      "learning_rate": 1.3471153846153846e-05,
+      "loss": 0.5712,
+      "step": 10442
+    },
+    {
+      "epoch": 95.81,
+      "learning_rate": 1.3442307692307691e-05,
+      "loss": 0.5141,
+      "step": 10443
+    },
+    {
+      "epoch": 95.82,
+      "learning_rate": 1.3413461538461538e-05,
+      "loss": 0.4636,
+      "step": 10444
+    },
+    {
+      "epoch": 95.83,
+      "learning_rate": 1.3384615384615383e-05,
+      "loss": 0.5502,
+      "step": 10445
+    },
+    {
+      "epoch": 95.83,
+      "learning_rate": 1.335576923076923e-05,
+      "loss": 0.4926,
+      "step": 10446
+    },
+    {
+      "epoch": 95.84,
+      "learning_rate": 1.3326923076923076e-05,
+      "loss": 0.4765,
+      "step": 10447
+    },
+    {
+      "epoch": 95.85,
+      "learning_rate": 1.3298076923076922e-05,
+      "loss": 0.4983,
+      "step": 10448
+    },
+    {
+      "epoch": 95.86,
+      "learning_rate": 1.3269230769230769e-05,
+      "loss": 0.4114,
+      "step": 10449
+    },
+    {
+      "epoch": 95.87,
+      "learning_rate": 1.3240384615384614e-05,
+      "loss": 0.5031,
+      "step": 10450
+    },
+    {
+      "epoch": 95.88,
+      "learning_rate": 1.321153846153846e-05,
+      "loss": 0.4539,
+      "step": 10451
+    },
+    {
+      "epoch": 95.89,
+      "learning_rate": 1.3182692307692306e-05,
+      "loss": 0.456,
+      "step": 10452
+    },
+    {
+      "epoch": 95.9,
+      "learning_rate": 1.3153846153846152e-05,
+      "loss": 0.6627,
+      "step": 10453
+    },
+    {
+      "epoch": 95.91,
+      "learning_rate": 1.3124999999999999e-05,
+      "loss": 0.5482,
+      "step": 10454
+    },
+    {
+      "epoch": 95.92,
+      "learning_rate": 1.3096153846153845e-05,
+      "loss": 0.5676,
+      "step": 10455
+    },
+    {
+      "epoch": 95.93,
+      "learning_rate": 1.3067307692307692e-05,
+      "loss": 0.5625,
+      "step": 10456
+    },
+    {
+      "epoch": 95.94,
+      "learning_rate": 1.3038461538461537e-05,
+      "loss": 0.5099,
+      "step": 10457
+    },
+    {
+      "epoch": 95.94,
+      "learning_rate": 1.3009615384615384e-05,
+      "loss": 0.4759,
+      "step": 10458
+    },
+    {
+      "epoch": 95.95,
+      "learning_rate": 1.2980769230769229e-05,
+      "loss": 0.5365,
+      "step": 10459
+    },
+    {
+      "epoch": 95.96,
+      "learning_rate": 1.2951923076923077e-05,
+      "loss": 0.5009,
+      "step": 10460
+    },
+    {
+      "epoch": 95.97,
+      "learning_rate": 1.2923076923076922e-05,
+      "loss": 0.3361,
+      "step": 10461
+    },
+    {
+      "epoch": 95.98,
+      "learning_rate": 1.2894230769230768e-05,
+      "loss": 0.4326,
+      "step": 10462
+    },
+    {
+      "epoch": 95.99,
+      "learning_rate": 1.2865384615384613e-05,
+      "loss": 0.575,
+      "step": 10463
+    },
+    {
+      "epoch": 96.0,
+      "learning_rate": 1.283653846153846e-05,
+      "loss": 0.436,
+      "step": 10464
+    },
+    {
+      "epoch": 96.01,
+      "learning_rate": 1.2807692307692305e-05,
+      "loss": 0.4866,
+      "step": 10465
+    },
+    {
+      "epoch": 96.02,
+      "learning_rate": 1.2778846153846153e-05,
+      "loss": 0.4713,
+      "step": 10466
+    },
+    {
+      "epoch": 96.03,
+      "learning_rate": 1.275e-05,
+      "loss": 0.5592,
+      "step": 10467
+    },
+    {
+      "epoch": 96.04,
+      "learning_rate": 1.2721153846153845e-05,
+      "loss": 0.4466,
+      "step": 10468
+    },
+    {
+      "epoch": 96.05,
+      "learning_rate": 1.2692307692307691e-05,
+      "loss": 0.5941,
+      "step": 10469
+    },
+    {
+      "epoch": 96.06,
+      "learning_rate": 1.2663461538461536e-05,
+      "loss": 0.431,
+      "step": 10470
+    },
+    {
+      "epoch": 96.06,
+      "learning_rate": 1.2634615384615385e-05,
+      "loss": 0.4496,
+      "step": 10471
+    },
+    {
+      "epoch": 96.07,
+      "learning_rate": 1.260576923076923e-05,
+      "loss": 0.4561,
+      "step": 10472
+    },
+    {
+      "epoch": 96.08,
+      "learning_rate": 1.2576923076923076e-05,
+      "loss": 0.4603,
+      "step": 10473
+    },
+    {
+      "epoch": 96.09,
+      "learning_rate": 1.2548076923076921e-05,
+      "loss": 0.5684,
+      "step": 10474
+    },
+    {
+      "epoch": 96.1,
+      "learning_rate": 1.2519230769230768e-05,
+      "loss": 0.5236,
+      "step": 10475
+    },
+    {
+      "epoch": 96.11,
+      "learning_rate": 1.2490384615384616e-05,
+      "loss": 0.4679,
+      "step": 10476
+    },
+    {
+      "epoch": 96.12,
+      "learning_rate": 1.2461538461538461e-05,
+      "loss": 0.4341,
+      "step": 10477
+    },
+    {
+      "epoch": 96.13,
+      "learning_rate": 1.2432692307692308e-05,
+      "loss": 0.4765,
+      "step": 10478
+    },
+    {
+      "epoch": 96.14,
+      "learning_rate": 1.2403846153846153e-05,
+      "loss": 0.5646,
+      "step": 10479
+    },
+    {
+      "epoch": 96.15,
+      "learning_rate": 1.2375e-05,
+      "loss": 0.4684,
+      "step": 10480
+    },
+    {
+      "epoch": 96.16,
+      "learning_rate": 1.2346153846153844e-05,
+      "loss": 0.5092,
+      "step": 10481
+    },
+    {
+      "epoch": 96.17,
+      "learning_rate": 1.231730769230769e-05,
+      "loss": 0.6077,
+      "step": 10482
+    },
+    {
+      "epoch": 96.17,
+      "learning_rate": 1.2288461538461537e-05,
+      "loss": 0.4722,
+      "step": 10483
+    },
+    {
+      "epoch": 96.18,
+      "learning_rate": 1.2259615384615384e-05,
+      "loss": 0.3738,
+      "step": 10484
+    },
+    {
+      "epoch": 96.19,
+      "learning_rate": 1.2230769230769229e-05,
+      "loss": 0.5311,
+      "step": 10485
+    },
+    {
+      "epoch": 96.2,
+      "learning_rate": 1.2201923076923076e-05,
+      "loss": 0.4995,
+      "step": 10486
+    },
+    {
+      "epoch": 96.21,
+      "learning_rate": 1.2173076923076922e-05,
+      "loss": 0.5714,
+      "step": 10487
+    },
+    {
+      "epoch": 96.22,
+      "learning_rate": 1.2144230769230767e-05,
+      "loss": 0.362,
+      "step": 10488
+    },
+    {
+      "epoch": 96.23,
+      "learning_rate": 1.2115384615384615e-05,
+      "loss": 0.4153,
+      "step": 10489
+    },
+    {
+      "epoch": 96.24,
+      "learning_rate": 1.208653846153846e-05,
+      "loss": 0.4636,
+      "step": 10490
+    },
+    {
+      "epoch": 96.25,
+      "learning_rate": 1.2057692307692307e-05,
+      "loss": 0.3948,
+      "step": 10491
+    },
+    {
+      "epoch": 96.26,
+      "learning_rate": 1.2028846153846152e-05,
+      "loss": 0.47,
+      "step": 10492
+    },
+    {
+      "epoch": 96.27,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 0.5668,
+      "step": 10493
+    },
+    {
+      "epoch": 96.28,
+      "learning_rate": 1.1971153846153844e-05,
+      "loss": 0.4818,
+      "step": 10494
+    },
+    {
+      "epoch": 96.28,
+      "learning_rate": 1.1942307692307692e-05,
+      "loss": 0.4891,
+      "step": 10495
+    },
+    {
+      "epoch": 96.29,
+      "learning_rate": 1.1913461538461538e-05,
+      "loss": 0.5322,
+      "step": 10496
+    },
+    {
+      "epoch": 96.3,
+      "learning_rate": 1.1884615384615383e-05,
+      "loss": 0.4703,
+      "step": 10497
+    },
+    {
+      "epoch": 96.31,
+      "learning_rate": 1.185576923076923e-05,
+      "loss": 0.4104,
+      "step": 10498
+    },
+    {
+      "epoch": 96.32,
+      "learning_rate": 1.1826923076923075e-05,
+      "loss": 0.5089,
+      "step": 10499
+    },
+    {
+      "epoch": 96.33,
+      "learning_rate": 1.1798076923076923e-05,
+      "loss": 0.4879,
+      "step": 10500
+    },
+    {
+      "epoch": 96.33,
+      "eval_cer": 0.08195025412674853,
+      "eval_loss": 0.3683013617992401,
+      "eval_runtime": 13.9843,
+      "eval_samples_per_second": 117.775,
+      "eval_steps_per_second": 1.859,
+      "eval_wer": 0.34184455111837403,
+      "step": 10500
+    },
+    {
+      "epoch": 96.34,
+      "learning_rate": 1.1769230769230768e-05,
+      "loss": 0.4789,
+      "step": 10501
+    },
+    {
+      "epoch": 96.35,
+      "learning_rate": 1.1740384615384615e-05,
+      "loss": 0.4829,
+      "step": 10502
+    },
+    {
+      "epoch": 96.36,
+      "learning_rate": 1.171153846153846e-05,
+      "loss": 0.5068,
+      "step": 10503
+    },
+    {
+      "epoch": 96.37,
+      "learning_rate": 1.1682692307692306e-05,
+      "loss": 0.5913,
+      "step": 10504
+    },
+    {
+      "epoch": 96.38,
+      "learning_rate": 1.1653846153846151e-05,
+      "loss": 0.4347,
+      "step": 10505
+    },
+    {
+      "epoch": 96.39,
+      "learning_rate": 1.1625e-05,
+      "loss": 0.5235,
+      "step": 10506
+    },
+    {
+      "epoch": 96.39,
+      "learning_rate": 1.1596153846153846e-05,
+      "loss": 0.4663,
+      "step": 10507
+    },
+    {
+      "epoch": 96.4,
+      "learning_rate": 1.1567307692307691e-05,
+      "loss": 0.6035,
+      "step": 10508
+    },
+    {
+      "epoch": 96.41,
+      "learning_rate": 1.1538461538461538e-05,
+      "loss": 0.5872,
+      "step": 10509
+    },
+    {
+      "epoch": 96.42,
+      "learning_rate": 1.1509615384615383e-05,
+      "loss": 0.4928,
+      "step": 10510
+    },
+    {
+      "epoch": 96.43,
+      "learning_rate": 1.148076923076923e-05,
+      "loss": 0.4826,
+      "step": 10511
+    },
+    {
+      "epoch": 96.44,
+      "learning_rate": 1.1451923076923076e-05,
+      "loss": 0.3963,
+      "step": 10512
+    },
+    {
+      "epoch": 96.45,
+      "learning_rate": 1.1423076923076923e-05,
+      "loss": 0.4665,
+      "step": 10513
+    },
+    {
+      "epoch": 96.46,
+      "learning_rate": 1.1394230769230768e-05,
+      "loss": 0.3711,
+      "step": 10514
+    },
+    {
+      "epoch": 96.47,
+      "learning_rate": 1.1365384615384614e-05,
+      "loss": 0.6131,
+      "step": 10515
+    },
+    {
+      "epoch": 96.48,
+      "learning_rate": 1.133653846153846e-05,
+      "loss": 0.4205,
+      "step": 10516
+    },
+    {
+      "epoch": 96.49,
+      "learning_rate": 1.1307692307692306e-05,
+      "loss": 0.4388,
+      "step": 10517
+    },
+    {
+      "epoch": 96.5,
+      "learning_rate": 1.1278846153846154e-05,
+      "loss": 0.5219,
+      "step": 10518
+    },
+    {
+      "epoch": 96.5,
+      "learning_rate": 1.1249999999999999e-05,
+      "loss": 0.5237,
+      "step": 10519
+    },
+    {
+      "epoch": 96.51,
+      "learning_rate": 1.1221153846153846e-05,
+      "loss": 0.4702,
+      "step": 10520
+    },
+    {
+      "epoch": 96.52,
+      "learning_rate": 1.119230769230769e-05,
+      "loss": 0.4991,
+      "step": 10521
+    },
+    {
+      "epoch": 96.53,
+      "learning_rate": 1.1163461538461537e-05,
+      "loss": 0.6018,
+      "step": 10522
+    },
+    {
+      "epoch": 96.54,
+      "learning_rate": 1.1134615384615382e-05,
+      "loss": 0.5145,
+      "step": 10523
+    },
+    {
+      "epoch": 96.55,
+      "learning_rate": 1.110576923076923e-05,
+      "loss": 0.4447,
+      "step": 10524
+    },
+    {
+      "epoch": 96.56,
+      "learning_rate": 1.1076923076923077e-05,
+      "loss": 0.5662,
+      "step": 10525
+    },
+    {
+      "epoch": 96.57,
+      "learning_rate": 1.1048076923076922e-05,
+      "loss": 0.4061,
+      "step": 10526
+    },
+    {
+      "epoch": 96.58,
+      "learning_rate": 1.1019230769230769e-05,
+      "loss": 0.4562,
+      "step": 10527
+    },
+    {
+      "epoch": 96.59,
+      "learning_rate": 1.0990384615384614e-05,
+      "loss": 0.5737,
+      "step": 10528
+    },
+    {
+      "epoch": 96.6,
+      "learning_rate": 1.0961538461538462e-05,
+      "loss": 0.4913,
+      "step": 10529
+    },
+    {
+      "epoch": 96.61,
+      "learning_rate": 1.0932692307692307e-05,
+      "loss": 0.4052,
+      "step": 10530
+    },
+    {
+      "epoch": 96.61,
+      "learning_rate": 1.0903846153846153e-05,
+      "loss": 0.5465,
+      "step": 10531
+    },
+    {
+      "epoch": 96.62,
+      "learning_rate": 1.0874999999999998e-05,
+      "loss": 0.5731,
+      "step": 10532
+    },
+    {
+      "epoch": 96.63,
+      "learning_rate": 1.0846153846153845e-05,
+      "loss": 0.407,
+      "step": 10533
+    },
+    {
+      "epoch": 96.64,
+      "learning_rate": 1.081730769230769e-05,
+      "loss": 0.4382,
+      "step": 10534
+    },
+    {
+      "epoch": 96.65,
+      "learning_rate": 1.0788461538461538e-05,
+      "loss": 0.4897,
+      "step": 10535
+    },
+    {
+      "epoch": 96.66,
+      "learning_rate": 1.0759615384615385e-05,
+      "loss": 0.5403,
+      "step": 10536
+    },
+    {
+      "epoch": 96.67,
+      "learning_rate": 1.073076923076923e-05,
+      "loss": 0.5983,
+      "step": 10537
+    },
+    {
+      "epoch": 96.68,
+      "learning_rate": 1.0701923076923076e-05,
+      "loss": 0.53,
+      "step": 10538
+    },
+    {
+      "epoch": 96.69,
+      "learning_rate": 1.0673076923076921e-05,
+      "loss": 0.4793,
+      "step": 10539
+    },
+    {
+      "epoch": 96.7,
+      "learning_rate": 1.0644230769230768e-05,
+      "loss": 0.5175,
+      "step": 10540
+    },
+    {
+      "epoch": 96.71,
+      "learning_rate": 1.0615384615384615e-05,
+      "loss": 0.4952,
+      "step": 10541
+    },
+    {
+      "epoch": 96.72,
+      "learning_rate": 1.0586538461538461e-05,
+      "loss": 0.5578,
+      "step": 10542
+    },
+    {
+      "epoch": 96.72,
+      "learning_rate": 1.0557692307692306e-05,
+      "loss": 0.4878,
+      "step": 10543
+    },
+    {
+      "epoch": 96.73,
+      "learning_rate": 1.0528846153846153e-05,
+      "loss": 0.4781,
+      "step": 10544
+    },
+    {
+      "epoch": 96.74,
+      "learning_rate": 1.05e-05,
+      "loss": 0.5048,
+      "step": 10545
+    },
+    {
+      "epoch": 96.75,
+      "learning_rate": 1.0471153846153844e-05,
+      "loss": 0.4718,
+      "step": 10546
+    },
+    {
+      "epoch": 96.76,
+      "learning_rate": 1.0442307692307693e-05,
+      "loss": 0.5597,
+      "step": 10547
+    },
+    {
+      "epoch": 96.77,
+      "learning_rate": 1.0413461538461538e-05,
+      "loss": 0.5351,
+      "step": 10548
+    },
+    {
+      "epoch": 96.78,
+      "learning_rate": 1.0384615384615384e-05,
+      "loss": 0.5086,
+      "step": 10549
+    },
+    {
+      "epoch": 96.79,
+      "learning_rate": 1.035576923076923e-05,
+      "loss": 0.4387,
+      "step": 10550
+    },
+    {
+      "epoch": 96.8,
+      "learning_rate": 1.0326923076923076e-05,
+      "loss": 0.5047,
+      "step": 10551
+    },
+    {
+      "epoch": 96.81,
+      "learning_rate": 1.029807692307692e-05,
+      "loss": 0.5438,
+      "step": 10552
+    },
+    {
+      "epoch": 96.82,
+      "learning_rate": 1.0269230769230769e-05,
+      "loss": 0.5038,
+      "step": 10553
+    },
+    {
+      "epoch": 96.83,
+      "learning_rate": 1.0240384615384614e-05,
+      "loss": 0.5054,
+      "step": 10554
+    },
+    {
+      "epoch": 96.83,
+      "learning_rate": 1.021153846153846e-05,
+      "loss": 0.5066,
+      "step": 10555
+    },
+    {
+      "epoch": 96.84,
+      "learning_rate": 1.0182692307692307e-05,
+      "loss": 0.5068,
+      "step": 10556
+    },
+    {
+      "epoch": 96.85,
+      "learning_rate": 1.0153846153846152e-05,
+      "loss": 0.4785,
+      "step": 10557
+    },
+    {
+      "epoch": 96.86,
+      "learning_rate": 1.0125e-05,
+      "loss": 0.3724,
+      "step": 10558
+    },
+    {
+      "epoch": 96.87,
+      "learning_rate": 1.0096153846153845e-05,
+      "loss": 0.4861,
+      "step": 10559
+    },
+    {
+      "epoch": 96.88,
+      "learning_rate": 1.0067307692307692e-05,
+      "loss": 0.4594,
+      "step": 10560
+    },
+    {
+      "epoch": 96.89,
+      "learning_rate": 1.0038461538461537e-05,
+      "loss": 0.456,
+      "step": 10561
+    },
+    {
+      "epoch": 96.9,
+      "learning_rate": 1.0009615384615384e-05,
+      "loss": 0.4712,
+      "step": 10562
+    },
+    {
+      "epoch": 96.91,
+      "learning_rate": 9.980769230769229e-06,
+      "loss": 0.4361,
+      "step": 10563
+    },
+    {
+      "epoch": 96.92,
+      "learning_rate": 9.951923076923077e-06,
+      "loss": 0.4702,
+      "step": 10564
+    },
+    {
+      "epoch": 96.93,
+      "learning_rate": 9.923076923076923e-06,
+      "loss": 0.5554,
+      "step": 10565
+    },
+    {
+      "epoch": 96.94,
+      "learning_rate": 9.894230769230768e-06,
+      "loss": 0.4738,
+      "step": 10566
+    },
+    {
+      "epoch": 96.94,
+      "learning_rate": 9.865384615384615e-06,
+      "loss": 0.3873,
+      "step": 10567
+    },
+    {
+      "epoch": 96.95,
+      "learning_rate": 9.83653846153846e-06,
+      "loss": 0.5689,
+      "step": 10568
+    },
+    {
+      "epoch": 96.96,
+      "learning_rate": 9.807692307692307e-06,
+      "loss": 0.5471,
+      "step": 10569
+    },
+    {
+      "epoch": 96.97,
+      "learning_rate": 9.778846153846153e-06,
+      "loss": 0.3801,
+      "step": 10570
+    },
+    {
+      "epoch": 96.98,
+      "learning_rate": 9.75e-06,
+      "loss": 0.4332,
+      "step": 10571
+    },
+    {
+      "epoch": 96.99,
+      "learning_rate": 9.721153846153845e-06,
+      "loss": 0.5042,
+      "step": 10572
+    },
+    {
+      "epoch": 97.0,
+      "learning_rate": 9.692307692307691e-06,
+      "loss": 0.5623,
+      "step": 10573
+    },
+    {
+      "epoch": 97.01,
+      "learning_rate": 9.663461538461536e-06,
+      "loss": 0.4784,
+      "step": 10574
+    },
+    {
+      "epoch": 97.02,
+      "learning_rate": 9.634615384615383e-06,
+      "loss": 0.4068,
+      "step": 10575
+    },
+    {
+      "epoch": 97.03,
+      "learning_rate": 9.605769230769231e-06,
+      "loss": 0.4322,
+      "step": 10576
+    },
+    {
+      "epoch": 97.04,
+      "learning_rate": 9.576923076923076e-06,
+      "loss": 0.5588,
+      "step": 10577
+    },
+    {
+      "epoch": 97.05,
+      "learning_rate": 9.548076923076923e-06,
+      "loss": 0.5654,
+      "step": 10578
+    },
+    {
+      "epoch": 97.06,
+      "learning_rate": 9.519230769230768e-06,
+      "loss": 0.5558,
+      "step": 10579
+    },
+    {
+      "epoch": 97.06,
+      "learning_rate": 9.490384615384614e-06,
+      "loss": 0.4174,
+      "step": 10580
+    },
+    {
+      "epoch": 97.07,
+      "learning_rate": 9.46153846153846e-06,
+      "loss": 0.3907,
+      "step": 10581
+    },
+    {
+      "epoch": 97.08,
+      "learning_rate": 9.432692307692308e-06,
+      "loss": 0.4249,
+      "step": 10582
+    },
+    {
+      "epoch": 97.09,
+      "learning_rate": 9.403846153846153e-06,
+      "loss": 0.4575,
+      "step": 10583
+    },
+    {
+      "epoch": 97.1,
+      "learning_rate": 9.375e-06,
+      "loss": 0.4407,
+      "step": 10584
+    },
+    {
+      "epoch": 97.11,
+      "learning_rate": 9.346153846153846e-06,
+      "loss": 0.4427,
+      "step": 10585
+    },
+    {
+      "epoch": 97.12,
+      "learning_rate": 9.31730769230769e-06,
+      "loss": 0.4652,
+      "step": 10586
+    },
+    {
+      "epoch": 97.13,
+      "learning_rate": 9.288461538461537e-06,
+      "loss": 0.5785,
+      "step": 10587
+    },
+    {
+      "epoch": 97.14,
+      "learning_rate": 9.259615384615384e-06,
+      "loss": 0.4651,
+      "step": 10588
+    },
+    {
+      "epoch": 97.15,
+      "learning_rate": 9.23076923076923e-06,
+      "loss": 0.581,
+      "step": 10589
+    },
+    {
+      "epoch": 97.16,
+      "learning_rate": 9.201923076923077e-06,
+      "loss": 0.4007,
+      "step": 10590
+    },
+    {
+      "epoch": 97.17,
+      "learning_rate": 9.173076923076922e-06,
+      "loss": 0.533,
+      "step": 10591
+    },
+    {
+      "epoch": 97.17,
+      "learning_rate": 9.144230769230769e-06,
+      "loss": 0.4925,
+      "step": 10592
+    },
+    {
+      "epoch": 97.18,
+      "learning_rate": 9.115384615384615e-06,
+      "loss": 0.4844,
+      "step": 10593
+    },
+    {
+      "epoch": 97.19,
+      "learning_rate": 9.08653846153846e-06,
+      "loss": 0.4508,
+      "step": 10594
+    },
+    {
+      "epoch": 97.2,
+      "learning_rate": 9.057692307692307e-06,
+      "loss": 0.4193,
+      "step": 10595
+    },
+    {
+      "epoch": 97.21,
+      "learning_rate": 9.028846153846154e-06,
+      "loss": 0.512,
+      "step": 10596
+    },
+    {
+      "epoch": 97.22,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 0.4429,
+      "step": 10597
+    },
+    {
+      "epoch": 97.23,
+      "learning_rate": 8.971153846153845e-06,
+      "loss": 0.4704,
+      "step": 10598
+    },
+    {
+      "epoch": 97.24,
+      "learning_rate": 8.942307692307692e-06,
+      "loss": 0.5947,
+      "step": 10599
+    },
+    {
+      "epoch": 97.25,
+      "learning_rate": 8.913461538461538e-06,
+      "loss": 0.4267,
+      "step": 10600
+    },
+    {
+      "epoch": 97.26,
+      "learning_rate": 8.884615384615383e-06,
+      "loss": 0.4318,
+      "step": 10601
+    },
+    {
+      "epoch": 97.27,
+      "learning_rate": 8.85576923076923e-06,
+      "loss": 0.4657,
+      "step": 10602
+    },
+    {
+      "epoch": 97.28,
+      "learning_rate": 8.826923076923077e-06,
+      "loss": 0.4178,
+      "step": 10603
+    },
+    {
+      "epoch": 97.28,
+      "learning_rate": 8.798076923076922e-06,
+      "loss": 0.5596,
+      "step": 10604
+    },
+    {
+      "epoch": 97.29,
+      "learning_rate": 8.769230769230768e-06,
+      "loss": 0.5365,
+      "step": 10605
+    },
+    {
+      "epoch": 97.3,
+      "learning_rate": 8.740384615384615e-06,
+      "loss": 0.3962,
+      "step": 10606
+    },
+    {
+      "epoch": 97.31,
+      "learning_rate": 8.71153846153846e-06,
+      "loss": 0.4697,
+      "step": 10607
+    },
+    {
+      "epoch": 97.32,
+      "learning_rate": 8.682692307692306e-06,
+      "loss": 0.5474,
+      "step": 10608
+    },
+    {
+      "epoch": 97.33,
+      "learning_rate": 8.653846153846153e-06,
+      "loss": 0.6144,
+      "step": 10609
+    },
+    {
+      "epoch": 97.34,
+      "learning_rate": 8.625e-06,
+      "loss": 0.6186,
+      "step": 10610
+    },
+    {
+      "epoch": 97.35,
+      "learning_rate": 8.596153846153846e-06,
+      "loss": 0.5461,
+      "step": 10611
+    },
+    {
+      "epoch": 97.36,
+      "learning_rate": 8.567307692307691e-06,
+      "loss": 0.4959,
+      "step": 10612
+    },
+    {
+      "epoch": 97.37,
+      "learning_rate": 8.538461538461538e-06,
+      "loss": 0.5361,
+      "step": 10613
+    },
+    {
+      "epoch": 97.38,
+      "learning_rate": 8.509615384615384e-06,
+      "loss": 0.3946,
+      "step": 10614
+    },
+    {
+      "epoch": 97.39,
+      "learning_rate": 8.48076923076923e-06,
+      "loss": 0.5346,
+      "step": 10615
+    },
+    {
+      "epoch": 97.39,
+      "learning_rate": 8.451923076923076e-06,
+      "loss": 0.513,
+      "step": 10616
+    },
+    {
+      "epoch": 97.4,
+      "learning_rate": 8.423076923076923e-06,
+      "loss": 0.3603,
+      "step": 10617
+    },
+    {
+      "epoch": 97.41,
+      "learning_rate": 8.394230769230768e-06,
+      "loss": 0.546,
+      "step": 10618
+    },
+    {
+      "epoch": 97.42,
+      "learning_rate": 8.365384615384616e-06,
+      "loss": 0.4732,
+      "step": 10619
+    },
+    {
+      "epoch": 97.43,
+      "learning_rate": 8.33653846153846e-06,
+      "loss": 0.5501,
+      "step": 10620
+    },
+    {
+      "epoch": 97.44,
+      "learning_rate": 8.307692307692307e-06,
+      "loss": 0.498,
+      "step": 10621
+    },
+    {
+      "epoch": 97.45,
+      "learning_rate": 8.278846153846154e-06,
+      "loss": 0.4606,
+      "step": 10622
+    },
+    {
+      "epoch": 97.46,
+      "learning_rate": 8.249999999999999e-06,
+      "loss": 0.5488,
+      "step": 10623
+    },
+    {
+      "epoch": 97.47,
+      "learning_rate": 8.221153846153846e-06,
+      "loss": 0.4953,
+      "step": 10624
+    },
+    {
+      "epoch": 97.48,
+      "learning_rate": 8.192307692307692e-06,
+      "loss": 0.4455,
+      "step": 10625
+    },
+    {
+      "epoch": 97.49,
+      "learning_rate": 8.163461538461537e-06,
+      "loss": 0.4074,
+      "step": 10626
+    },
+    {
+      "epoch": 97.5,
+      "learning_rate": 8.134615384615384e-06,
+      "loss": 0.4458,
+      "step": 10627
+    },
+    {
+      "epoch": 97.5,
+      "learning_rate": 8.10576923076923e-06,
+      "loss": 0.4771,
+      "step": 10628
+    },
+    {
+      "epoch": 97.51,
+      "learning_rate": 8.076923076923077e-06,
+      "loss": 0.4396,
+      "step": 10629
+    },
+    {
+      "epoch": 97.52,
+      "learning_rate": 8.048076923076922e-06,
+      "loss": 0.4935,
+      "step": 10630
+    },
+    {
+      "epoch": 97.53,
+      "learning_rate": 8.019230769230769e-06,
+      "loss": 0.5361,
+      "step": 10631
+    },
+    {
+      "epoch": 97.54,
+      "learning_rate": 7.990384615384615e-06,
+      "loss": 0.4568,
+      "step": 10632
+    },
+    {
+      "epoch": 97.55,
+      "learning_rate": 7.96153846153846e-06,
+      "loss": 0.4829,
+      "step": 10633
+    },
+    {
+      "epoch": 97.56,
+      "learning_rate": 7.932692307692307e-06,
+      "loss": 0.6546,
+      "step": 10634
+    },
+    {
+      "epoch": 97.57,
+      "learning_rate": 7.903846153846153e-06,
+      "loss": 0.5182,
+      "step": 10635
+    },
+    {
+      "epoch": 97.58,
+      "learning_rate": 7.874999999999998e-06,
+      "loss": 0.4563,
+      "step": 10636
+    },
+    {
+      "epoch": 97.59,
+      "learning_rate": 7.846153846153845e-06,
+      "loss": 0.3695,
+      "step": 10637
+    },
+    {
+      "epoch": 97.6,
+      "learning_rate": 7.817307692307692e-06,
+      "loss": 0.5018,
+      "step": 10638
+    },
+    {
+      "epoch": 97.61,
+      "learning_rate": 7.788461538461538e-06,
+      "loss": 0.4599,
+      "step": 10639
+    },
+    {
+      "epoch": 97.61,
+      "learning_rate": 7.759615384615385e-06,
+      "loss": 0.4014,
+      "step": 10640
+    },
+    {
+      "epoch": 97.62,
+      "learning_rate": 7.73076923076923e-06,
+      "loss": 0.5452,
+      "step": 10641
+    },
+    {
+      "epoch": 97.63,
+      "learning_rate": 7.701923076923076e-06,
+      "loss": 0.4374,
+      "step": 10642
+    },
+    {
+      "epoch": 97.64,
+      "learning_rate": 7.673076923076923e-06,
+      "loss": 0.4259,
+      "step": 10643
+    },
+    {
+      "epoch": 97.65,
+      "learning_rate": 7.644230769230768e-06,
+      "loss": 0.3588,
+      "step": 10644
+    },
+    {
+      "epoch": 97.66,
+      "learning_rate": 7.6153846153846145e-06,
+      "loss": 0.4996,
+      "step": 10645
+    },
+    {
+      "epoch": 97.67,
+      "learning_rate": 7.58653846153846e-06,
+      "loss": 0.522,
+      "step": 10646
+    },
+    {
+      "epoch": 97.68,
+      "learning_rate": 7.557692307692307e-06,
+      "loss": 0.5794,
+      "step": 10647
+    },
+    {
+      "epoch": 97.69,
+      "learning_rate": 7.528846153846153e-06,
+      "loss": 0.4423,
+      "step": 10648
+    },
+    {
+      "epoch": 97.7,
+      "learning_rate": 7.499999999999999e-06,
+      "loss": 0.489,
+      "step": 10649
+    },
+    {
+      "epoch": 97.71,
+      "learning_rate": 7.471153846153846e-06,
+      "loss": 0.4609,
+      "step": 10650
+    },
+    {
+      "epoch": 97.72,
+      "learning_rate": 7.442307692307692e-06,
+      "loss": 0.484,
+      "step": 10651
+    },
+    {
+      "epoch": 97.72,
+      "learning_rate": 7.4134615384615375e-06,
+      "loss": 0.6258,
+      "step": 10652
+    },
+    {
+      "epoch": 97.73,
+      "learning_rate": 7.384615384615384e-06,
+      "loss": 0.6187,
+      "step": 10653
+    },
+    {
+      "epoch": 97.74,
+      "learning_rate": 7.35576923076923e-06,
+      "loss": 0.5638,
+      "step": 10654
+    },
+    {
+      "epoch": 97.75,
+      "learning_rate": 7.326923076923076e-06,
+      "loss": 0.4456,
+      "step": 10655
+    },
+    {
+      "epoch": 97.76,
+      "learning_rate": 7.298076923076922e-06,
+      "loss": 0.614,
+      "step": 10656
+    },
+    {
+      "epoch": 97.77,
+      "learning_rate": 7.269230769230768e-06,
+      "loss": 0.4231,
+      "step": 10657
+    },
+    {
+      "epoch": 97.78,
+      "learning_rate": 7.240384615384616e-06,
+      "loss": 0.4113,
+      "step": 10658
+    },
+    {
+      "epoch": 97.79,
+      "learning_rate": 7.211538461538461e-06,
+      "loss": 0.4618,
+      "step": 10659
+    },
+    {
+      "epoch": 97.8,
+      "learning_rate": 7.182692307692307e-06,
+      "loss": 0.4883,
+      "step": 10660
+    },
+    {
+      "epoch": 97.81,
+      "learning_rate": 7.153846153846154e-06,
+      "loss": 0.5655,
+      "step": 10661
+    },
+    {
+      "epoch": 97.82,
+      "learning_rate": 7.1249999999999995e-06,
+      "loss": 0.6072,
+      "step": 10662
+    },
+    {
+      "epoch": 97.83,
+      "learning_rate": 7.096153846153845e-06,
+      "loss": 0.5849,
+      "step": 10663
+    },
+    {
+      "epoch": 97.83,
+      "learning_rate": 7.067307692307692e-06,
+      "loss": 0.4342,
+      "step": 10664
+    },
+    {
+      "epoch": 97.84,
+      "learning_rate": 7.038461538461538e-06,
+      "loss": 0.5356,
+      "step": 10665
+    },
+    {
+      "epoch": 97.85,
+      "learning_rate": 7.0096153846153835e-06,
+      "loss": 0.491,
+      "step": 10666
+    },
+    {
+      "epoch": 97.86,
+      "learning_rate": 6.98076923076923e-06,
+      "loss": 0.4788,
+      "step": 10667
+    },
+    {
+      "epoch": 97.87,
+      "learning_rate": 6.951923076923077e-06,
+      "loss": 0.478,
+      "step": 10668
+    },
+    {
+      "epoch": 97.88,
+      "learning_rate": 6.9230769230769225e-06,
+      "loss": 0.4186,
+      "step": 10669
+    },
+    {
+      "epoch": 97.89,
+      "learning_rate": 6.894230769230769e-06,
+      "loss": 0.5699,
+      "step": 10670
+    },
+    {
+      "epoch": 97.9,
+      "learning_rate": 6.865384615384615e-06,
+      "loss": 0.5299,
+      "step": 10671
+    },
+    {
+      "epoch": 97.91,
+      "learning_rate": 6.836538461538461e-06,
+      "loss": 0.5876,
+      "step": 10672
+    },
+    {
+      "epoch": 97.92,
+      "learning_rate": 6.807692307692307e-06,
+      "loss": 0.576,
+      "step": 10673
+    },
+    {
+      "epoch": 97.93,
+      "learning_rate": 6.778846153846153e-06,
+      "loss": 0.5648,
+      "step": 10674
+    },
+    {
+      "epoch": 97.94,
+      "learning_rate": 6.749999999999999e-06,
+      "loss": 0.4585,
+      "step": 10675
+    },
+    {
+      "epoch": 97.94,
+      "learning_rate": 6.7211538461538455e-06,
+      "loss": 0.5213,
+      "step": 10676
+    },
+    {
+      "epoch": 97.95,
+      "learning_rate": 6.692307692307691e-06,
+      "loss": 0.5095,
+      "step": 10677
+    },
+    {
+      "epoch": 97.96,
+      "learning_rate": 6.663461538461538e-06,
+      "loss": 0.5125,
+      "step": 10678
+    },
+    {
+      "epoch": 97.97,
+      "learning_rate": 6.6346153846153846e-06,
+      "loss": 0.4641,
+      "step": 10679
+    },
+    {
+      "epoch": 97.98,
+      "learning_rate": 6.60576923076923e-06,
+      "loss": 0.5539,
+      "step": 10680
+    },
+    {
+      "epoch": 97.99,
+      "learning_rate": 6.576923076923076e-06,
+      "loss": 0.4604,
+      "step": 10681
+    },
+    {
+      "epoch": 98.0,
+      "learning_rate": 6.548076923076923e-06,
+      "loss": 0.4737,
+      "step": 10682
+    },
+    {
+      "epoch": 98.01,
+      "learning_rate": 6.5192307692307685e-06,
+      "loss": 0.5196,
+      "step": 10683
+    },
+    {
+      "epoch": 98.02,
+      "learning_rate": 6.490384615384614e-06,
+      "loss": 0.413,
+      "step": 10684
+    },
+    {
+      "epoch": 98.03,
+      "learning_rate": 6.461538461538461e-06,
+      "loss": 0.5654,
+      "step": 10685
+    },
+    {
+      "epoch": 98.04,
+      "learning_rate": 6.432692307692307e-06,
+      "loss": 0.5122,
+      "step": 10686
+    },
+    {
+      "epoch": 98.05,
+      "learning_rate": 6.4038461538461525e-06,
+      "loss": 0.5844,
+      "step": 10687
+    },
+    {
+      "epoch": 98.06,
+      "learning_rate": 6.375e-06,
+      "loss": 0.5242,
+      "step": 10688
+    },
+    {
+      "epoch": 98.06,
+      "learning_rate": 6.346153846153846e-06,
+      "loss": 0.4652,
+      "step": 10689
+    },
+    {
+      "epoch": 98.07,
+      "learning_rate": 6.317307692307692e-06,
+      "loss": 0.5406,
+      "step": 10690
+    },
+    {
+      "epoch": 98.08,
+      "learning_rate": 6.288461538461538e-06,
+      "loss": 0.4553,
+      "step": 10691
+    },
+    {
+      "epoch": 98.09,
+      "learning_rate": 6.259615384615384e-06,
+      "loss": 0.4813,
+      "step": 10692
+    },
+    {
+      "epoch": 98.1,
+      "learning_rate": 6.2307692307692305e-06,
+      "loss": 0.5769,
+      "step": 10693
+    },
+    {
+      "epoch": 98.11,
+      "learning_rate": 6.201923076923076e-06,
+      "loss": 0.4901,
+      "step": 10694
+    },
+    {
+      "epoch": 98.12,
+      "learning_rate": 6.173076923076922e-06,
+      "loss": 0.4373,
+      "step": 10695
+    },
+    {
+      "epoch": 98.13,
+      "learning_rate": 6.144230769230769e-06,
+      "loss": 0.4648,
+      "step": 10696
+    },
+    {
+      "epoch": 98.14,
+      "learning_rate": 6.1153846153846145e-06,
+      "loss": 0.4494,
+      "step": 10697
+    },
+    {
+      "epoch": 98.15,
+      "learning_rate": 6.086538461538461e-06,
+      "loss": 0.6071,
+      "step": 10698
+    },
+    {
+      "epoch": 98.16,
+      "learning_rate": 6.057692307692308e-06,
+      "loss": 0.5759,
+      "step": 10699
+    },
+    {
+      "epoch": 98.17,
+      "learning_rate": 6.0288461538461535e-06,
+      "loss": 0.5264,
+      "step": 10700
+    },
+    {
+      "epoch": 98.17,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 0.4987,
+      "step": 10701
+    },
+    {
+      "epoch": 98.18,
+      "learning_rate": 5.971153846153846e-06,
+      "loss": 0.5015,
+      "step": 10702
+    },
+    {
+      "epoch": 98.19,
+      "learning_rate": 5.942307692307692e-06,
+      "loss": 0.5663,
+      "step": 10703
+    },
+    {
+      "epoch": 98.2,
+      "learning_rate": 5.9134615384615375e-06,
+      "loss": 0.6075,
+      "step": 10704
+    },
+    {
+      "epoch": 98.21,
+      "learning_rate": 5.884615384615384e-06,
+      "loss": 0.4452,
+      "step": 10705
+    },
+    {
+      "epoch": 98.22,
+      "learning_rate": 5.85576923076923e-06,
+      "loss": 0.4686,
+      "step": 10706
+    },
+    {
+      "epoch": 98.23,
+      "learning_rate": 5.826923076923076e-06,
+      "loss": 0.3875,
+      "step": 10707
+    },
+    {
+      "epoch": 98.24,
+      "learning_rate": 5.798076923076923e-06,
+      "loss": 0.597,
+      "step": 10708
+    },
+    {
+      "epoch": 98.25,
+      "learning_rate": 5.769230769230769e-06,
+      "loss": 0.375,
+      "step": 10709
+    },
+    {
+      "epoch": 98.26,
+      "learning_rate": 5.740384615384615e-06,
+      "loss": 0.6227,
+      "step": 10710
+    },
+    {
+      "epoch": 98.27,
+      "learning_rate": 5.711538461538461e-06,
+      "loss": 0.4436,
+      "step": 10711
+    },
+    {
+      "epoch": 98.28,
+      "learning_rate": 5.682692307692307e-06,
+      "loss": 0.4834,
+      "step": 10712
+    },
+    {
+      "epoch": 98.28,
+      "learning_rate": 5.653846153846153e-06,
+      "loss": 0.4633,
+      "step": 10713
+    },
+    {
+      "epoch": 98.29,
+      "learning_rate": 5.6249999999999995e-06,
+      "loss": 0.5155,
+      "step": 10714
+    },
+    {
+      "epoch": 98.3,
+      "learning_rate": 5.596153846153845e-06,
+      "loss": 0.4272,
+      "step": 10715
+    },
+    {
+      "epoch": 98.31,
+      "learning_rate": 5.567307692307691e-06,
+      "loss": 0.4869,
+      "step": 10716
+    },
+    {
+      "epoch": 98.32,
+      "learning_rate": 5.5384615384615385e-06,
+      "loss": 0.3867,
+      "step": 10717
+    },
+    {
+      "epoch": 98.33,
+      "learning_rate": 5.509615384615384e-06,
+      "loss": 0.4339,
+      "step": 10718
+    },
+    {
+      "epoch": 98.34,
+      "learning_rate": 5.480769230769231e-06,
+      "loss": 0.4084,
+      "step": 10719
+    },
+    {
+      "epoch": 98.35,
+      "learning_rate": 5.451923076923077e-06,
+      "loss": 0.477,
+      "step": 10720
+    },
+    {
+      "epoch": 98.36,
+      "learning_rate": 5.4230769230769225e-06,
+      "loss": 0.5191,
+      "step": 10721
+    },
+    {
+      "epoch": 98.37,
+      "learning_rate": 5.394230769230769e-06,
+      "loss": 0.3856,
+      "step": 10722
+    },
+    {
+      "epoch": 98.38,
+      "learning_rate": 5.365384615384615e-06,
+      "loss": 0.6559,
+      "step": 10723
+    },
+    {
+      "epoch": 98.39,
+      "learning_rate": 5.336538461538461e-06,
+      "loss": 0.5153,
+      "step": 10724
+    },
+    {
+      "epoch": 98.39,
+      "learning_rate": 5.307692307692307e-06,
+      "loss": 0.3942,
+      "step": 10725
+    },
+    {
+      "epoch": 98.4,
+      "learning_rate": 5.278846153846153e-06,
+      "loss": 0.4613,
+      "step": 10726
+    },
+    {
+      "epoch": 98.41,
+      "learning_rate": 5.25e-06,
+      "loss": 0.5467,
+      "step": 10727
+    },
+    {
+      "epoch": 98.42,
+      "learning_rate": 5.221153846153846e-06,
+      "loss": 0.4602,
+      "step": 10728
+    },
+    {
+      "epoch": 98.43,
+      "learning_rate": 5.192307692307692e-06,
+      "loss": 0.5793,
+      "step": 10729
+    },
+    {
+      "epoch": 98.44,
+      "learning_rate": 5.163461538461538e-06,
+      "loss": 0.5253,
+      "step": 10730
+    },
+    {
+      "epoch": 98.45,
+      "learning_rate": 5.1346153846153845e-06,
+      "loss": 0.4755,
+      "step": 10731
+    },
+    {
+      "epoch": 98.46,
+      "learning_rate": 5.10576923076923e-06,
+      "loss": 0.5579,
+      "step": 10732
+    },
+    {
+      "epoch": 98.47,
+      "learning_rate": 5.076923076923076e-06,
+      "loss": 0.4928,
+      "step": 10733
+    },
+    {
+      "epoch": 98.48,
+      "learning_rate": 5.048076923076923e-06,
+      "loss": 0.4629,
+      "step": 10734
+    },
+    {
+      "epoch": 98.49,
+      "learning_rate": 5.0192307692307685e-06,
+      "loss": 0.5069,
+      "step": 10735
+    },
+    {
+      "epoch": 98.5,
+      "learning_rate": 4.990384615384614e-06,
+      "loss": 0.4304,
+      "step": 10736
+    },
+    {
+      "epoch": 98.5,
+      "learning_rate": 4.961538461538462e-06,
+      "loss": 0.4334,
+      "step": 10737
+    },
+    {
+      "epoch": 98.51,
+      "learning_rate": 4.9326923076923075e-06,
+      "loss": 0.4551,
+      "step": 10738
+    },
+    {
+      "epoch": 98.52,
+      "learning_rate": 4.903846153846153e-06,
+      "loss": 0.4769,
+      "step": 10739
+    },
+    {
+      "epoch": 98.53,
+      "learning_rate": 4.875e-06,
+      "loss": 0.5425,
+      "step": 10740
+    },
+    {
+      "epoch": 98.54,
+      "learning_rate": 4.846153846153846e-06,
+      "loss": 0.6584,
+      "step": 10741
+    },
+    {
+      "epoch": 98.55,
+      "learning_rate": 4.8173076923076915e-06,
+      "loss": 0.5349,
+      "step": 10742
+    },
+    {
+      "epoch": 98.56,
+      "learning_rate": 4.788461538461538e-06,
+      "loss": 0.5275,
+      "step": 10743
+    },
+    {
+      "epoch": 98.57,
+      "learning_rate": 4.759615384615384e-06,
+      "loss": 0.5709,
+      "step": 10744
+    },
+    {
+      "epoch": 98.58,
+      "learning_rate": 4.73076923076923e-06,
+      "loss": 0.4646,
+      "step": 10745
+    },
+    {
+      "epoch": 98.59,
+      "learning_rate": 4.701923076923076e-06,
+      "loss": 0.5694,
+      "step": 10746
+    },
+    {
+      "epoch": 98.6,
+      "learning_rate": 4.673076923076923e-06,
+      "loss": 0.4014,
+      "step": 10747
+    },
+    {
+      "epoch": 98.61,
+      "learning_rate": 4.644230769230769e-06,
+      "loss": 0.4563,
+      "step": 10748
+    },
+    {
+      "epoch": 98.61,
+      "learning_rate": 4.615384615384615e-06,
+      "loss": 0.4344,
+      "step": 10749
+    },
+    {
+      "epoch": 98.62,
+      "learning_rate": 4.586538461538461e-06,
+      "loss": 0.5559,
+      "step": 10750
+    },
+    {
+      "epoch": 98.63,
+      "learning_rate": 4.557692307692308e-06,
+      "loss": 0.3557,
+      "step": 10751
+    },
+    {
+      "epoch": 98.64,
+      "learning_rate": 4.5288461538461535e-06,
+      "loss": 0.4277,
+      "step": 10752
+    },
+    {
+      "epoch": 98.65,
+      "learning_rate": 4.499999999999999e-06,
+      "loss": 0.4375,
+      "step": 10753
+    },
+    {
+      "epoch": 98.66,
+      "learning_rate": 4.471153846153846e-06,
+      "loss": 0.4465,
+      "step": 10754
+    },
+    {
+      "epoch": 98.67,
+      "learning_rate": 4.442307692307692e-06,
+      "loss": 0.4996,
+      "step": 10755
+    },
+    {
+      "epoch": 98.68,
+      "learning_rate": 4.413461538461538e-06,
+      "loss": 0.5292,
+      "step": 10756
+    },
+    {
+      "epoch": 98.69,
+      "learning_rate": 4.384615384615384e-06,
+      "loss": 0.4705,
+      "step": 10757
+    },
+    {
+      "epoch": 98.7,
+      "learning_rate": 4.35576923076923e-06,
+      "loss": 0.4991,
+      "step": 10758
+    },
+    {
+      "epoch": 98.71,
+      "learning_rate": 4.3269230769230765e-06,
+      "loss": 0.4598,
+      "step": 10759
+    },
+    {
+      "epoch": 98.72,
+      "learning_rate": 4.298076923076923e-06,
+      "loss": 0.4689,
+      "step": 10760
+    },
+    {
+      "epoch": 98.72,
+      "learning_rate": 4.269230769230769e-06,
+      "loss": 0.4474,
+      "step": 10761
+    },
+    {
+      "epoch": 98.73,
+      "learning_rate": 4.240384615384615e-06,
+      "loss": 0.538,
+      "step": 10762
+    },
+    {
+      "epoch": 98.74,
+      "learning_rate": 4.211538461538461e-06,
+      "loss": 0.5634,
+      "step": 10763
+    },
+    {
+      "epoch": 98.75,
+      "learning_rate": 4.182692307692308e-06,
+      "loss": 0.5783,
+      "step": 10764
+    },
+    {
+      "epoch": 98.76,
+      "learning_rate": 4.153846153846154e-06,
+      "loss": 0.4525,
+      "step": 10765
+    },
+    {
+      "epoch": 98.77,
+      "learning_rate": 4.1249999999999995e-06,
+      "loss": 0.4447,
+      "step": 10766
+    },
+    {
+      "epoch": 98.78,
+      "learning_rate": 4.096153846153846e-06,
+      "loss": 0.4241,
+      "step": 10767
+    },
+    {
+      "epoch": 98.79,
+      "learning_rate": 4.067307692307692e-06,
+      "loss": 0.6721,
+      "step": 10768
+    },
+    {
+      "epoch": 98.8,
+      "learning_rate": 4.0384615384615385e-06,
+      "loss": 0.4231,
+      "step": 10769
+    },
+    {
+      "epoch": 98.81,
+      "learning_rate": 4.009615384615384e-06,
+      "loss": 0.5298,
+      "step": 10770
+    },
+    {
+      "epoch": 98.82,
+      "learning_rate": 3.98076923076923e-06,
+      "loss": 0.4825,
+      "step": 10771
+    },
+    {
+      "epoch": 98.83,
+      "learning_rate": 3.951923076923077e-06,
+      "loss": 0.4898,
+      "step": 10772
+    },
+    {
+      "epoch": 98.83,
+      "learning_rate": 3.9230769230769225e-06,
+      "loss": 0.4647,
+      "step": 10773
+    },
+    {
+      "epoch": 98.84,
+      "learning_rate": 3.894230769230769e-06,
+      "loss": 0.6299,
+      "step": 10774
+    },
+    {
+      "epoch": 98.85,
+      "learning_rate": 3.865384615384615e-06,
+      "loss": 0.446,
+      "step": 10775
+    },
+    {
+      "epoch": 98.86,
+      "learning_rate": 3.8365384615384615e-06,
+      "loss": 0.586,
+      "step": 10776
+    },
+    {
+      "epoch": 98.87,
+      "learning_rate": 3.8076923076923073e-06,
+      "loss": 0.4488,
+      "step": 10777
+    },
+    {
+      "epoch": 98.88,
+      "learning_rate": 3.7788461538461535e-06,
+      "loss": 0.4277,
+      "step": 10778
+    },
+    {
+      "epoch": 98.89,
+      "learning_rate": 3.7499999999999997e-06,
+      "loss": 0.4653,
+      "step": 10779
+    },
+    {
+      "epoch": 98.9,
+      "learning_rate": 3.721153846153846e-06,
+      "loss": 0.4293,
+      "step": 10780
+    },
+    {
+      "epoch": 98.91,
+      "learning_rate": 3.692307692307692e-06,
+      "loss": 0.4812,
+      "step": 10781
+    },
+    {
+      "epoch": 98.92,
+      "learning_rate": 3.663461538461538e-06,
+      "loss": 0.5053,
+      "step": 10782
+    },
+    {
+      "epoch": 98.93,
+      "learning_rate": 3.634615384615384e-06,
+      "loss": 0.447,
+      "step": 10783
+    },
+    {
+      "epoch": 98.94,
+      "learning_rate": 3.6057692307692307e-06,
+      "loss": 0.5245,
+      "step": 10784
+    },
+    {
+      "epoch": 98.94,
+      "learning_rate": 3.576923076923077e-06,
+      "loss": 0.4962,
+      "step": 10785
+    },
+    {
+      "epoch": 98.95,
+      "learning_rate": 3.5480769230769227e-06,
+      "loss": 0.554,
+      "step": 10786
+    },
+    {
+      "epoch": 98.96,
+      "learning_rate": 3.519230769230769e-06,
+      "loss": 0.428,
+      "step": 10787
+    },
+    {
+      "epoch": 98.97,
+      "learning_rate": 3.490384615384615e-06,
+      "loss": 0.538,
+      "step": 10788
+    },
+    {
+      "epoch": 98.98,
+      "learning_rate": 3.4615384615384613e-06,
+      "loss": 0.4881,
+      "step": 10789
+    },
+    {
+      "epoch": 98.99,
+      "learning_rate": 3.4326923076923075e-06,
+      "loss": 0.4652,
+      "step": 10790
+    },
+    {
+      "epoch": 99.0,
+      "learning_rate": 3.4038461538461537e-06,
+      "loss": 0.5251,
+      "step": 10791
+    },
+    {
+      "epoch": 99.01,
+      "learning_rate": 3.3749999999999995e-06,
+      "loss": 0.471,
+      "step": 10792
+    },
+    {
+      "epoch": 99.02,
+      "learning_rate": 3.3461538461538457e-06,
+      "loss": 0.5135,
+      "step": 10793
+    },
+    {
+      "epoch": 99.03,
+      "learning_rate": 3.3173076923076923e-06,
+      "loss": 0.4757,
+      "step": 10794
+    },
+    {
+      "epoch": 99.04,
+      "learning_rate": 3.288461538461538e-06,
+      "loss": 0.5308,
+      "step": 10795
+    },
+    {
+      "epoch": 99.05,
+      "learning_rate": 3.2596153846153843e-06,
+      "loss": 0.4576,
+      "step": 10796
+    },
+    {
+      "epoch": 99.06,
+      "learning_rate": 3.2307692307692305e-06,
+      "loss": 0.4762,
+      "step": 10797
+    },
+    {
+      "epoch": 99.06,
+      "learning_rate": 3.2019230769230762e-06,
+      "loss": 0.5775,
+      "step": 10798
+    },
+    {
+      "epoch": 99.07,
+      "learning_rate": 3.173076923076923e-06,
+      "loss": 0.3935,
+      "step": 10799
+    },
+    {
+      "epoch": 99.08,
+      "learning_rate": 3.144230769230769e-06,
+      "loss": 0.5904,
+      "step": 10800
+    },
+    {
+      "epoch": 99.09,
+      "learning_rate": 3.1153846153846153e-06,
+      "loss": 0.5368,
+      "step": 10801
+    },
+    {
+      "epoch": 99.1,
+      "learning_rate": 3.086538461538461e-06,
+      "loss": 0.6232,
+      "step": 10802
+    },
+    {
+      "epoch": 99.11,
+      "learning_rate": 3.0576923076923072e-06,
+      "loss": 0.561,
+      "step": 10803
+    },
+    {
+      "epoch": 99.12,
+      "learning_rate": 3.028846153846154e-06,
+      "loss": 0.4694,
+      "step": 10804
+    },
+    {
+      "epoch": 99.13,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 0.5618,
+      "step": 10805
+    },
+    {
+      "epoch": 99.14,
+      "learning_rate": 2.971153846153846e-06,
+      "loss": 0.5166,
+      "step": 10806
+    },
+    {
+      "epoch": 99.15,
+      "learning_rate": 2.942307692307692e-06,
+      "loss": 0.42,
+      "step": 10807
+    },
+    {
+      "epoch": 99.16,
+      "learning_rate": 2.913461538461538e-06,
+      "loss": 0.5312,
+      "step": 10808
+    },
+    {
+      "epoch": 99.17,
+      "learning_rate": 2.8846153846153845e-06,
+      "loss": 0.3964,
+      "step": 10809
+    },
+    {
+      "epoch": 99.17,
+      "learning_rate": 2.8557692307692307e-06,
+      "loss": 0.3949,
+      "step": 10810
+    },
+    {
+      "epoch": 99.18,
+      "learning_rate": 2.8269230769230764e-06,
+      "loss": 0.5269,
+      "step": 10811
+    },
+    {
+      "epoch": 99.19,
+      "learning_rate": 2.7980769230769226e-06,
+      "loss": 0.4087,
+      "step": 10812
+    },
+    {
+      "epoch": 99.2,
+      "learning_rate": 2.7692307692307693e-06,
+      "loss": 0.4831,
+      "step": 10813
+    },
+    {
+      "epoch": 99.21,
+      "learning_rate": 2.7403846153846155e-06,
+      "loss": 0.5554,
+      "step": 10814
+    },
+    {
+      "epoch": 99.22,
+      "learning_rate": 2.7115384615384612e-06,
+      "loss": 0.5563,
+      "step": 10815
+    },
+    {
+      "epoch": 99.23,
+      "learning_rate": 2.6826923076923075e-06,
+      "loss": 0.4486,
+      "step": 10816
+    },
+    {
+      "epoch": 99.24,
+      "learning_rate": 2.6538461538461537e-06,
+      "loss": 0.5088,
+      "step": 10817
+    },
+    {
+      "epoch": 99.25,
+      "learning_rate": 2.625e-06,
+      "loss": 0.4979,
+      "step": 10818
+    },
+    {
+      "epoch": 99.26,
+      "learning_rate": 2.596153846153846e-06,
+      "loss": 0.4409,
+      "step": 10819
+    },
+    {
+      "epoch": 99.27,
+      "learning_rate": 2.5673076923076923e-06,
+      "loss": 0.6143,
+      "step": 10820
+    },
+    {
+      "epoch": 99.28,
+      "learning_rate": 2.538461538461538e-06,
+      "loss": 0.4298,
+      "step": 10821
+    },
+    {
+      "epoch": 99.28,
+      "learning_rate": 2.5096153846153842e-06,
+      "loss": 0.5193,
+      "step": 10822
+    },
+    {
+      "epoch": 99.29,
+      "learning_rate": 2.480769230769231e-06,
+      "loss": 0.4715,
+      "step": 10823
+    },
+    {
+      "epoch": 99.3,
+      "learning_rate": 2.4519230769230766e-06,
+      "loss": 0.5072,
+      "step": 10824
+    },
+    {
+      "epoch": 99.31,
+      "learning_rate": 2.423076923076923e-06,
+      "loss": 0.404,
+      "step": 10825
+    },
+    {
+      "epoch": 99.32,
+      "learning_rate": 2.394230769230769e-06,
+      "loss": 0.5714,
+      "step": 10826
+    },
+    {
+      "epoch": 99.33,
+      "learning_rate": 2.365384615384615e-06,
+      "loss": 0.347,
+      "step": 10827
+    },
+    {
+      "epoch": 99.34,
+      "learning_rate": 2.3365384615384615e-06,
+      "loss": 0.3448,
+      "step": 10828
+    },
+    {
+      "epoch": 99.35,
+      "learning_rate": 2.3076923076923077e-06,
+      "loss": 0.4541,
+      "step": 10829
+    },
+    {
+      "epoch": 99.36,
+      "learning_rate": 2.278846153846154e-06,
+      "loss": 0.5319,
+      "step": 10830
+    },
+    {
+      "epoch": 99.37,
+      "learning_rate": 2.2499999999999996e-06,
+      "loss": 0.5179,
+      "step": 10831
+    },
+    {
+      "epoch": 99.38,
+      "learning_rate": 2.221153846153846e-06,
+      "loss": 0.6121,
+      "step": 10832
+    },
+    {
+      "epoch": 99.39,
+      "learning_rate": 2.192307692307692e-06,
+      "loss": 0.6196,
+      "step": 10833
+    },
+    {
+      "epoch": 99.39,
+      "learning_rate": 2.1634615384615382e-06,
+      "loss": 0.5711,
+      "step": 10834
+    },
+    {
+      "epoch": 99.4,
+      "learning_rate": 2.1346153846153844e-06,
+      "loss": 0.4116,
+      "step": 10835
+    },
+    {
+      "epoch": 99.41,
+      "learning_rate": 2.1057692307692306e-06,
+      "loss": 0.4176,
+      "step": 10836
+    },
+    {
+      "epoch": 99.42,
+      "learning_rate": 2.076923076923077e-06,
+      "loss": 0.4309,
+      "step": 10837
+    },
+    {
+      "epoch": 99.43,
+      "learning_rate": 2.048076923076923e-06,
+      "loss": 0.5822,
+      "step": 10838
+    },
+    {
+      "epoch": 99.44,
+      "learning_rate": 2.0192307692307692e-06,
+      "loss": 0.6158,
+      "step": 10839
+    },
+    {
+      "epoch": 99.45,
+      "learning_rate": 1.990384615384615e-06,
+      "loss": 0.5643,
+      "step": 10840
+    },
+    {
+      "epoch": 99.46,
+      "learning_rate": 1.9615384615384612e-06,
+      "loss": 0.5073,
+      "step": 10841
+    },
+    {
+      "epoch": 99.47,
+      "learning_rate": 1.9326923076923074e-06,
+      "loss": 0.5464,
+      "step": 10842
+    },
+    {
+      "epoch": 99.48,
+      "learning_rate": 1.9038461538461536e-06,
+      "loss": 0.5519,
+      "step": 10843
+    },
+    {
+      "epoch": 99.49,
+      "learning_rate": 1.8749999999999998e-06,
+      "loss": 0.4587,
+      "step": 10844
+    },
+    {
+      "epoch": 99.5,
+      "learning_rate": 1.846153846153846e-06,
+      "loss": 0.5134,
+      "step": 10845
+    },
+    {
+      "epoch": 99.5,
+      "learning_rate": 1.817307692307692e-06,
+      "loss": 0.4545,
+      "step": 10846
+    },
+    {
+      "epoch": 99.51,
+      "learning_rate": 1.7884615384615384e-06,
+      "loss": 0.5497,
+      "step": 10847
+    },
+    {
+      "epoch": 99.52,
+      "learning_rate": 1.7596153846153844e-06,
+      "loss": 0.5889,
+      "step": 10848
+    },
+    {
+      "epoch": 99.53,
+      "learning_rate": 1.7307692307692306e-06,
+      "loss": 0.4812,
+      "step": 10849
+    },
+    {
+      "epoch": 99.54,
+      "learning_rate": 1.7019230769230768e-06,
+      "loss": 0.4847,
+      "step": 10850
+    },
+    {
+      "epoch": 99.55,
+      "learning_rate": 1.6730769230769228e-06,
+      "loss": 0.5063,
+      "step": 10851
+    },
+    {
+      "epoch": 99.56,
+      "learning_rate": 1.644230769230769e-06,
+      "loss": 0.5214,
+      "step": 10852
+    },
+    {
+      "epoch": 99.57,
+      "learning_rate": 1.6153846153846152e-06,
+      "loss": 0.4699,
+      "step": 10853
+    },
+    {
+      "epoch": 99.58,
+      "learning_rate": 1.5865384615384614e-06,
+      "loss": 0.4391,
+      "step": 10854
+    },
+    {
+      "epoch": 99.59,
+      "learning_rate": 1.5576923076923076e-06,
+      "loss": 0.5438,
+      "step": 10855
+    },
+    {
+      "epoch": 99.6,
+      "learning_rate": 1.5288461538461536e-06,
+      "loss": 0.5279,
+      "step": 10856
+    },
+    {
+      "epoch": 99.61,
+      "learning_rate": 1.4999999999999998e-06,
+      "loss": 0.3757,
+      "step": 10857
+    },
+    {
+      "epoch": 99.61,
+      "learning_rate": 1.471153846153846e-06,
+      "loss": 0.5417,
+      "step": 10858
+    },
+    {
+      "epoch": 99.62,
+      "learning_rate": 1.4423076923076922e-06,
+      "loss": 0.5223,
+      "step": 10859
+    },
+    {
+      "epoch": 99.63,
+      "learning_rate": 1.4134615384615382e-06,
+      "loss": 0.5038,
+      "step": 10860
+    },
+    {
+      "epoch": 99.64,
+      "learning_rate": 1.3846153846153846e-06,
+      "loss": 0.4346,
+      "step": 10861
+    },
+    {
+      "epoch": 99.65,
+      "learning_rate": 1.3557692307692306e-06,
+      "loss": 0.4731,
+      "step": 10862
+    },
+    {
+      "epoch": 99.66,
+      "learning_rate": 1.3269230769230768e-06,
+      "loss": 0.5827,
+      "step": 10863
+    },
+    {
+      "epoch": 99.67,
+      "learning_rate": 1.298076923076923e-06,
+      "loss": 0.4207,
+      "step": 10864
+    },
+    {
+      "epoch": 99.68,
+      "learning_rate": 1.269230769230769e-06,
+      "loss": 0.5278,
+      "step": 10865
+    },
+    {
+      "epoch": 99.69,
+      "learning_rate": 1.2403846153846154e-06,
+      "loss": 0.4947,
+      "step": 10866
+    },
+    {
+      "epoch": 99.7,
+      "learning_rate": 1.2115384615384614e-06,
+      "loss": 0.5914,
+      "step": 10867
+    },
+    {
+      "epoch": 99.71,
+      "learning_rate": 1.1826923076923074e-06,
+      "loss": 0.5331,
+      "step": 10868
+    },
+    {
+      "epoch": 99.72,
+      "learning_rate": 1.1538461538461538e-06,
+      "loss": 0.4542,
+      "step": 10869
+    },
+    {
+      "epoch": 99.72,
+      "learning_rate": 1.1249999999999998e-06,
+      "loss": 0.6115,
+      "step": 10870
+    },
+    {
+      "epoch": 99.73,
+      "learning_rate": 1.096153846153846e-06,
+      "loss": 0.5196,
+      "step": 10871
+    },
+    {
+      "epoch": 99.74,
+      "learning_rate": 1.0673076923076922e-06,
+      "loss": 0.5751,
+      "step": 10872
+    },
+    {
+      "epoch": 99.75,
+      "learning_rate": 1.0384615384615384e-06,
+      "loss": 0.5412,
+      "step": 10873
+    },
+    {
+      "epoch": 99.76,
+      "learning_rate": 1.0096153846153846e-06,
+      "loss": 0.4416,
+      "step": 10874
+    },
+    {
+      "epoch": 99.77,
+      "learning_rate": 9.807692307692306e-07,
+      "loss": 0.5436,
+      "step": 10875
+    },
+    {
+      "epoch": 99.78,
+      "learning_rate": 9.519230769230768e-07,
+      "loss": 0.5724,
+      "step": 10876
+    },
+    {
+      "epoch": 99.79,
+      "learning_rate": 9.23076923076923e-07,
+      "loss": 0.466,
+      "step": 10877
+    },
+    {
+      "epoch": 99.8,
+      "learning_rate": 8.942307692307692e-07,
+      "loss": 0.4693,
+      "step": 10878
+    },
+    {
+      "epoch": 99.81,
+      "learning_rate": 8.653846153846153e-07,
+      "loss": 0.5421,
+      "step": 10879
+    },
+    {
+      "epoch": 99.82,
+      "learning_rate": 8.365384615384614e-07,
+      "loss": 0.49,
+      "step": 10880
+    },
+    {
+      "epoch": 99.83,
+      "learning_rate": 8.076923076923076e-07,
+      "loss": 0.5,
+      "step": 10881
+    },
+    {
+      "epoch": 99.83,
+      "learning_rate": 7.788461538461538e-07,
+      "loss": 0.385,
+      "step": 10882
+    },
+    {
+      "epoch": 99.84,
+      "learning_rate": 7.499999999999999e-07,
+      "loss": 0.4884,
+      "step": 10883
+    },
+    {
+      "epoch": 99.85,
+      "learning_rate": 7.211538461538461e-07,
+      "loss": 0.5333,
+      "step": 10884
+    },
+    {
+      "epoch": 99.86,
+      "learning_rate": 6.923076923076923e-07,
+      "loss": 0.5281,
+      "step": 10885
+    },
+    {
+      "epoch": 99.87,
+      "learning_rate": 6.634615384615384e-07,
+      "loss": 0.4867,
+      "step": 10886
+    },
+    {
+      "epoch": 99.88,
+      "learning_rate": 6.346153846153845e-07,
+      "loss": 0.5369,
+      "step": 10887
+    },
+    {
+      "epoch": 99.89,
+      "learning_rate": 6.057692307692307e-07,
+      "loss": 0.4444,
+      "step": 10888
+    },
+    {
+      "epoch": 99.9,
+      "learning_rate": 5.769230769230769e-07,
+      "loss": 0.3578,
+      "step": 10889
+    },
+    {
+      "epoch": 99.91,
+      "learning_rate": 5.48076923076923e-07,
+      "loss": 0.4838,
+      "step": 10890
+    },
+    {
+      "epoch": 99.92,
+      "learning_rate": 5.192307692307692e-07,
+      "loss": 0.4562,
+      "step": 10891
+    },
+    {
+      "epoch": 99.93,
+      "learning_rate": 4.903846153846153e-07,
+      "loss": 0.5972,
+      "step": 10892
+    },
+    {
+      "epoch": 99.94,
+      "learning_rate": 4.615384615384615e-07,
+      "loss": 0.4555,
+      "step": 10893
+    },
+    {
+      "epoch": 99.94,
+      "learning_rate": 4.3269230769230766e-07,
+      "loss": 0.5331,
+      "step": 10894
+    },
+    {
+      "epoch": 99.95,
+      "learning_rate": 4.038461538461538e-07,
+      "loss": 0.4378,
+      "step": 10895
+    },
+    {
+      "epoch": 99.96,
+      "learning_rate": 3.7499999999999996e-07,
+      "loss": 0.5256,
+      "step": 10896
+    },
+    {
+      "epoch": 99.97,
+      "learning_rate": 3.4615384615384616e-07,
+      "loss": 0.4553,
+      "step": 10897
+    },
+    {
+      "epoch": 99.98,
+      "learning_rate": 3.1730769230769225e-07,
+      "loss": 0.4523,
+      "step": 10898
+    },
+    {
+      "epoch": 99.99,
+      "learning_rate": 2.8846153846153846e-07,
+      "loss": 0.5134,
+      "step": 10899
+    },
+    {
+      "epoch": 100.0,
+      "learning_rate": 2.596153846153846e-07,
+      "loss": 0.5448,
+      "step": 10900
+    },
+    {
+      "epoch": 100.0,
+      "step": 10900,
+      "total_flos": 4.71831915651169e+19,
+      "train_loss": 0.935905982969551,
+      "train_runtime": 6038.1706,
+      "train_samples_per_second": 57.6,
+      "train_steps_per_second": 1.805
+    }
+  ],
+  "max_steps": 10900,
+  "num_train_epochs": 100,
+  "total_flos": 4.71831915651169e+19,
+  "trial_name": null,
+  "trial_params": null
+}