diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,9979 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 15.0,
+  "global_step": 1635,
+  "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": 15.3209,
+      "step": 1
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0,
+      "loss": 13.147,
+      "step": 2
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 6e-07,
+      "loss": 14.9178,
+      "step": 3
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6e-07,
+      "loss": 15.015,
+      "step": 4
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.2e-06,
+      "loss": 14.0947,
+      "step": 5
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8e-06,
+      "loss": 14.3373,
+      "step": 6
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.4e-06,
+      "loss": 13.6543,
+      "step": 7
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 15.5996,
+      "step": 8
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 3.6e-06,
+      "loss": 18.7089,
+      "step": 9
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.2e-06,
+      "loss": 12.8116,
+      "step": 10
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.8e-06,
+      "loss": 13.158,
+      "step": 11
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.399999999999999e-06,
+      "loss": 18.8176,
+      "step": 12
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 17.4918,
+      "step": 13
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.599999999999999e-06,
+      "loss": 15.8307,
+      "step": 14
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.2e-06,
+      "loss": 13.977,
+      "step": 15
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 7.799999999999998e-06,
+      "loss": 14.3042,
+      "step": 16
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.4e-06,
+      "loss": 14.0154,
+      "step": 17
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 11.3614,
+      "step": 18
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.6e-06,
+      "loss": 12.2551,
+      "step": 19
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.02e-05,
+      "loss": 16.5141,
+      "step": 20
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.0799999999999998e-05,
+      "loss": 16.2688,
+      "step": 21
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.14e-05,
+      "loss": 16.1565,
+      "step": 22
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 17.7785,
+      "step": 23
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.26e-05,
+      "loss": 15.2434,
+      "step": 24
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.3199999999999997e-05,
+      "loss": 10.8856,
+      "step": 25
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.3799999999999998e-05,
+      "loss": 17.2036,
+      "step": 26
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.44e-05,
+      "loss": 16.1396,
+      "step": 27
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 14.0828,
+      "step": 28
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.5599999999999996e-05,
+      "loss": 19.5756,
+      "step": 29
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6199999999999997e-05,
+      "loss": 11.2803,
+      "step": 30
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68e-05,
+      "loss": 11.4906,
+      "step": 31
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.74e-05,
+      "loss": 15.734,
+      "step": 32
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 17.4073,
+      "step": 33
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.8599999999999998e-05,
+      "loss": 13.4076,
+      "step": 34
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.92e-05,
+      "loss": 15.0625,
+      "step": 35
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.98e-05,
+      "loss": 12.6117,
+      "step": 36
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.04e-05,
+      "loss": 13.0859,
+      "step": 37
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.1e-05,
+      "loss": 19.2712,
+      "step": 38
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.1599999999999996e-05,
+      "loss": 13.7634,
+      "step": 39
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2199999999999998e-05,
+      "loss": 15.5088,
+      "step": 40
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.28e-05,
+      "loss": 12.7346,
+      "step": 41
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.34e-05,
+      "loss": 13.4754,
+      "step": 42
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 11.2926,
+      "step": 43
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.4599999999999998e-05,
+      "loss": 11.7751,
+      "step": 44
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.52e-05,
+      "loss": 13.547,
+      "step": 45
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.5799999999999997e-05,
+      "loss": 12.8372,
+      "step": 46
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.6399999999999995e-05,
+      "loss": 9.8682,
+      "step": 47
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 11.9647,
+      "step": 48
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.7599999999999997e-05,
+      "loss": 9.8382,
+      "step": 49
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 9.2951,
+      "step": 50
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.88e-05,
+      "loss": 9.8869,
+      "step": 51
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.94e-05,
+      "loss": 9.8975,
+      "step": 52
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 8.9872,
+      "step": 53
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.06e-05,
+      "loss": 8.3323,
+      "step": 54
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.119999999999999e-05,
+      "loss": 8.7692,
+      "step": 55
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.1799999999999994e-05,
+      "loss": 10.5133,
+      "step": 56
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2399999999999995e-05,
+      "loss": 9.5492,
+      "step": 57
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 7.4448,
+      "step": 58
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.36e-05,
+      "loss": 6.0625,
+      "step": 59
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.42e-05,
+      "loss": 5.8929,
+      "step": 60
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.48e-05,
+      "loss": 6.1602,
+      "step": 61
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.539999999999999e-05,
+      "loss": 6.1044,
+      "step": 62
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 6.3962,
+      "step": 63
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6599999999999995e-05,
+      "loss": 5.5909,
+      "step": 64
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7199999999999996e-05,
+      "loss": 5.5223,
+      "step": 65
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.78e-05,
+      "loss": 4.7758,
+      "step": 66
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.84e-05,
+      "loss": 4.6541,
+      "step": 67
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9e-05,
+      "loss": 4.7552,
+      "step": 68
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.96e-05,
+      "loss": 4.4414,
+      "step": 69
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.02e-05,
+      "loss": 4.8146,
+      "step": 70
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.08e-05,
+      "loss": 4.8635,
+      "step": 71
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.14e-05,
+      "loss": 4.7246,
+      "step": 72
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.2e-05,
+      "loss": 4.6529,
+      "step": 73
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.259999999999999e-05,
+      "loss": 4.152,
+      "step": 74
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.319999999999999e-05,
+      "loss": 4.0489,
+      "step": 75
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3799999999999994e-05,
+      "loss": 4.3434,
+      "step": 76
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.4399999999999995e-05,
+      "loss": 4.0352,
+      "step": 77
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 4.3773,
+      "step": 78
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.56e-05,
+      "loss": 3.7607,
+      "step": 79
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.62e-05,
+      "loss": 3.8717,
+      "step": 80
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.68e-05,
+      "loss": 4.2613,
+      "step": 81
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.7399999999999993e-05,
+      "loss": 3.8807,
+      "step": 82
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 3.7571,
+      "step": 83
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.8599999999999995e-05,
+      "loss": 3.7799,
+      "step": 84
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9199999999999997e-05,
+      "loss": 3.9747,
+      "step": 85
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.98e-05,
+      "loss": 3.8203,
+      "step": 86
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.04e-05,
+      "loss": 4.2831,
+      "step": 87
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.1e-05,
+      "loss": 3.8184,
+      "step": 88
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.1599999999999994e-05,
+      "loss": 3.6599,
+      "step": 89
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.2199999999999995e-05,
+      "loss": 3.6153,
+      "step": 90
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.279999999999999e-05,
+      "loss": 3.6842,
+      "step": 91
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.339999999999999e-05,
+      "loss": 3.5289,
+      "step": 92
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 3.6055,
+      "step": 93
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.459999999999999e-05,
+      "loss": 3.7251,
+      "step": 94
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.519999999999999e-05,
+      "loss": 3.6014,
+      "step": 95
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.5799999999999994e-05,
+      "loss": 3.5082,
+      "step": 96
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 3.6797,
+      "step": 97
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 3.4808,
+      "step": 98
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.76e-05,
+      "loss": 3.3553,
+      "step": 99
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.82e-05,
+      "loss": 3.5571,
+      "step": 100
+    },
+    {
+      "epoch": 0.92,
+      "eval_loss": 3.5946969985961914,
+      "eval_runtime": 21.4628,
+      "eval_samples_per_second": 76.737,
+      "eval_steps_per_second": 1.211,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.88e-05,
+      "loss": 3.4312,
+      "step": 101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.94e-05,
+      "loss": 3.3556,
+      "step": 102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 3.4018,
+      "step": 103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.0599999999999996e-05,
+      "loss": 3.4495,
+      "step": 104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.12e-05,
+      "loss": 3.3712,
+      "step": 105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.18e-05,
+      "loss": 3.7355,
+      "step": 106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.239999999999999e-05,
+      "loss": 3.4153,
+      "step": 107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 3.345,
+      "step": 108
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.359999999999999e-05,
+      "loss": 3.3741,
+      "step": 109
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 6.419999999999999e-05,
+      "loss": 3.3753,
+      "step": 110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.479999999999999e-05,
+      "loss": 3.4028,
+      "step": 111
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 6.539999999999999e-05,
+      "loss": 3.3035,
+      "step": 112
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 3.2704,
+      "step": 113
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 6.659999999999999e-05,
+      "loss": 3.27,
+      "step": 114
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 6.72e-05,
+      "loss": 3.4096,
+      "step": 115
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 6.78e-05,
+      "loss": 3.2671,
+      "step": 116
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 6.84e-05,
+      "loss": 3.295,
+      "step": 117
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 6.9e-05,
+      "loss": 3.27,
+      "step": 118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 6.96e-05,
+      "loss": 3.3279,
+      "step": 119
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 7.02e-05,
+      "loss": 3.1743,
+      "step": 120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 7.079999999999999e-05,
+      "loss": 3.2365,
+      "step": 121
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.139999999999999e-05,
+      "loss": 3.3368,
+      "step": 122
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 3.2057,
+      "step": 123
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 7.259999999999999e-05,
+      "loss": 3.2941,
+      "step": 124
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 7.319999999999999e-05,
+      "loss": 3.238,
+      "step": 125
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 7.379999999999999e-05,
+      "loss": 3.2477,
+      "step": 126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.439999999999999e-05,
+      "loss": 3.3111,
+      "step": 127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 7.5e-05,
+      "loss": 3.3144,
+      "step": 128
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 7.56e-05,
+      "loss": 3.1385,
+      "step": 129
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.62e-05,
+      "loss": 3.1887,
+      "step": 130
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 7.68e-05,
+      "loss": 3.238,
+      "step": 131
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 7.74e-05,
+      "loss": 3.2702,
+      "step": 132
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.8e-05,
+      "loss": 3.1548,
+      "step": 133
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.86e-05,
+      "loss": 3.2073,
+      "step": 134
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.92e-05,
+      "loss": 3.279,
+      "step": 135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.98e-05,
+      "loss": 3.1951,
+      "step": 136
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.04e-05,
+      "loss": 3.1772,
+      "step": 137
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.1e-05,
+      "loss": 3.1085,
+      "step": 138
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.16e-05,
+      "loss": 3.1511,
+      "step": 139
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.22e-05,
+      "loss": 3.212,
+      "step": 140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.28e-05,
+      "loss": 3.3054,
+      "step": 141
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.34e-05,
+      "loss": 3.1368,
+      "step": 142
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.4e-05,
+      "loss": 3.1332,
+      "step": 143
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.459999999999998e-05,
+      "loss": 3.1879,
+      "step": 144
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.519999999999998e-05,
+      "loss": 3.1274,
+      "step": 145
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.579999999999998e-05,
+      "loss": 3.1131,
+      "step": 146
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.639999999999999e-05,
+      "loss": 3.2303,
+      "step": 147
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 3.1415,
+      "step": 148
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.759999999999999e-05,
+      "loss": 3.1264,
+      "step": 149
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.819999999999999e-05,
+      "loss": 2.9943,
+      "step": 150
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.879999999999999e-05,
+      "loss": 3.1302,
+      "step": 151
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.939999999999999e-05,
+      "loss": 3.0538,
+      "step": 152
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 3.0965,
+      "step": 153
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.059999999999999e-05,
+      "loss": 3.0977,
+      "step": 154
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 9.12e-05,
+      "loss": 3.1288,
+      "step": 155
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 9.18e-05,
+      "loss": 3.1918,
+      "step": 156
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.24e-05,
+      "loss": 3.11,
+      "step": 157
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 9.3e-05,
+      "loss": 3.0668,
+      "step": 158
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 9.36e-05,
+      "loss": 3.1779,
+      "step": 159
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.419999999999999e-05,
+      "loss": 3.1454,
+      "step": 160
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.479999999999999e-05,
+      "loss": 3.1836,
+      "step": 161
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 9.539999999999999e-05,
+      "loss": 3.0508,
+      "step": 162
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 3.119,
+      "step": 163
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.659999999999999e-05,
+      "loss": 3.1405,
+      "step": 164
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.719999999999999e-05,
+      "loss": 3.1866,
+      "step": 165
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.779999999999999e-05,
+      "loss": 3.1043,
+      "step": 166
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.839999999999999e-05,
+      "loss": 3.017,
+      "step": 167
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.9e-05,
+      "loss": 3.0101,
+      "step": 168
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.96e-05,
+      "loss": 3.0992,
+      "step": 169
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001002,
+      "loss": 3.1166,
+      "step": 170
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001008,
+      "loss": 3.1263,
+      "step": 171
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001014,
+      "loss": 3.1716,
+      "step": 172
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.000102,
+      "loss": 3.1659,
+      "step": 173
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001026,
+      "loss": 3.1435,
+      "step": 174
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00010319999999999999,
+      "loss": 2.9801,
+      "step": 175
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00010379999999999999,
+      "loss": 3.0866,
+      "step": 176
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00010439999999999999,
+      "loss": 3.1187,
+      "step": 177
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 2.996,
+      "step": 178
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00010559999999999998,
+      "loss": 3.083,
+      "step": 179
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00010619999999999998,
+      "loss": 3.0235,
+      "step": 180
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00010679999999999998,
+      "loss": 3.0856,
+      "step": 181
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00010739999999999998,
+      "loss": 2.9631,
+      "step": 182
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 3.0156,
+      "step": 183
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00010859999999999998,
+      "loss": 3.1007,
+      "step": 184
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00010919999999999998,
+      "loss": 2.9965,
+      "step": 185
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00010979999999999999,
+      "loss": 2.9686,
+      "step": 186
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00011039999999999999,
+      "loss": 3.0729,
+      "step": 187
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 3.0888,
+      "step": 188
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00011159999999999999,
+      "loss": 3.0148,
+      "step": 189
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00011219999999999999,
+      "loss": 2.9947,
+      "step": 190
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011279999999999999,
+      "loss": 3.0686,
+      "step": 191
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011339999999999999,
+      "loss": 3.0438,
+      "step": 192
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 3.0385,
+      "step": 193
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001146,
+      "loss": 3.0348,
+      "step": 194
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001152,
+      "loss": 3.0041,
+      "step": 195
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001158,
+      "loss": 3.0538,
+      "step": 196
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001164,
+      "loss": 3.054,
+      "step": 197
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.000117,
+      "loss": 3.0025,
+      "step": 198
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001176,
+      "loss": 3.01,
+      "step": 199
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001182,
+      "loss": 2.9269,
+      "step": 200
+    },
+    {
+      "epoch": 1.83,
+      "eval_loss": 3.009141206741333,
+      "eval_runtime": 20.4711,
+      "eval_samples_per_second": 80.455,
+      "eval_steps_per_second": 1.27,
+      "eval_wer": 0.9998978653865795,
+      "step": 200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001188,
+      "loss": 2.9697,
+      "step": 201
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001194,
+      "loss": 2.91,
+      "step": 202
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 3.0262,
+      "step": 203
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00012059999999999999,
+      "loss": 2.9469,
+      "step": 204
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00012119999999999999,
+      "loss": 2.9809,
+      "step": 205
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00012179999999999999,
+      "loss": 2.9119,
+      "step": 206
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001224,
+      "loss": 2.8735,
+      "step": 207
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 2.9211,
+      "step": 208
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001236,
+      "loss": 3.0556,
+      "step": 209
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00012419999999999998,
+      "loss": 2.9629,
+      "step": 210
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00012479999999999997,
+      "loss": 2.9254,
+      "step": 211
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00012539999999999999,
+      "loss": 2.9069,
+      "step": 212
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 2.9407,
+      "step": 213
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001266,
+      "loss": 2.9854,
+      "step": 214
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00012719999999999997,
+      "loss": 2.8615,
+      "step": 215
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0001278,
+      "loss": 2.981,
+      "step": 216
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00012839999999999998,
+      "loss": 2.8677,
+      "step": 217
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.000129,
+      "loss": 2.9314,
+      "step": 218
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00012959999999999998,
+      "loss": 2.9306,
+      "step": 219
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0001302,
+      "loss": 2.912,
+      "step": 220
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00013079999999999998,
+      "loss": 2.8527,
+      "step": 221
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0001314,
+      "loss": 2.7591,
+      "step": 222
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 2.8076,
+      "step": 223
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0001326,
+      "loss": 2.9137,
+      "step": 224
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00013319999999999999,
+      "loss": 2.8311,
+      "step": 225
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0001338,
+      "loss": 2.8666,
+      "step": 226
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0001344,
+      "loss": 2.7577,
+      "step": 227
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.000135,
+      "loss": 2.6859,
+      "step": 228
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0001356,
+      "loss": 2.6957,
+      "step": 229
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0001362,
+      "loss": 2.6024,
+      "step": 230
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0001368,
+      "loss": 2.6208,
+      "step": 231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0001374,
+      "loss": 2.5696,
+      "step": 232
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.000138,
+      "loss": 2.5933,
+      "step": 233
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0001386,
+      "loss": 2.6552,
+      "step": 234
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0001392,
+      "loss": 2.5771,
+      "step": 235
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00013979999999999998,
+      "loss": 2.4974,
+      "step": 236
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0001404,
+      "loss": 2.5139,
+      "step": 237
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 2.6153,
+      "step": 238
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00014159999999999997,
+      "loss": 2.5096,
+      "step": 239
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0001422,
+      "loss": 2.5168,
+      "step": 240
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00014279999999999997,
+      "loss": 2.5182,
+      "step": 241
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0001434,
+      "loss": 2.4636,
+      "step": 242
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 2.2496,
+      "step": 243
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0001446,
+      "loss": 2.3962,
+      "step": 244
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00014519999999999998,
+      "loss": 2.3955,
+      "step": 245
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0001458,
+      "loss": 2.2126,
+      "step": 246
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00014639999999999998,
+      "loss": 2.3604,
+      "step": 247
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000147,
+      "loss": 2.3044,
+      "step": 248
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00014759999999999998,
+      "loss": 2.2575,
+      "step": 249
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0001482,
+      "loss": 2.139,
+      "step": 250
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00014879999999999998,
+      "loss": 2.2387,
+      "step": 251
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0001494,
+      "loss": 2.2734,
+      "step": 252
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00015,
+      "loss": 2.0964,
+      "step": 253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00015059999999999997,
+      "loss": 2.2254,
+      "step": 254
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0001512,
+      "loss": 1.9322,
+      "step": 255
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00015179999999999998,
+      "loss": 2.0015,
+      "step": 256
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0001524,
+      "loss": 1.8817,
+      "step": 257
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 1.7881,
+      "step": 258
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0001536,
+      "loss": 1.8138,
+      "step": 259
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00015419999999999998,
+      "loss": 1.6331,
+      "step": 260
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0001548,
+      "loss": 1.6477,
+      "step": 261
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00015539999999999998,
+      "loss": 1.6959,
+      "step": 262
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.000156,
+      "loss": 1.8084,
+      "step": 263
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00015659999999999998,
+      "loss": 1.5961,
+      "step": 264
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0001572,
+      "loss": 1.662,
+      "step": 265
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0001578,
+      "loss": 1.4023,
+      "step": 266
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0001584,
+      "loss": 1.6648,
+      "step": 267
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.000159,
+      "loss": 1.4555,
+      "step": 268
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0001596,
+      "loss": 1.6013,
+      "step": 269
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0001602,
+      "loss": 1.3135,
+      "step": 270
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0001608,
+      "loss": 1.5808,
+      "step": 271
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0001614,
+      "loss": 1.7059,
+      "step": 272
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000162,
+      "loss": 1.3777,
+      "step": 273
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0001626,
+      "loss": 1.2551,
+      "step": 274
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0001632,
+      "loss": 1.3435,
+      "step": 275
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0001638,
+      "loss": 1.4827,
+      "step": 276
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0001644,
+      "loss": 1.2668,
+      "step": 277
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.000165,
+      "loss": 1.2435,
+      "step": 278
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0001656,
+      "loss": 1.286,
+      "step": 279
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0001662,
+      "loss": 1.6001,
+      "step": 280
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0001668,
+      "loss": 1.2683,
+      "step": 281
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0001674,
+      "loss": 1.4337,
+      "step": 282
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.000168,
+      "loss": 0.9217,
+      "step": 283
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0001686,
+      "loss": 1.2177,
+      "step": 284
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00016919999999999997,
+      "loss": 1.0393,
+      "step": 285
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00016979999999999998,
+      "loss": 1.1669,
+      "step": 286
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00017039999999999997,
+      "loss": 0.9628,
+      "step": 287
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 1.1075,
+      "step": 288
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00017159999999999997,
+      "loss": 0.9624,
+      "step": 289
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00017219999999999998,
+      "loss": 0.9703,
+      "step": 290
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00017279999999999997,
+      "loss": 1.0271,
+      "step": 291
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00017339999999999996,
+      "loss": 1.2429,
+      "step": 292
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 1.0191,
+      "step": 293
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00017459999999999996,
+      "loss": 0.8872,
+      "step": 294
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00017519999999999998,
+      "loss": 0.9615,
+      "step": 295
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00017579999999999996,
+      "loss": 0.8433,
+      "step": 296
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00017639999999999998,
+      "loss": 1.0205,
+      "step": 297
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 1.1264,
+      "step": 298
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00017759999999999998,
+      "loss": 1.069,
+      "step": 299
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00017819999999999997,
+      "loss": 0.7958,
+      "step": 300
+    },
+    {
+      "epoch": 2.75,
+      "eval_loss": 0.9092942476272583,
+      "eval_runtime": 20.7101,
+      "eval_samples_per_second": 79.526,
+      "eval_steps_per_second": 1.255,
+      "eval_wer": 0.805637830660811,
+      "step": 300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.00017879999999999998,
+      "loss": 0.9554,
+      "step": 301
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00017939999999999997,
+      "loss": 1.1591,
+      "step": 302
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 1.1534,
+      "step": 303
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00018059999999999997,
+      "loss": 1.109,
+      "step": 304
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00018119999999999999,
+      "loss": 0.8138,
+      "step": 305
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00018179999999999997,
+      "loss": 0.8022,
+      "step": 306
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0001824,
+      "loss": 0.8003,
+      "step": 307
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 0.893,
+      "step": 308
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0001836,
+      "loss": 1.0309,
+      "step": 309
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00018419999999999998,
+      "loss": 0.9358,
+      "step": 310
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0001848,
+      "loss": 0.738,
+      "step": 311
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00018539999999999998,
+      "loss": 0.657,
+      "step": 312
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.000186,
+      "loss": 0.9043,
+      "step": 313
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00018659999999999998,
+      "loss": 0.7584,
+      "step": 314
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0001872,
+      "loss": 0.8219,
+      "step": 315
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00018779999999999998,
+      "loss": 0.673,
+      "step": 316
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00018839999999999997,
+      "loss": 0.8785,
+      "step": 317
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 1.2493,
+      "step": 318
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.00018959999999999997,
+      "loss": 0.8458,
+      "step": 319
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0001902,
+      "loss": 0.8645,
+      "step": 320
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00019079999999999998,
+      "loss": 0.8572,
+      "step": 321
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0001914,
+      "loss": 0.7713,
+      "step": 322
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 0.7779,
+      "step": 323
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0001926,
+      "loss": 1.1098,
+      "step": 324
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00019319999999999998,
+      "loss": 0.719,
+      "step": 325
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0001938,
+      "loss": 0.6919,
+      "step": 326
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00019439999999999998,
+      "loss": 0.7991,
+      "step": 327
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.000195,
+      "loss": 0.5577,
+      "step": 328
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00019559999999999998,
+      "loss": 0.5955,
+      "step": 329
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0001962,
+      "loss": 0.7619,
+      "step": 330
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00019679999999999999,
+      "loss": 0.649,
+      "step": 331
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0001974,
+      "loss": 0.6284,
+      "step": 332
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.000198,
+      "loss": 0.7406,
+      "step": 333
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0001986,
+      "loss": 0.591,
+      "step": 334
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0001992,
+      "loss": 0.896,
+      "step": 335
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0001998,
+      "loss": 0.5065,
+      "step": 336
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002004,
+      "loss": 0.4696,
+      "step": 337
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.000201,
+      "loss": 0.8985,
+      "step": 338
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002016,
+      "loss": 0.8283,
+      "step": 339
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002022,
+      "loss": 0.7275,
+      "step": 340
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002028,
+      "loss": 0.4211,
+      "step": 341
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00020339999999999998,
+      "loss": 0.5277,
+      "step": 342
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.000204,
+      "loss": 0.9494,
+      "step": 343
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00020459999999999999,
+      "loss": 0.6271,
+      "step": 344
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002052,
+      "loss": 0.7927,
+      "step": 345
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002058,
+      "loss": 0.6099,
+      "step": 346
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00020639999999999998,
+      "loss": 0.7043,
+      "step": 347
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 0.7124,
+      "step": 348
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00020759999999999998,
+      "loss": 0.5296,
+      "step": 349
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00020819999999999996,
+      "loss": 0.7337,
+      "step": 350
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00020879999999999998,
+      "loss": 0.7838,
+      "step": 351
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00020939999999999997,
+      "loss": 0.7439,
+      "step": 352
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 0.8067,
+      "step": 353
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00021059999999999997,
+      "loss": 0.6174,
+      "step": 354
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00021119999999999996,
+      "loss": 0.6622,
+      "step": 355
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00021179999999999997,
+      "loss": 0.6704,
+      "step": 356
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00021239999999999996,
+      "loss": 0.544,
+      "step": 357
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 0.4399,
+      "step": 358
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00021359999999999996,
+      "loss": 0.355,
+      "step": 359
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00021419999999999998,
+      "loss": 0.4169,
+      "step": 360
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00021479999999999996,
+      "loss": 0.6259,
+      "step": 361
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00021539999999999998,
+      "loss": 0.8317,
+      "step": 362
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 0.4265,
+      "step": 363
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00021659999999999998,
+      "loss": 0.9351,
+      "step": 364
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00021719999999999997,
+      "loss": 0.4256,
+      "step": 365
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00021779999999999998,
+      "loss": 0.7961,
+      "step": 366
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00021839999999999997,
+      "loss": 0.4537,
+      "step": 367
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 0.4643,
+      "step": 368
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00021959999999999997,
+      "loss": 0.4404,
+      "step": 369
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00022019999999999999,
+      "loss": 0.5353,
+      "step": 370
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00022079999999999997,
+      "loss": 0.6515,
+      "step": 371
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002214,
+      "loss": 0.4657,
+      "step": 372
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 0.6809,
+      "step": 373
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002226,
+      "loss": 0.4238,
+      "step": 374
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00022319999999999998,
+      "loss": 0.616,
+      "step": 375
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002238,
+      "loss": 0.501,
+      "step": 376
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00022439999999999998,
+      "loss": 0.6904,
+      "step": 377
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.000225,
+      "loss": 0.4766,
+      "step": 378
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00022559999999999998,
+      "loss": 0.5959,
+      "step": 379
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00022619999999999997,
+      "loss": 0.8233,
+      "step": 380
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00022679999999999998,
+      "loss": 0.9804,
+      "step": 381
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00022739999999999997,
+      "loss": 0.7232,
+      "step": 382
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 0.6951,
+      "step": 383
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00022859999999999997,
+      "loss": 0.381,
+      "step": 384
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002292,
+      "loss": 0.5776,
+      "step": 385
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.00022979999999999997,
+      "loss": 0.7077,
+      "step": 386
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002304,
+      "loss": 0.5536,
+      "step": 387
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 0.7417,
+      "step": 388
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002316,
+      "loss": 0.6552,
+      "step": 389
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00023219999999999998,
+      "loss": 0.2999,
+      "step": 390
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002328,
+      "loss": 0.5058,
+      "step": 391
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00023339999999999998,
+      "loss": 0.4915,
+      "step": 392
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.000234,
+      "loss": 0.6512,
+      "step": 393
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00023459999999999998,
+      "loss": 0.4409,
+      "step": 394
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002352,
+      "loss": 0.5179,
+      "step": 395
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00023579999999999999,
+      "loss": 0.5347,
+      "step": 396
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002364,
+      "loss": 0.6048,
+      "step": 397
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.000237,
+      "loss": 0.5515,
+      "step": 398
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002376,
+      "loss": 0.5012,
+      "step": 399
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002382,
+      "loss": 0.6542,
+      "step": 400
+    },
+    {
+      "epoch": 3.67,
+      "eval_loss": 0.5963820815086365,
+      "eval_runtime": 20.672,
+      "eval_samples_per_second": 79.673,
+      "eval_steps_per_second": 1.258,
+      "eval_wer": 0.609641507506894,
+      "step": 400
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002388,
+      "loss": 0.7643,
+      "step": 401
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002394,
+      "loss": 0.6319,
+      "step": 402
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.8255,
+      "step": 403
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002406,
+      "loss": 0.7163,
+      "step": 404
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00024119999999999998,
+      "loss": 0.7377,
+      "step": 405
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002418,
+      "loss": 0.5242,
+      "step": 406
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00024239999999999998,
+      "loss": 0.7683,
+      "step": 407
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.000243,
+      "loss": 0.7626,
+      "step": 408
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00024359999999999999,
+      "loss": 0.4409,
+      "step": 409
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00024419999999999997,
+      "loss": 0.5297,
+      "step": 410
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002448,
+      "loss": 0.3854,
+      "step": 411
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00024539999999999995,
+      "loss": 0.5129,
+      "step": 412
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 0.6171,
+      "step": 413
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002466,
+      "loss": 0.4093,
+      "step": 414
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0002472,
+      "loss": 0.5943,
+      "step": 415
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00024779999999999995,
+      "loss": 0.4809,
+      "step": 416
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00024839999999999997,
+      "loss": 0.6241,
+      "step": 417
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000249,
+      "loss": 0.6133,
+      "step": 418
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00024959999999999994,
+      "loss": 0.916,
+      "step": 419
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00025019999999999996,
+      "loss": 0.4945,
+      "step": 420
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00025079999999999997,
+      "loss": 0.5202,
+      "step": 421
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0002514,
+      "loss": 0.3996,
+      "step": 422
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 0.3398,
+      "step": 423
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00025259999999999996,
+      "loss": 0.5793,
+      "step": 424
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.0002532,
+      "loss": 0.533,
+      "step": 425
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0002538,
+      "loss": 0.4147,
+      "step": 426
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00025439999999999995,
+      "loss": 0.3975,
+      "step": 427
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 0.6206,
+      "step": 428
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002556,
+      "loss": 0.8133,
+      "step": 429
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0002562,
+      "loss": 0.5858,
+      "step": 430
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00025679999999999995,
+      "loss": 0.7453,
+      "step": 431
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00025739999999999997,
+      "loss": 0.4893,
+      "step": 432
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000258,
+      "loss": 0.516,
+      "step": 433
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0002586,
+      "loss": 0.4342,
+      "step": 434
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00025919999999999996,
+      "loss": 0.4436,
+      "step": 435
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00025979999999999997,
+      "loss": 1.1074,
+      "step": 436
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002604,
+      "loss": 0.2794,
+      "step": 437
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.000261,
+      "loss": 0.3083,
+      "step": 438
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.00026159999999999996,
+      "loss": 0.2644,
+      "step": 439
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002622,
+      "loss": 0.2665,
+      "step": 440
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0002628,
+      "loss": 0.32,
+      "step": 441
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00026339999999999995,
+      "loss": 0.5356,
+      "step": 442
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 0.2211,
+      "step": 443
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002646,
+      "loss": 0.4744,
+      "step": 444
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002652,
+      "loss": 0.3793,
+      "step": 445
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.00026579999999999996,
+      "loss": 0.275,
+      "step": 446
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00026639999999999997,
+      "loss": 0.3713,
+      "step": 447
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.000267,
+      "loss": 0.7011,
+      "step": 448
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002676,
+      "loss": 0.1956,
+      "step": 449
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00026819999999999996,
+      "loss": 0.5551,
+      "step": 450
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0002688,
+      "loss": 0.5263,
+      "step": 451
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002694,
+      "loss": 0.3182,
+      "step": 452
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00027,
+      "loss": 0.5185,
+      "step": 453
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00027059999999999996,
+      "loss": 0.3978,
+      "step": 454
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0002712,
+      "loss": 0.1956,
+      "step": 455
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0002718,
+      "loss": 0.5373,
+      "step": 456
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0002724,
+      "loss": 0.3572,
+      "step": 457
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 0.355,
+      "step": 458
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0002736,
+      "loss": 0.451,
+      "step": 459
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002742,
+      "loss": 0.487,
+      "step": 460
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002748,
+      "loss": 0.3208,
+      "step": 461
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.00027539999999999997,
+      "loss": 0.221,
+      "step": 462
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.000276,
+      "loss": 0.6257,
+      "step": 463
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.0002766,
+      "loss": 0.3882,
+      "step": 464
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0002772,
+      "loss": 0.4665,
+      "step": 465
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002778,
+      "loss": 0.5998,
+      "step": 466
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002784,
+      "loss": 0.2316,
+      "step": 467
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.000279,
+      "loss": 0.419,
+      "step": 468
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00027959999999999997,
+      "loss": 0.4161,
+      "step": 469
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0002802,
+      "loss": 0.2653,
+      "step": 470
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002808,
+      "loss": 0.3274,
+      "step": 471
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00028139999999999996,
+      "loss": 0.3944,
+      "step": 472
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 0.4319,
+      "step": 473
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0002826,
+      "loss": 0.2908,
+      "step": 474
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.00028319999999999994,
+      "loss": 0.4891,
+      "step": 475
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00028379999999999996,
+      "loss": 0.2903,
+      "step": 476
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0002844,
+      "loss": 0.6048,
+      "step": 477
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.000285,
+      "loss": 0.5717,
+      "step": 478
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00028559999999999995,
+      "loss": 0.2615,
+      "step": 479
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00028619999999999996,
+      "loss": 0.1936,
+      "step": 480
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002868,
+      "loss": 0.316,
+      "step": 481
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.00028739999999999994,
+      "loss": 0.4063,
+      "step": 482
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 0.4697,
+      "step": 483
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00028859999999999997,
+      "loss": 0.2494,
+      "step": 484
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.0002892,
+      "loss": 0.2499,
+      "step": 485
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00028979999999999994,
+      "loss": 0.7209,
+      "step": 486
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00029039999999999996,
+      "loss": 0.2894,
+      "step": 487
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 0.5228,
+      "step": 488
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002916,
+      "loss": 0.2896,
+      "step": 489
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00029219999999999995,
+      "loss": 0.4919,
+      "step": 490
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00029279999999999996,
+      "loss": 0.2049,
+      "step": 491
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002934,
+      "loss": 0.6207,
+      "step": 492
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.000294,
+      "loss": 0.2978,
+      "step": 493
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00029459999999999995,
+      "loss": 0.271,
+      "step": 494
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00029519999999999997,
+      "loss": 0.4462,
+      "step": 495
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.0002958,
+      "loss": 0.3649,
+      "step": 496
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.0002964,
+      "loss": 0.3818,
+      "step": 497
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 0.428,
+      "step": 498
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00029759999999999997,
+      "loss": 0.2635,
+      "step": 499
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.0002982,
+      "loss": 0.4491,
+      "step": 500
+    },
+    {
+      "epoch": 4.59,
+      "eval_loss": 0.5220645666122437,
+      "eval_runtime": 20.7654,
+      "eval_samples_per_second": 79.314,
+      "eval_steps_per_second": 1.252,
+      "eval_wer": 0.533040547441528,
+      "step": 500
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0002988,
+      "loss": 0.472,
+      "step": 501
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.00029939999999999996,
+      "loss": 0.4015,
+      "step": 502
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0003,
+      "loss": 0.6159,
+      "step": 503
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00029973568281938326,
+      "loss": 0.8724,
+      "step": 504
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.0002994713656387665,
+      "loss": 0.3187,
+      "step": 505
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.0002992070484581498,
+      "loss": 0.3546,
+      "step": 506
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.000298942731277533,
+      "loss": 0.6029,
+      "step": 507
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00029867841409691624,
+      "loss": 0.3422,
+      "step": 508
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0002984140969162995,
+      "loss": 0.6309,
+      "step": 509
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0002981497797356828,
+      "loss": 0.3484,
+      "step": 510
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00029788546255506604,
+      "loss": 0.5814,
+      "step": 511
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0002976211453744493,
+      "loss": 0.4075,
+      "step": 512
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0002973568281938326,
+      "loss": 0.3013,
+      "step": 513
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00029709251101321584,
+      "loss": 0.2557,
+      "step": 514
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00029682819383259907,
+      "loss": 0.263,
+      "step": 515
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00029656387665198236,
+      "loss": 0.1715,
+      "step": 516
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0002962995594713656,
+      "loss": 0.3834,
+      "step": 517
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0002960352422907489,
+      "loss": 0.3456,
+      "step": 518
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00029577092511013216,
+      "loss": 0.3018,
+      "step": 519
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0002955066079295154,
+      "loss": 0.38,
+      "step": 520
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0002952422907488987,
+      "loss": 0.3134,
+      "step": 521
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0002949779735682819,
+      "loss": 0.4132,
+      "step": 522
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002947136563876652,
+      "loss": 0.0814,
+      "step": 523
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0002944493392070484,
+      "loss": 0.2914,
+      "step": 524
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002941850220264317,
+      "loss": 0.2797,
+      "step": 525
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00029392070484581494,
+      "loss": 0.2037,
+      "step": 526
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0002936563876651982,
+      "loss": 0.4366,
+      "step": 527
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0002933920704845815,
+      "loss": 0.4666,
+      "step": 528
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.00029312775330396474,
+      "loss": 0.5651,
+      "step": 529
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00029286343612334797,
+      "loss": 0.7109,
+      "step": 530
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00029259911894273126,
+      "loss": 0.3089,
+      "step": 531
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0002923348017621145,
+      "loss": 0.3013,
+      "step": 532
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.00029207048458149777,
+      "loss": 0.4648,
+      "step": 533
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00029180616740088106,
+      "loss": 0.3372,
+      "step": 534
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0002915418502202643,
+      "loss": 0.8511,
+      "step": 535
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.0002912775330396475,
+      "loss": 0.2407,
+      "step": 536
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0002910132158590308,
+      "loss": 0.5248,
+      "step": 537
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002907488986784141,
+      "loss": 0.2598,
+      "step": 538
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002904845814977973,
+      "loss": 0.2276,
+      "step": 539
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0002902202643171806,
+      "loss": 0.3091,
+      "step": 540
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00028995594713656384,
+      "loss": 0.5105,
+      "step": 541
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.0002896916299559471,
+      "loss": 0.3395,
+      "step": 542
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00028942731277533035,
+      "loss": 0.3658,
+      "step": 543
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00028916299559471364,
+      "loss": 0.3329,
+      "step": 544
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00028889867841409687,
+      "loss": 0.3864,
+      "step": 545
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.00028863436123348016,
+      "loss": 0.2192,
+      "step": 546
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00028837004405286344,
+      "loss": 0.3083,
+      "step": 547
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00028810572687224667,
+      "loss": 0.2032,
+      "step": 548
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00028784140969162996,
+      "loss": 0.2958,
+      "step": 549
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0002875770925110132,
+      "loss": 0.2712,
+      "step": 550
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002873127753303964,
+      "loss": 0.3198,
+      "step": 551
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002870484581497797,
+      "loss": 0.1019,
+      "step": 552
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.000286784140969163,
+      "loss": 0.0742,
+      "step": 553
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0002865198237885462,
+      "loss": 0.1326,
+      "step": 554
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002862555066079295,
+      "loss": 0.2875,
+      "step": 555
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00028599118942731274,
+      "loss": 0.4097,
+      "step": 556
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.000285726872246696,
+      "loss": 0.2043,
+      "step": 557
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00028546255506607925,
+      "loss": 0.0699,
+      "step": 558
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00028519823788546254,
+      "loss": 0.1614,
+      "step": 559
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00028493392070484577,
+      "loss": 0.1843,
+      "step": 560
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00028466960352422905,
+      "loss": 0.1655,
+      "step": 561
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00028440528634361234,
+      "loss": 0.2239,
+      "step": 562
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00028414096916299557,
+      "loss": 0.3169,
+      "step": 563
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00028387665198237886,
+      "loss": 0.2458,
+      "step": 564
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0002836123348017621,
+      "loss": 0.2688,
+      "step": 565
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0002833480176211453,
+      "loss": 0.2578,
+      "step": 566
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0002830837004405286,
+      "loss": 0.2673,
+      "step": 567
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.0002828193832599119,
+      "loss": 0.1164,
+      "step": 568
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002825550660792951,
+      "loss": 0.3038,
+      "step": 569
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0002822907488986784,
+      "loss": 0.3031,
+      "step": 570
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00028202643171806164,
+      "loss": 0.2212,
+      "step": 571
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00028176211453744487,
+      "loss": 0.2269,
+      "step": 572
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00028149779735682815,
+      "loss": 0.3686,
+      "step": 573
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.00028123348017621144,
+      "loss": 0.2948,
+      "step": 574
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00028096916299559467,
+      "loss": 0.2818,
+      "step": 575
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00028070484581497795,
+      "loss": 0.4305,
+      "step": 576
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00028044052863436124,
+      "loss": 0.1088,
+      "step": 577
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00028017621145374447,
+      "loss": 0.1838,
+      "step": 578
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0002799118942731277,
+      "loss": 0.3772,
+      "step": 579
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.000279647577092511,
+      "loss": 0.2409,
+      "step": 580
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.0002793832599118942,
+      "loss": 0.5439,
+      "step": 581
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0002791189427312775,
+      "loss": 0.2108,
+      "step": 582
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0002788546255506608,
+      "loss": 0.0994,
+      "step": 583
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.000278590308370044,
+      "loss": 0.258,
+      "step": 584
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002783259911894273,
+      "loss": 0.2255,
+      "step": 585
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0002780616740088106,
+      "loss": 0.2307,
+      "step": 586
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0002777973568281938,
+      "loss": 0.3816,
+      "step": 587
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00027753303964757705,
+      "loss": 0.2057,
+      "step": 588
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00027726872246696034,
+      "loss": 0.3418,
+      "step": 589
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00027700440528634357,
+      "loss": 0.2163,
+      "step": 590
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00027674008810572685,
+      "loss": 0.148,
+      "step": 591
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00027647577092511014,
+      "loss": 0.1867,
+      "step": 592
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00027621145374449337,
+      "loss": 0.3414,
+      "step": 593
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.0002759471365638766,
+      "loss": 0.4121,
+      "step": 594
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0002756828193832599,
+      "loss": 0.2797,
+      "step": 595
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002754185022026431,
+      "loss": 0.2184,
+      "step": 596
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.0002751541850220264,
+      "loss": 0.0965,
+      "step": 597
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.0002748898678414097,
+      "loss": 0.4278,
+      "step": 598
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002746255506607929,
+      "loss": 0.323,
+      "step": 599
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002743612334801762,
+      "loss": 0.279,
+      "step": 600
+    },
+    {
+      "epoch": 5.5,
+      "eval_loss": 0.4335188865661621,
+      "eval_runtime": 20.6813,
+      "eval_samples_per_second": 79.637,
+      "eval_steps_per_second": 1.257,
+      "eval_wer": 0.4752323562455316,
+      "step": 600
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00027409691629955943,
+      "loss": 0.3165,
+      "step": 601
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002738325991189427,
+      "loss": 0.1385,
+      "step": 602
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.00027356828193832595,
+      "loss": 0.2958,
+      "step": 603
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00027330396475770924,
+      "loss": 0.229,
+      "step": 604
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.00027303964757709247,
+      "loss": 0.1618,
+      "step": 605
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00027277533039647575,
+      "loss": 0.2673,
+      "step": 606
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00027251101321585904,
+      "loss": 0.3943,
+      "step": 607
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00027224669603524227,
+      "loss": 0.2042,
+      "step": 608
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.0002719823788546255,
+      "loss": 0.2477,
+      "step": 609
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002717180616740088,
+      "loss": 0.2228,
+      "step": 610
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00027145374449339207,
+      "loss": 0.1296,
+      "step": 611
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0002711894273127753,
+      "loss": 0.3615,
+      "step": 612
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.0002709251101321586,
+      "loss": 0.1003,
+      "step": 613
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0002706607929515418,
+      "loss": 0.1588,
+      "step": 614
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.0002703964757709251,
+      "loss": 0.5409,
+      "step": 615
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00027013215859030833,
+      "loss": 0.2075,
+      "step": 616
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0002698678414096916,
+      "loss": 0.1424,
+      "step": 617
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.00026960352422907485,
+      "loss": 0.0893,
+      "step": 618
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00026933920704845814,
+      "loss": 0.3982,
+      "step": 619
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00026907488986784137,
+      "loss": 0.2859,
+      "step": 620
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00026881057268722465,
+      "loss": 0.1736,
+      "step": 621
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00026854625550660794,
+      "loss": 0.3298,
+      "step": 622
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00026828193832599117,
+      "loss": 0.1526,
+      "step": 623
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0002680176211453744,
+      "loss": 0.1985,
+      "step": 624
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0002677533039647577,
+      "loss": 0.3154,
+      "step": 625
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00026748898678414097,
+      "loss": 0.2743,
+      "step": 626
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002672246696035242,
+      "loss": 0.1987,
+      "step": 627
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.0002669603524229075,
+      "loss": 0.2059,
+      "step": 628
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0002666960352422907,
+      "loss": 0.4123,
+      "step": 629
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00026643171806167395,
+      "loss": 0.213,
+      "step": 630
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00026616740088105723,
+      "loss": 0.1625,
+      "step": 631
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0002659030837004405,
+      "loss": 0.2305,
+      "step": 632
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.00026563876651982375,
+      "loss": 0.1201,
+      "step": 633
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00026537444933920703,
+      "loss": 0.1377,
+      "step": 634
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0002651101321585903,
+      "loss": 0.1392,
+      "step": 635
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00026484581497797355,
+      "loss": 0.3582,
+      "step": 636
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0002645814977973568,
+      "loss": 0.434,
+      "step": 637
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00026431718061674007,
+      "loss": 0.2696,
+      "step": 638
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002640528634361233,
+      "loss": 0.309,
+      "step": 639
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0002637885462555066,
+      "loss": 0.1253,
+      "step": 640
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00026352422907488987,
+      "loss": 0.1943,
+      "step": 641
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0002632599118942731,
+      "loss": 0.211,
+      "step": 642
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0002629955947136564,
+      "loss": 0.1014,
+      "step": 643
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0002627312775330396,
+      "loss": 0.4637,
+      "step": 644
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00026246696035242285,
+      "loss": 0.4484,
+      "step": 645
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00026220264317180613,
+      "loss": 0.2543,
+      "step": 646
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0002619383259911894,
+      "loss": 0.1228,
+      "step": 647
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00026167400881057265,
+      "loss": 0.3288,
+      "step": 648
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00026140969162995593,
+      "loss": 0.1228,
+      "step": 649
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0002611453744493392,
+      "loss": 0.2023,
+      "step": 650
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00026088105726872245,
+      "loss": 0.5295,
+      "step": 651
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.0002606167400881057,
+      "loss": 0.0814,
+      "step": 652
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00026035242290748897,
+      "loss": 0.2237,
+      "step": 653
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0002600881057268722,
+      "loss": 0.2197,
+      "step": 654
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0002598237885462555,
+      "loss": 0.1151,
+      "step": 655
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00025955947136563877,
+      "loss": 0.244,
+      "step": 656
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.000259295154185022,
+      "loss": 0.1785,
+      "step": 657
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0002590308370044053,
+      "loss": 0.2579,
+      "step": 658
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0002587665198237885,
+      "loss": 0.2572,
+      "step": 659
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0002585022026431718,
+      "loss": 0.2055,
+      "step": 660
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00025823788546255503,
+      "loss": 0.1603,
+      "step": 661
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0002579735682819383,
+      "loss": 0.3012,
+      "step": 662
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00025770925110132155,
+      "loss": 0.1636,
+      "step": 663
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00025744493392070483,
+      "loss": 0.294,
+      "step": 664
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.0002571806167400881,
+      "loss": 0.2247,
+      "step": 665
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00025691629955947135,
+      "loss": 0.3456,
+      "step": 666
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0002566519823788546,
+      "loss": 0.0356,
+      "step": 667
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00025638766519823787,
+      "loss": 0.2314,
+      "step": 668
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002561233480176211,
+      "loss": 0.25,
+      "step": 669
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0002558590308370044,
+      "loss": 0.1275,
+      "step": 670
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00025559471365638767,
+      "loss": 0.4078,
+      "step": 671
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0002553303964757709,
+      "loss": 0.2236,
+      "step": 672
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00025506607929515413,
+      "loss": 0.1344,
+      "step": 673
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.0002548017621145374,
+      "loss": 0.0682,
+      "step": 674
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.0002545374449339207,
+      "loss": 0.2414,
+      "step": 675
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00025427312775330393,
+      "loss": 0.2495,
+      "step": 676
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0002540088105726872,
+      "loss": 0.3136,
+      "step": 677
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00025374449339207045,
+      "loss": 0.1107,
+      "step": 678
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.00025348017621145373,
+      "loss": 0.3332,
+      "step": 679
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.000253215859030837,
+      "loss": 0.0433,
+      "step": 680
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00025295154185022025,
+      "loss": 0.1173,
+      "step": 681
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.0002526872246696035,
+      "loss": 0.2195,
+      "step": 682
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00025242290748898676,
+      "loss": 0.2436,
+      "step": 683
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00025215859030837005,
+      "loss": 0.1624,
+      "step": 684
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0002518942731277533,
+      "loss": 0.0986,
+      "step": 685
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00025162995594713657,
+      "loss": 0.1562,
+      "step": 686
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.0002513656387665198,
+      "loss": 0.2453,
+      "step": 687
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00025110132158590303,
+      "loss": 0.1245,
+      "step": 688
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0002508370044052863,
+      "loss": 0.2067,
+      "step": 689
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.0002505726872246696,
+      "loss": 0.3202,
+      "step": 690
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00025030837004405283,
+      "loss": 0.1954,
+      "step": 691
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0002500440528634361,
+      "loss": 0.1485,
+      "step": 692
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00024977973568281935,
+      "loss": 0.1808,
+      "step": 693
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00024951541850220263,
+      "loss": 0.1978,
+      "step": 694
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00024925110132158586,
+      "loss": 0.2984,
+      "step": 695
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00024898678414096915,
+      "loss": 0.2245,
+      "step": 696
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0002487224669603524,
+      "loss": 0.2584,
+      "step": 697
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.00024845814977973566,
+      "loss": 0.3164,
+      "step": 698
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00024819383259911895,
+      "loss": 0.0461,
+      "step": 699
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.0002479295154185022,
+      "loss": 0.2251,
+      "step": 700
+    },
+    {
+      "epoch": 6.42,
+      "eval_loss": 0.4514606297016144,
+      "eval_runtime": 20.6661,
+      "eval_samples_per_second": 79.696,
+      "eval_steps_per_second": 1.258,
+      "eval_wer": 0.46532529874374423,
+      "step": 700
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00024766519823788547,
+      "loss": 0.2806,
+      "step": 701
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.0002474008810572687,
+      "loss": 0.2108,
+      "step": 702
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00024713656387665193,
+      "loss": 0.2462,
+      "step": 703
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.0002468722466960352,
+      "loss": 0.1329,
+      "step": 704
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0002466079295154185,
+      "loss": 0.115,
+      "step": 705
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00024634361233480173,
+      "loss": 0.1838,
+      "step": 706
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.000246079295154185,
+      "loss": 0.1293,
+      "step": 707
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00024581497797356825,
+      "loss": 0.1638,
+      "step": 708
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00024555066079295153,
+      "loss": 0.2114,
+      "step": 709
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00024528634361233476,
+      "loss": 0.2921,
+      "step": 710
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00024502202643171805,
+      "loss": 0.1246,
+      "step": 711
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0002447577092511013,
+      "loss": 0.1765,
+      "step": 712
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00024449339207048456,
+      "loss": 0.136,
+      "step": 713
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00024422907488986785,
+      "loss": 0.3438,
+      "step": 714
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00024396475770925108,
+      "loss": 0.297,
+      "step": 715
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00024370044052863436,
+      "loss": 0.3483,
+      "step": 716
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0002434361233480176,
+      "loss": 0.2205,
+      "step": 717
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00024317180616740085,
+      "loss": 0.2058,
+      "step": 718
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0002429074889867841,
+      "loss": 0.1514,
+      "step": 719
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00024264317180616737,
+      "loss": 0.2167,
+      "step": 720
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00024237885462555066,
+      "loss": 0.3287,
+      "step": 721
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00024211453744493391,
+      "loss": 0.1381,
+      "step": 722
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00024185022026431717,
+      "loss": 0.1109,
+      "step": 723
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.0002415859030837004,
+      "loss": 0.3767,
+      "step": 724
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00024132158590308366,
+      "loss": 0.1166,
+      "step": 725
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00024105726872246695,
+      "loss": 0.2113,
+      "step": 726
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.0002407929515418502,
+      "loss": 0.5353,
+      "step": 727
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00024052863436123346,
+      "loss": 0.2502,
+      "step": 728
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00024026431718061672,
+      "loss": 0.1811,
+      "step": 729
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.3299,
+      "step": 730
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.0002397356828193832,
+      "loss": 0.2254,
+      "step": 731
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0002394713656387665,
+      "loss": 0.2988,
+      "step": 732
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00023920704845814975,
+      "loss": 0.4192,
+      "step": 733
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.000238942731277533,
+      "loss": 0.1991,
+      "step": 734
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00023867841409691627,
+      "loss": 0.0516,
+      "step": 735
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00023841409691629955,
+      "loss": 0.1398,
+      "step": 736
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.0002381497797356828,
+      "loss": 0.1648,
+      "step": 737
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00023788546255506604,
+      "loss": 0.1819,
+      "step": 738
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0002376211453744493,
+      "loss": 0.1757,
+      "step": 739
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00023735682819383256,
+      "loss": 0.1051,
+      "step": 740
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00023709251101321585,
+      "loss": 0.1857,
+      "step": 741
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0002368281938325991,
+      "loss": 0.1663,
+      "step": 742
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00023656387665198236,
+      "loss": 0.2691,
+      "step": 743
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00023629955947136562,
+      "loss": 0.3249,
+      "step": 744
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0002360352422907489,
+      "loss": 0.1311,
+      "step": 745
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00023577092511013214,
+      "loss": 0.2204,
+      "step": 746
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.0002355066079295154,
+      "loss": 0.0752,
+      "step": 747
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00023524229074889865,
+      "loss": 0.3081,
+      "step": 748
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0002349779735682819,
+      "loss": 0.1372,
+      "step": 749
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0002347136563876652,
+      "loss": 0.2608,
+      "step": 750
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00023444933920704845,
+      "loss": 0.2186,
+      "step": 751
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0002341850220264317,
+      "loss": 0.1016,
+      "step": 752
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00023392070484581494,
+      "loss": 0.2295,
+      "step": 753
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0002336563876651982,
+      "loss": 0.7775,
+      "step": 754
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00023339207048458146,
+      "loss": 0.2247,
+      "step": 755
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00023312775330396474,
+      "loss": 0.0718,
+      "step": 756
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.000232863436123348,
+      "loss": 0.1169,
+      "step": 757
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00023259911894273126,
+      "loss": 0.1307,
+      "step": 758
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00023233480176211452,
+      "loss": 0.1339,
+      "step": 759
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00023207048458149775,
+      "loss": 0.3346,
+      "step": 760
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00023180616740088104,
+      "loss": 0.2375,
+      "step": 761
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0002315418502202643,
+      "loss": 0.28,
+      "step": 762
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00023127753303964755,
+      "loss": 0.7263,
+      "step": 763
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0002310132158590308,
+      "loss": 0.082,
+      "step": 764
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.0002307488986784141,
+      "loss": 0.1817,
+      "step": 765
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00023048458149779735,
+      "loss": 0.1222,
+      "step": 766
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00023022026431718058,
+      "loss": 0.0708,
+      "step": 767
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00022995594713656384,
+      "loss": 0.1115,
+      "step": 768
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0002296916299559471,
+      "loss": 0.2776,
+      "step": 769
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00022942731277533039,
+      "loss": 0.1938,
+      "step": 770
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00022916299559471364,
+      "loss": 0.1434,
+      "step": 771
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0002288986784140969,
+      "loss": 0.1644,
+      "step": 772
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00022863436123348016,
+      "loss": 0.2931,
+      "step": 773
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00022837004405286345,
+      "loss": 0.3356,
+      "step": 774
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00022810572687224668,
+      "loss": 0.0805,
+      "step": 775
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00022784140969162993,
+      "loss": 0.1871,
+      "step": 776
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0002275770925110132,
+      "loss": 0.1044,
+      "step": 777
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00022731277533039645,
+      "loss": 0.1177,
+      "step": 778
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.0002270484581497797,
+      "loss": 0.0673,
+      "step": 779
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.000226784140969163,
+      "loss": 0.1424,
+      "step": 780
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00022651982378854625,
+      "loss": 0.2642,
+      "step": 781
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00022625550660792948,
+      "loss": 0.1243,
+      "step": 782
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00022599118942731274,
+      "loss": 0.1178,
+      "step": 783
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.000225726872246696,
+      "loss": 0.0624,
+      "step": 784
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00022546255506607929,
+      "loss": 0.1298,
+      "step": 785
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00022519823788546254,
+      "loss": 0.0901,
+      "step": 786
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0002249339207048458,
+      "loss": 0.0384,
+      "step": 787
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00022466960352422906,
+      "loss": 0.4056,
+      "step": 788
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.0002244052863436123,
+      "loss": 0.0839,
+      "step": 789
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00022414096916299558,
+      "loss": 0.2449,
+      "step": 790
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00022387665198237883,
+      "loss": 0.2228,
+      "step": 791
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0002236123348017621,
+      "loss": 0.2951,
+      "step": 792
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00022334801762114535,
+      "loss": 0.1661,
+      "step": 793
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00022308370044052864,
+      "loss": 0.0737,
+      "step": 794
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.0002228193832599119,
+      "loss": 0.1839,
+      "step": 795
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00022255506607929512,
+      "loss": 0.2112,
+      "step": 796
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00022229074889867838,
+      "loss": 0.1011,
+      "step": 797
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00022202643171806164,
+      "loss": 0.1505,
+      "step": 798
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0002217621145374449,
+      "loss": 0.1657,
+      "step": 799
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00022149779735682818,
+      "loss": 0.2484,
+      "step": 800
+    },
+    {
+      "epoch": 7.34,
+      "eval_loss": 0.42423081398010254,
+      "eval_runtime": 20.65,
+      "eval_samples_per_second": 79.758,
+      "eval_steps_per_second": 1.259,
+      "eval_wer": 0.43080379940761926,
+      "step": 800
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00022123348017621144,
+      "loss": 0.0301,
+      "step": 801
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0002209691629955947,
+      "loss": 0.1522,
+      "step": 802
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00022070484581497796,
+      "loss": 0.1847,
+      "step": 803
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0002204405286343612,
+      "loss": 0.1052,
+      "step": 804
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00022017621145374448,
+      "loss": 0.1157,
+      "step": 805
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00021991189427312773,
+      "loss": 0.1991,
+      "step": 806
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.000219647577092511,
+      "loss": 0.0215,
+      "step": 807
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00021938325991189425,
+      "loss": 0.2818,
+      "step": 808
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00021911894273127753,
+      "loss": 0.4506,
+      "step": 809
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0002188546255506608,
+      "loss": 0.0946,
+      "step": 810
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00021859030837004402,
+      "loss": 0.2227,
+      "step": 811
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00021832599118942728,
+      "loss": 0.0904,
+      "step": 812
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00021806167400881054,
+      "loss": 0.0721,
+      "step": 813
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00021779735682819383,
+      "loss": 0.0401,
+      "step": 814
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00021753303964757708,
+      "loss": 0.2042,
+      "step": 815
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00021726872246696034,
+      "loss": 0.1771,
+      "step": 816
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0002170044052863436,
+      "loss": 0.2348,
+      "step": 817
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00021674008810572683,
+      "loss": 0.1744,
+      "step": 818
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00021647577092511012,
+      "loss": 0.3288,
+      "step": 819
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00021621145374449337,
+      "loss": 0.0313,
+      "step": 820
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00021594713656387663,
+      "loss": 0.1395,
+      "step": 821
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.0002156828193832599,
+      "loss": 0.1808,
+      "step": 822
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00021541850220264315,
+      "loss": 0.1466,
+      "step": 823
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.00021515418502202643,
+      "loss": 0.2761,
+      "step": 824
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00021488986784140967,
+      "loss": 0.1311,
+      "step": 825
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00021462555066079292,
+      "loss": 0.0924,
+      "step": 826
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00021436123348017618,
+      "loss": 0.2073,
+      "step": 827
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00021409691629955944,
+      "loss": 0.0939,
+      "step": 828
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00021383259911894272,
+      "loss": 0.2884,
+      "step": 829
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00021356828193832598,
+      "loss": 0.2573,
+      "step": 830
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00021330396475770924,
+      "loss": 0.3691,
+      "step": 831
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.00021303964757709247,
+      "loss": 0.1369,
+      "step": 832
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00021277533039647573,
+      "loss": 0.2572,
+      "step": 833
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00021251101321585902,
+      "loss": 0.2259,
+      "step": 834
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00021224669603524227,
+      "loss": 0.1174,
+      "step": 835
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00021198237885462553,
+      "loss": 0.1251,
+      "step": 836
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.0002117180616740088,
+      "loss": 0.1277,
+      "step": 837
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00021145374449339208,
+      "loss": 0.1955,
+      "step": 838
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00021118942731277533,
+      "loss": 0.1053,
+      "step": 839
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00021092511013215856,
+      "loss": 0.0803,
+      "step": 840
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00021066079295154182,
+      "loss": 0.0584,
+      "step": 841
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00021039647577092508,
+      "loss": 0.0735,
+      "step": 842
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00021013215859030837,
+      "loss": 0.067,
+      "step": 843
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00020986784140969162,
+      "loss": 0.179,
+      "step": 844
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00020960352422907488,
+      "loss": 0.0502,
+      "step": 845
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00020933920704845814,
+      "loss": 0.0921,
+      "step": 846
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00020907488986784137,
+      "loss": 0.0663,
+      "step": 847
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.00020881057268722463,
+      "loss": 0.1086,
+      "step": 848
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00020854625550660791,
+      "loss": 0.1018,
+      "step": 849
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00020828193832599117,
+      "loss": 0.116,
+      "step": 850
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00020801762114537443,
+      "loss": 0.102,
+      "step": 851
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0002077533039647577,
+      "loss": 0.1418,
+      "step": 852
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00020748898678414097,
+      "loss": 0.1389,
+      "step": 853
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.0002072246696035242,
+      "loss": 0.0987,
+      "step": 854
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00020696035242290746,
+      "loss": 0.1255,
+      "step": 855
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00020669603524229072,
+      "loss": 0.2231,
+      "step": 856
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00020643171806167398,
+      "loss": 0.0419,
+      "step": 857
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00020616740088105727,
+      "loss": 0.1175,
+      "step": 858
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00020590308370044052,
+      "loss": 0.3016,
+      "step": 859
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00020563876651982378,
+      "loss": 0.2679,
+      "step": 860
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.000205374449339207,
+      "loss": 0.1522,
+      "step": 861
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00020511013215859027,
+      "loss": 0.2269,
+      "step": 862
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00020484581497797356,
+      "loss": 0.1744,
+      "step": 863
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00020458149779735681,
+      "loss": 0.0432,
+      "step": 864
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00020431718061674007,
+      "loss": 0.1846,
+      "step": 865
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00020405286343612333,
+      "loss": 0.1334,
+      "step": 866
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.0002037885462555066,
+      "loss": 0.1163,
+      "step": 867
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00020352422907488987,
+      "loss": 0.0384,
+      "step": 868
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0002032599118942731,
+      "loss": 0.2286,
+      "step": 869
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00020299559471365636,
+      "loss": 0.2996,
+      "step": 870
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00020273127753303962,
+      "loss": 0.1687,
+      "step": 871
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00020246696035242288,
+      "loss": 0.096,
+      "step": 872
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00020220264317180616,
+      "loss": 0.1478,
+      "step": 873
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00020193832599118942,
+      "loss": 0.0517,
+      "step": 874
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00020167400881057268,
+      "loss": 0.0563,
+      "step": 875
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0002014096916299559,
+      "loss": 0.1561,
+      "step": 876
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00020114537444933917,
+      "loss": 0.2611,
+      "step": 877
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00020088105726872246,
+      "loss": 0.0455,
+      "step": 878
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0002006167400881057,
+      "loss": 0.0628,
+      "step": 879
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00020035242290748897,
+      "loss": 0.0533,
+      "step": 880
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00020008810572687223,
+      "loss": 0.1227,
+      "step": 881
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00019982378854625552,
+      "loss": 0.0321,
+      "step": 882
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00019955947136563875,
+      "loss": 0.076,
+      "step": 883
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.000199295154185022,
+      "loss": 0.1473,
+      "step": 884
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00019903083700440526,
+      "loss": 0.0247,
+      "step": 885
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00019876651982378852,
+      "loss": 0.1521,
+      "step": 886
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0001985022026431718,
+      "loss": 0.0484,
+      "step": 887
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00019823788546255506,
+      "loss": 0.1446,
+      "step": 888
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00019797356828193832,
+      "loss": 0.1958,
+      "step": 889
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00019770925110132155,
+      "loss": 0.2739,
+      "step": 890
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.0001974449339207048,
+      "loss": 0.0719,
+      "step": 891
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00019718061674008807,
+      "loss": 0.0611,
+      "step": 892
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00019691629955947135,
+      "loss": 0.2907,
+      "step": 893
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0001966519823788546,
+      "loss": 0.1619,
+      "step": 894
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00019638766519823787,
+      "loss": 0.172,
+      "step": 895
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00019612334801762113,
+      "loss": 0.1681,
+      "step": 896
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00019585903083700436,
+      "loss": 0.2517,
+      "step": 897
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00019559471365638765,
+      "loss": 0.1715,
+      "step": 898
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.0001953303964757709,
+      "loss": 0.0867,
+      "step": 899
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00019506607929515416,
+      "loss": 0.0662,
+      "step": 900
+    },
+    {
+      "epoch": 8.26,
+      "eval_loss": 0.4253427982330322,
+      "eval_runtime": 20.7694,
+      "eval_samples_per_second": 79.299,
+      "eval_steps_per_second": 1.252,
+      "eval_wer": 0.4199775303850475,
+      "step": 900
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.00019480176211453742,
+      "loss": 0.1637,
+      "step": 901
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0001945374449339207,
+      "loss": 0.2237,
+      "step": 902
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00019427312775330396,
+      "loss": 0.3601,
+      "step": 903
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00019400881057268722,
+      "loss": 0.0801,
+      "step": 904
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00019374449339207045,
+      "loss": 0.1015,
+      "step": 905
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.0001934801762114537,
+      "loss": 0.2919,
+      "step": 906
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.000193215859030837,
+      "loss": 0.0813,
+      "step": 907
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00019295154185022025,
+      "loss": 0.135,
+      "step": 908
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0001926872246696035,
+      "loss": 0.2061,
+      "step": 909
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00019242290748898677,
+      "loss": 0.0547,
+      "step": 910
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00019215859030837003,
+      "loss": 0.1612,
+      "step": 911
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00019189427312775329,
+      "loss": 0.1312,
+      "step": 912
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00019162995594713654,
+      "loss": 0.1269,
+      "step": 913
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0001913656387665198,
+      "loss": 0.1227,
+      "step": 914
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00019110132158590306,
+      "loss": 0.2029,
+      "step": 915
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00019083700440528632,
+      "loss": 0.2047,
+      "step": 916
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0001905726872246696,
+      "loss": 0.1262,
+      "step": 917
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00019030837004405286,
+      "loss": 0.196,
+      "step": 918
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001900440528634361,
+      "loss": 0.1668,
+      "step": 919
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00018977973568281935,
+      "loss": 0.2997,
+      "step": 920
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.0001895154185022026,
+      "loss": 0.1234,
+      "step": 921
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0001892511013215859,
+      "loss": 0.1882,
+      "step": 922
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00018898678414096915,
+      "loss": 0.1301,
+      "step": 923
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0001887224669603524,
+      "loss": 0.0419,
+      "step": 924
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00018845814977973567,
+      "loss": 0.1122,
+      "step": 925
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0001881938325991189,
+      "loss": 0.164,
+      "step": 926
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00018792951541850219,
+      "loss": 0.1365,
+      "step": 927
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00018766519823788544,
+      "loss": 0.0273,
+      "step": 928
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.0001874008810572687,
+      "loss": 0.0772,
+      "step": 929
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00018713656387665196,
+      "loss": 0.2237,
+      "step": 930
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.00018687224669603525,
+      "loss": 0.1092,
+      "step": 931
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0001866079295154185,
+      "loss": 0.1982,
+      "step": 932
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00018634361233480176,
+      "loss": 0.0967,
+      "step": 933
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.000186079295154185,
+      "loss": 0.0681,
+      "step": 934
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00018581497797356825,
+      "loss": 0.0122,
+      "step": 935
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.0001855506607929515,
+      "loss": 0.0698,
+      "step": 936
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0001852863436123348,
+      "loss": 0.1847,
+      "step": 937
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00018502202643171805,
+      "loss": 0.1181,
+      "step": 938
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0001847577092511013,
+      "loss": 0.1836,
+      "step": 939
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00018449339207048457,
+      "loss": 0.0655,
+      "step": 940
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.0001842290748898678,
+      "loss": 0.0121,
+      "step": 941
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.00018396475770925108,
+      "loss": 0.1478,
+      "step": 942
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00018370044052863434,
+      "loss": 0.2164,
+      "step": 943
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.0001834361233480176,
+      "loss": 0.0181,
+      "step": 944
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00018317180616740086,
+      "loss": 0.1777,
+      "step": 945
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00018290748898678414,
+      "loss": 0.0739,
+      "step": 946
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0001826431718061674,
+      "loss": 0.2477,
+      "step": 947
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00018237885462555063,
+      "loss": 0.1838,
+      "step": 948
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0001821145374449339,
+      "loss": 0.1223,
+      "step": 949
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00018185022026431715,
+      "loss": 0.1231,
+      "step": 950
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00018158590308370044,
+      "loss": 0.1147,
+      "step": 951
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.0001813215859030837,
+      "loss": 0.1009,
+      "step": 952
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00018105726872246695,
+      "loss": 0.0984,
+      "step": 953
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0001807929515418502,
+      "loss": 0.0805,
+      "step": 954
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00018052863436123344,
+      "loss": 0.092,
+      "step": 955
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00018026431718061673,
+      "loss": 0.0831,
+      "step": 956
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.1146,
+      "step": 957
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00017973568281938324,
+      "loss": 0.0176,
+      "step": 958
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0001794713656387665,
+      "loss": 0.142,
+      "step": 959
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017920704845814976,
+      "loss": 0.1375,
+      "step": 960
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00017894273127753304,
+      "loss": 0.2318,
+      "step": 961
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0001786784140969163,
+      "loss": 0.1032,
+      "step": 962
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017841409691629953,
+      "loss": 0.1061,
+      "step": 963
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0001781497797356828,
+      "loss": 0.1211,
+      "step": 964
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017788546255506605,
+      "loss": 0.1968,
+      "step": 965
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017762114537444933,
+      "loss": 0.1638,
+      "step": 966
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.0001773568281938326,
+      "loss": 0.1946,
+      "step": 967
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00017709251101321585,
+      "loss": 0.0795,
+      "step": 968
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.0001768281938325991,
+      "loss": 0.1048,
+      "step": 969
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017656387665198234,
+      "loss": 0.2536,
+      "step": 970
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00017629955947136563,
+      "loss": 0.0545,
+      "step": 971
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00017603524229074888,
+      "loss": 0.1176,
+      "step": 972
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017577092511013214,
+      "loss": 0.1689,
+      "step": 973
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0001755066079295154,
+      "loss": 0.0643,
+      "step": 974
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017524229074889869,
+      "loss": 0.1415,
+      "step": 975
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00017497797356828194,
+      "loss": 0.2591,
+      "step": 976
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00017471365638766517,
+      "loss": 0.1184,
+      "step": 977
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00017444933920704843,
+      "loss": 0.0827,
+      "step": 978
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0001741850220264317,
+      "loss": 0.2077,
+      "step": 979
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00017392070484581495,
+      "loss": 0.1109,
+      "step": 980
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00017365638766519823,
+      "loss": 0.0973,
+      "step": 981
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.0001733920704845815,
+      "loss": 0.2408,
+      "step": 982
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00017312775330396475,
+      "loss": 0.1775,
+      "step": 983
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00017286343612334798,
+      "loss": 0.0986,
+      "step": 984
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00017259911894273124,
+      "loss": 0.0497,
+      "step": 985
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00017233480176211452,
+      "loss": 0.1033,
+      "step": 986
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017207048458149778,
+      "loss": 0.0335,
+      "step": 987
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00017180616740088104,
+      "loss": 0.0739,
+      "step": 988
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.0001715418502202643,
+      "loss": 0.0578,
+      "step": 989
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00017127753303964758,
+      "loss": 0.1152,
+      "step": 990
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00017101321585903082,
+      "loss": 0.1202,
+      "step": 991
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00017074889867841407,
+      "loss": 0.0409,
+      "step": 992
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00017048458149779733,
+      "loss": 0.0745,
+      "step": 993
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.0001702202643171806,
+      "loss": 0.0842,
+      "step": 994
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00016995594713656388,
+      "loss": 0.0268,
+      "step": 995
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00016969162995594713,
+      "loss": 0.1557,
+      "step": 996
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.0001694273127753304,
+      "loss": 0.1976,
+      "step": 997
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00016916299559471365,
+      "loss": 0.0835,
+      "step": 998
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00016889867841409688,
+      "loss": 0.0988,
+      "step": 999
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00016863436123348017,
+      "loss": 0.026,
+      "step": 1000
+    },
+    {
+      "epoch": 9.17,
+      "eval_loss": 0.4305838644504547,
+      "eval_runtime": 20.6887,
+      "eval_samples_per_second": 79.609,
+      "eval_steps_per_second": 1.257,
+      "eval_wer": 0.3977121846593811,
+      "step": 1000
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00016837004405286342,
+      "loss": 0.0366,
+      "step": 1001
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.00016810572687224668,
+      "loss": 0.0934,
+      "step": 1002
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00016784140969162994,
+      "loss": 0.1707,
+      "step": 1003
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.0001675770925110132,
+      "loss": 0.2058,
+      "step": 1004
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00016731277533039648,
+      "loss": 0.0756,
+      "step": 1005
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016704845814977971,
+      "loss": 0.1635,
+      "step": 1006
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00016678414096916297,
+      "loss": 0.064,
+      "step": 1007
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00016651982378854623,
+      "loss": 0.1056,
+      "step": 1008
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0001662555066079295,
+      "loss": 0.0921,
+      "step": 1009
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016599118942731277,
+      "loss": 0.1664,
+      "step": 1010
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00016572687224669603,
+      "loss": 0.06,
+      "step": 1011
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0001654625550660793,
+      "loss": 0.3552,
+      "step": 1012
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00016519823788546252,
+      "loss": 0.0468,
+      "step": 1013
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016493392070484578,
+      "loss": 0.0764,
+      "step": 1014
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016466960352422907,
+      "loss": 0.0278,
+      "step": 1015
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016440528634361232,
+      "loss": 0.2475,
+      "step": 1016
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00016414096916299558,
+      "loss": 0.0884,
+      "step": 1017
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016387665198237884,
+      "loss": 0.0725,
+      "step": 1018
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016361233480176212,
+      "loss": 0.0853,
+      "step": 1019
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.00016334801762114536,
+      "loss": 0.0224,
+      "step": 1020
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016308370044052861,
+      "loss": 0.0818,
+      "step": 1021
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00016281938325991187,
+      "loss": 0.0638,
+      "step": 1022
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016255506607929513,
+      "loss": 0.1895,
+      "step": 1023
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016229074889867842,
+      "loss": 0.1989,
+      "step": 1024
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016202643171806167,
+      "loss": 0.0489,
+      "step": 1025
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00016176211453744493,
+      "loss": 0.0539,
+      "step": 1026
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0001614977973568282,
+      "loss": 0.1095,
+      "step": 1027
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016123348017621142,
+      "loss": 0.1311,
+      "step": 1028
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00016096916299559468,
+      "loss": 0.2165,
+      "step": 1029
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016070484581497796,
+      "loss": 0.2518,
+      "step": 1030
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016044052863436122,
+      "loss": 0.194,
+      "step": 1031
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00016017621145374448,
+      "loss": 0.0433,
+      "step": 1032
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00015991189427312774,
+      "loss": 0.1345,
+      "step": 1033
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.00015964757709251102,
+      "loss": 0.0617,
+      "step": 1034
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00015938325991189425,
+      "loss": 0.0166,
+      "step": 1035
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.0001591189427312775,
+      "loss": 0.122,
+      "step": 1036
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00015885462555066077,
+      "loss": 0.2003,
+      "step": 1037
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.00015859030837004403,
+      "loss": 0.0104,
+      "step": 1038
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00015832599118942731,
+      "loss": 0.0416,
+      "step": 1039
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00015806167400881057,
+      "loss": 0.0921,
+      "step": 1040
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00015779735682819383,
+      "loss": 0.033,
+      "step": 1041
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.00015753303964757706,
+      "loss": 0.0738,
+      "step": 1042
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00015726872246696032,
+      "loss": 0.1938,
+      "step": 1043
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.0001570044052863436,
+      "loss": 0.1505,
+      "step": 1044
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00015674008810572686,
+      "loss": 0.0179,
+      "step": 1045
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00015647577092511012,
+      "loss": 0.0985,
+      "step": 1046
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00015621145374449338,
+      "loss": 0.069,
+      "step": 1047
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00015594713656387664,
+      "loss": 0.0911,
+      "step": 1048
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00015568281938325987,
+      "loss": 0.1279,
+      "step": 1049
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00015541850220264315,
+      "loss": 0.0434,
+      "step": 1050
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0001551541850220264,
+      "loss": 0.0661,
+      "step": 1051
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00015488986784140967,
+      "loss": 0.0986,
+      "step": 1052
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00015462555066079293,
+      "loss": 0.0177,
+      "step": 1053
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00015436123348017621,
+      "loss": 0.1361,
+      "step": 1054
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00015409691629955947,
+      "loss": 0.2669,
+      "step": 1055
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00015383259911894273,
+      "loss": 0.2784,
+      "step": 1056
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00015356828193832596,
+      "loss": 0.0447,
+      "step": 1057
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00015330396475770922,
+      "loss": 0.1157,
+      "step": 1058
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0001530396475770925,
+      "loss": 0.07,
+      "step": 1059
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00015277533039647576,
+      "loss": 0.1426,
+      "step": 1060
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00015251101321585902,
+      "loss": 0.0845,
+      "step": 1061
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00015224669603524228,
+      "loss": 0.1183,
+      "step": 1062
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.00015198237885462556,
+      "loss": 0.0618,
+      "step": 1063
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.0001517180616740088,
+      "loss": 0.0363,
+      "step": 1064
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00015145374449339205,
+      "loss": 0.1415,
+      "step": 1065
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.0001511894273127753,
+      "loss": 0.298,
+      "step": 1066
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.00015092511013215857,
+      "loss": 0.2451,
+      "step": 1067
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00015066079295154186,
+      "loss": 0.0523,
+      "step": 1068
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0001503964757709251,
+      "loss": 0.0771,
+      "step": 1069
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00015013215859030837,
+      "loss": 0.1646,
+      "step": 1070
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00014986784140969163,
+      "loss": 0.1512,
+      "step": 1071
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0001496035242290749,
+      "loss": 0.2151,
+      "step": 1072
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00014933920704845812,
+      "loss": 0.0516,
+      "step": 1073
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0001490748898678414,
+      "loss": 0.1517,
+      "step": 1074
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00014881057268722466,
+      "loss": 0.0105,
+      "step": 1075
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00014854625550660792,
+      "loss": 0.2626,
+      "step": 1076
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00014828193832599118,
+      "loss": 0.0807,
+      "step": 1077
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00014801762114537444,
+      "loss": 0.0719,
+      "step": 1078
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0001477533039647577,
+      "loss": 0.0692,
+      "step": 1079
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00014748898678414095,
+      "loss": 0.2708,
+      "step": 1080
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.0001472246696035242,
+      "loss": 0.1532,
+      "step": 1081
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00014696035242290747,
+      "loss": 0.0944,
+      "step": 1082
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00014669603524229075,
+      "loss": 0.0627,
+      "step": 1083
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00014643171806167399,
+      "loss": 0.0111,
+      "step": 1084
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00014616740088105724,
+      "loss": 0.1838,
+      "step": 1085
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00014590308370044053,
+      "loss": 0.124,
+      "step": 1086
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00014563876651982376,
+      "loss": 0.3301,
+      "step": 1087
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00014537444933920705,
+      "loss": 0.0831,
+      "step": 1088
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.0001451101321585903,
+      "loss": 0.0239,
+      "step": 1089
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00014484581497797356,
+      "loss": 0.0986,
+      "step": 1090
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00014458149779735682,
+      "loss": 0.1712,
+      "step": 1091
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00014431718061674008,
+      "loss": 0.0778,
+      "step": 1092
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00014405286343612334,
+      "loss": 0.0694,
+      "step": 1093
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0001437885462555066,
+      "loss": 0.0306,
+      "step": 1094
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00014352422907488985,
+      "loss": 0.0767,
+      "step": 1095
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0001432599118942731,
+      "loss": 0.0196,
+      "step": 1096
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00014299559471365637,
+      "loss": 0.1471,
+      "step": 1097
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00014273127753303963,
+      "loss": 0.1974,
+      "step": 1098
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00014246696035242288,
+      "loss": 0.0335,
+      "step": 1099
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00014220264317180617,
+      "loss": 0.1612,
+      "step": 1100
+    },
+    {
+      "epoch": 10.09,
+      "eval_loss": 0.41547149419784546,
+      "eval_runtime": 20.6636,
+      "eval_samples_per_second": 79.705,
+      "eval_steps_per_second": 1.258,
+      "eval_wer": 0.3940353385762435,
+      "step": 1100
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00014193832599118943,
+      "loss": 0.0846,
+      "step": 1101
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00014167400881057266,
+      "loss": 0.113,
+      "step": 1102
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00014140969162995594,
+      "loss": 0.0441,
+      "step": 1103
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0001411453744493392,
+      "loss": 0.0779,
+      "step": 1104
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00014088105726872243,
+      "loss": 0.1555,
+      "step": 1105
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00014061674008810572,
+      "loss": 0.134,
+      "step": 1106
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00014035242290748898,
+      "loss": 0.1617,
+      "step": 1107
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00014008810572687224,
+      "loss": 0.2659,
+      "step": 1108
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0001398237885462555,
+      "loss": 0.0412,
+      "step": 1109
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.00013955947136563875,
+      "loss": 0.0203,
+      "step": 1110
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.000139295154185022,
+      "loss": 0.0422,
+      "step": 1111
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0001390308370044053,
+      "loss": 0.1288,
+      "step": 1112
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00013876651982378853,
+      "loss": 0.0405,
+      "step": 1113
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00013850220264317178,
+      "loss": 0.0212,
+      "step": 1114
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00013823788546255507,
+      "loss": 0.1987,
+      "step": 1115
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0001379735682819383,
+      "loss": 0.1412,
+      "step": 1116
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00013770925110132156,
+      "loss": 0.0884,
+      "step": 1117
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00013744493392070484,
+      "loss": 0.034,
+      "step": 1118
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0001371806167400881,
+      "loss": 0.1052,
+      "step": 1119
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00013691629955947136,
+      "loss": 0.0381,
+      "step": 1120
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00013665198237885462,
+      "loss": 0.1033,
+      "step": 1121
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.00013638766519823788,
+      "loss": 0.0502,
+      "step": 1122
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00013612334801762113,
+      "loss": 0.0298,
+      "step": 1123
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.0001358590308370044,
+      "loss": 0.0509,
+      "step": 1124
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00013559471365638765,
+      "loss": 0.019,
+      "step": 1125
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0001353303964757709,
+      "loss": 0.1202,
+      "step": 1126
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00013506607929515417,
+      "loss": 0.0374,
+      "step": 1127
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00013480176211453743,
+      "loss": 0.0664,
+      "step": 1128
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00013453744493392068,
+      "loss": 0.0343,
+      "step": 1129
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00013427312775330397,
+      "loss": 0.2851,
+      "step": 1130
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.0001340088105726872,
+      "loss": 0.0988,
+      "step": 1131
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00013374449339207048,
+      "loss": 0.1847,
+      "step": 1132
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00013348017621145374,
+      "loss": 0.169,
+      "step": 1133
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00013321585903083697,
+      "loss": 0.1133,
+      "step": 1134
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00013295154185022026,
+      "loss": 0.1928,
+      "step": 1135
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00013268722466960352,
+      "loss": 0.0393,
+      "step": 1136
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00013242290748898678,
+      "loss": 0.0797,
+      "step": 1137
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.00013215859030837003,
+      "loss": 0.0458,
+      "step": 1138
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0001318942731277533,
+      "loss": 0.1561,
+      "step": 1139
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00013162995594713655,
+      "loss": 0.2915,
+      "step": 1140
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0001313656387665198,
+      "loss": 0.0497,
+      "step": 1141
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00013110132158590307,
+      "loss": 0.0186,
+      "step": 1142
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00013083700440528632,
+      "loss": 0.0383,
+      "step": 1143
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0001305726872246696,
+      "loss": 0.1133,
+      "step": 1144
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00013030837004405284,
+      "loss": 0.0824,
+      "step": 1145
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0001300440528634361,
+      "loss": 0.041,
+      "step": 1146
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00012977973568281938,
+      "loss": 0.0726,
+      "step": 1147
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00012951541850220264,
+      "loss": 0.1824,
+      "step": 1148
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0001292511013215859,
+      "loss": 0.179,
+      "step": 1149
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00012898678414096916,
+      "loss": 0.1867,
+      "step": 1150
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00012872246696035242,
+      "loss": 0.0349,
+      "step": 1151
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00012845814977973567,
+      "loss": 0.021,
+      "step": 1152
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00012819383259911893,
+      "loss": 0.0297,
+      "step": 1153
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0001279295154185022,
+      "loss": 0.0401,
+      "step": 1154
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00012766519823788545,
+      "loss": 0.0526,
+      "step": 1155
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0001274008810572687,
+      "loss": 0.05,
+      "step": 1156
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00012713656387665197,
+      "loss": 0.2185,
+      "step": 1157
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00012687224669603522,
+      "loss": 0.2737,
+      "step": 1158
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.0001266079295154185,
+      "loss": 0.0264,
+      "step": 1159
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00012634361233480174,
+      "loss": 0.0145,
+      "step": 1160
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00012607929515418503,
+      "loss": 0.0146,
+      "step": 1161
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00012581497797356828,
+      "loss": 0.018,
+      "step": 1162
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.00012555066079295151,
+      "loss": 0.0417,
+      "step": 1163
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.0001252863436123348,
+      "loss": 0.0879,
+      "step": 1164
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00012502202643171806,
+      "loss": 0.4337,
+      "step": 1165
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00012475770925110132,
+      "loss": 0.0612,
+      "step": 1166
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.00012449339207048457,
+      "loss": 0.1251,
+      "step": 1167
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00012422907488986783,
+      "loss": 0.1144,
+      "step": 1168
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.0001239647577092511,
+      "loss": 0.0304,
+      "step": 1169
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00012370044052863435,
+      "loss": 0.045,
+      "step": 1170
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0001234361233480176,
+      "loss": 0.0635,
+      "step": 1171
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00012317180616740086,
+      "loss": 0.0528,
+      "step": 1172
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00012290748898678412,
+      "loss": 0.1042,
+      "step": 1173
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00012264317180616738,
+      "loss": 0.0854,
+      "step": 1174
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00012237885462555064,
+      "loss": 0.1391,
+      "step": 1175
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00012211453744493392,
+      "loss": 0.0317,
+      "step": 1176
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00012185022026431718,
+      "loss": 0.0199,
+      "step": 1177
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00012158590308370043,
+      "loss": 0.0127,
+      "step": 1178
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00012132158590308369,
+      "loss": 0.0877,
+      "step": 1179
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00012105726872246696,
+      "loss": 0.0666,
+      "step": 1180
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0001207929515418502,
+      "loss": 0.0905,
+      "step": 1181
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00012052863436123347,
+      "loss": 0.1055,
+      "step": 1182
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00012026431718061673,
+      "loss": 0.0774,
+      "step": 1183
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 0.2295,
+      "step": 1184
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.00011973568281938325,
+      "loss": 0.0886,
+      "step": 1185
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0001194713656387665,
+      "loss": 0.1329,
+      "step": 1186
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00011920704845814978,
+      "loss": 0.1235,
+      "step": 1187
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00011894273127753302,
+      "loss": 0.1383,
+      "step": 1188
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00011867841409691628,
+      "loss": 0.1825,
+      "step": 1189
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00011841409691629955,
+      "loss": 0.1815,
+      "step": 1190
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00011814977973568281,
+      "loss": 0.0338,
+      "step": 1191
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00011788546255506607,
+      "loss": 0.1743,
+      "step": 1192
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00011762114537444933,
+      "loss": 0.0463,
+      "step": 1193
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0001173568281938326,
+      "loss": 0.2198,
+      "step": 1194
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00011709251101321586,
+      "loss": 0.0623,
+      "step": 1195
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.0001168281938325991,
+      "loss": 0.4023,
+      "step": 1196
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00011656387665198237,
+      "loss": 0.032,
+      "step": 1197
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00011629955947136563,
+      "loss": 0.1228,
+      "step": 1198
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00011603524229074888,
+      "loss": 0.4277,
+      "step": 1199
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00011577092511013215,
+      "loss": 0.0361,
+      "step": 1200
+    },
+    {
+      "epoch": 11.01,
+      "eval_loss": 0.4063771069049835,
+      "eval_runtime": 20.6338,
+      "eval_samples_per_second": 79.821,
+      "eval_steps_per_second": 1.26,
+      "eval_wer": 0.38137064651210295,
+      "step": 1200
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0001155066079295154,
+      "loss": 0.098,
+      "step": 1201
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00011524229074889868,
+      "loss": 0.0977,
+      "step": 1202
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00011497797356828192,
+      "loss": 0.172,
+      "step": 1203
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00011471365638766519,
+      "loss": 0.0306,
+      "step": 1204
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00011444933920704845,
+      "loss": 0.0912,
+      "step": 1205
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00011418502202643172,
+      "loss": 0.0066,
+      "step": 1206
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00011392070484581497,
+      "loss": 0.0043,
+      "step": 1207
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00011365638766519823,
+      "loss": 0.0598,
+      "step": 1208
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.0001133920704845815,
+      "loss": 0.0382,
+      "step": 1209
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00011312775330396474,
+      "loss": 0.0981,
+      "step": 1210
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.000112863436123348,
+      "loss": 0.0896,
+      "step": 1211
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00011259911894273127,
+      "loss": 0.0155,
+      "step": 1212
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.00011233480176211453,
+      "loss": 0.1246,
+      "step": 1213
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00011207048458149779,
+      "loss": 0.1021,
+      "step": 1214
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.00011180616740088105,
+      "loss": 0.0923,
+      "step": 1215
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00011154185022026432,
+      "loss": 0.1,
+      "step": 1216
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00011127753303964756,
+      "loss": 0.1227,
+      "step": 1217
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00011101321585903082,
+      "loss": 0.0793,
+      "step": 1218
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00011074889867841409,
+      "loss": 0.0147,
+      "step": 1219
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00011048458149779735,
+      "loss": 0.0674,
+      "step": 1220
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.0001102202643171806,
+      "loss": 0.1031,
+      "step": 1221
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00010995594713656387,
+      "loss": 0.1209,
+      "step": 1222
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00010969162995594712,
+      "loss": 0.0937,
+      "step": 1223
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0001094273127753304,
+      "loss": 0.1815,
+      "step": 1224
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00010916299559471364,
+      "loss": 0.0044,
+      "step": 1225
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00010889867841409691,
+      "loss": 0.0965,
+      "step": 1226
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00010863436123348017,
+      "loss": 0.0338,
+      "step": 1227
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00010837004405286342,
+      "loss": 0.2002,
+      "step": 1228
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00010810572687224669,
+      "loss": 0.0335,
+      "step": 1229
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00010784140969162995,
+      "loss": 0.0613,
+      "step": 1230
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00010757709251101322,
+      "loss": 0.021,
+      "step": 1231
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00010731277533039646,
+      "loss": 0.0231,
+      "step": 1232
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.00010704845814977972,
+      "loss": 0.0403,
+      "step": 1233
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00010678414096916299,
+      "loss": 0.0667,
+      "step": 1234
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00010651982378854624,
+      "loss": 0.12,
+      "step": 1235
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00010625550660792951,
+      "loss": 0.148,
+      "step": 1236
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00010599118942731277,
+      "loss": 0.0038,
+      "step": 1237
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00010572687224669604,
+      "loss": 0.0368,
+      "step": 1238
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00010546255506607928,
+      "loss": 0.0863,
+      "step": 1239
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00010519823788546254,
+      "loss": 0.1584,
+      "step": 1240
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00010493392070484581,
+      "loss": 0.125,
+      "step": 1241
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00010466960352422907,
+      "loss": 0.0085,
+      "step": 1242
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00010440528634361231,
+      "loss": 0.0323,
+      "step": 1243
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00010414096916299559,
+      "loss": 0.0881,
+      "step": 1244
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00010387665198237884,
+      "loss": 0.0232,
+      "step": 1245
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0001036123348017621,
+      "loss": 0.0615,
+      "step": 1246
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00010334801762114536,
+      "loss": 0.0673,
+      "step": 1247
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00010308370044052863,
+      "loss": 0.1671,
+      "step": 1248
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00010281938325991189,
+      "loss": 0.1117,
+      "step": 1249
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00010255506607929514,
+      "loss": 0.1038,
+      "step": 1250
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00010229074889867841,
+      "loss": 0.0688,
+      "step": 1251
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00010202643171806167,
+      "loss": 0.0377,
+      "step": 1252
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00010176211453744494,
+      "loss": 0.1531,
+      "step": 1253
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00010149779735682818,
+      "loss": 0.2099,
+      "step": 1254
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00010123348017621144,
+      "loss": 0.1054,
+      "step": 1255
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00010096916299559471,
+      "loss": 0.1238,
+      "step": 1256
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00010070484581497796,
+      "loss": 0.0528,
+      "step": 1257
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00010044052863436123,
+      "loss": 0.1224,
+      "step": 1258
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00010017621145374449,
+      "loss": 0.0079,
+      "step": 1259
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 9.991189427312776e-05,
+      "loss": 0.168,
+      "step": 1260
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 9.9647577092511e-05,
+      "loss": 0.0762,
+      "step": 1261
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 9.938325991189426e-05,
+      "loss": 0.005,
+      "step": 1262
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 9.911894273127753e-05,
+      "loss": 0.1368,
+      "step": 1263
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 9.885462555066078e-05,
+      "loss": 0.0252,
+      "step": 1264
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 9.859030837004403e-05,
+      "loss": 0.1424,
+      "step": 1265
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 9.83259911894273e-05,
+      "loss": 0.0178,
+      "step": 1266
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 9.806167400881056e-05,
+      "loss": 0.0588,
+      "step": 1267
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 9.779735682819382e-05,
+      "loss": 0.0411,
+      "step": 1268
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 9.753303964757708e-05,
+      "loss": 0.0209,
+      "step": 1269
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 9.726872246696035e-05,
+      "loss": 0.0338,
+      "step": 1270
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 9.700440528634361e-05,
+      "loss": 0.043,
+      "step": 1271
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 9.674008810572686e-05,
+      "loss": 0.2563,
+      "step": 1272
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 9.647577092511013e-05,
+      "loss": 0.1368,
+      "step": 1273
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 9.621145374449339e-05,
+      "loss": 0.2456,
+      "step": 1274
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 9.594713656387664e-05,
+      "loss": 0.1254,
+      "step": 1275
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 9.56828193832599e-05,
+      "loss": 0.0701,
+      "step": 1276
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 9.541850220264316e-05,
+      "loss": 0.2732,
+      "step": 1277
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 9.515418502202643e-05,
+      "loss": 0.0051,
+      "step": 1278
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 9.488986784140968e-05,
+      "loss": 0.09,
+      "step": 1279
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 9.462555066079295e-05,
+      "loss": 0.0433,
+      "step": 1280
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 9.43612334801762e-05,
+      "loss": 0.0772,
+      "step": 1281
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 9.409691629955945e-05,
+      "loss": 0.0325,
+      "step": 1282
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 9.383259911894272e-05,
+      "loss": 0.128,
+      "step": 1283
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 9.356828193832598e-05,
+      "loss": 0.1902,
+      "step": 1284
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 9.330396475770925e-05,
+      "loss": 0.1468,
+      "step": 1285
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 9.30396475770925e-05,
+      "loss": 0.1449,
+      "step": 1286
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 9.277533039647575e-05,
+      "loss": 0.0936,
+      "step": 1287
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 9.251101321585903e-05,
+      "loss": 0.1288,
+      "step": 1288
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 9.224669603524228e-05,
+      "loss": 0.0486,
+      "step": 1289
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 9.198237885462554e-05,
+      "loss": 0.2769,
+      "step": 1290
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 9.17180616740088e-05,
+      "loss": 0.0666,
+      "step": 1291
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 9.145374449339207e-05,
+      "loss": 0.0057,
+      "step": 1292
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 9.118942731277532e-05,
+      "loss": 0.0037,
+      "step": 1293
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 9.092511013215858e-05,
+      "loss": 0.0723,
+      "step": 1294
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 9.066079295154185e-05,
+      "loss": 0.0664,
+      "step": 1295
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 9.03964757709251e-05,
+      "loss": 0.0593,
+      "step": 1296
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 9.013215859030836e-05,
+      "loss": 0.0637,
+      "step": 1297
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 8.986784140969162e-05,
+      "loss": 0.1042,
+      "step": 1298
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 8.960352422907488e-05,
+      "loss": 0.0449,
+      "step": 1299
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 8.933920704845815e-05,
+      "loss": 0.0765,
+      "step": 1300
+    },
+    {
+      "epoch": 11.93,
+      "eval_loss": 0.4122539162635803,
+      "eval_runtime": 20.8663,
+      "eval_samples_per_second": 78.931,
+      "eval_steps_per_second": 1.246,
+      "eval_wer": 0.3790215504034317,
+      "step": 1300
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 8.90748898678414e-05,
+      "loss": 0.037,
+      "step": 1301
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 8.881057268722467e-05,
+      "loss": 0.0692,
+      "step": 1302
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 8.854625550660793e-05,
+      "loss": 0.1068,
+      "step": 1303
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 8.828193832599117e-05,
+      "loss": 0.1471,
+      "step": 1304
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 8.801762114537444e-05,
+      "loss": 0.0839,
+      "step": 1305
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 8.77533039647577e-05,
+      "loss": 0.1384,
+      "step": 1306
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 8.748898678414097e-05,
+      "loss": 0.0119,
+      "step": 1307
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 8.722466960352422e-05,
+      "loss": 0.0395,
+      "step": 1308
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 8.696035242290747e-05,
+      "loss": 0.0878,
+      "step": 1309
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 8.669603524229075e-05,
+      "loss": 0.0417,
+      "step": 1310
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 8.643171806167399e-05,
+      "loss": 0.0702,
+      "step": 1311
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 8.616740088105726e-05,
+      "loss": 0.1372,
+      "step": 1312
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 8.590308370044052e-05,
+      "loss": 0.0868,
+      "step": 1313
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 8.563876651982379e-05,
+      "loss": 0.1699,
+      "step": 1314
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 8.537444933920704e-05,
+      "loss": 0.0346,
+      "step": 1315
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 8.51101321585903e-05,
+      "loss": 0.006,
+      "step": 1316
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 8.484581497797357e-05,
+      "loss": 0.0909,
+      "step": 1317
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 8.458149779735682e-05,
+      "loss": 0.0352,
+      "step": 1318
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 8.431718061674008e-05,
+      "loss": 0.0857,
+      "step": 1319
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 8.405286343612334e-05,
+      "loss": 0.0242,
+      "step": 1320
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 8.37885462555066e-05,
+      "loss": 0.0154,
+      "step": 1321
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 8.352422907488986e-05,
+      "loss": 0.0565,
+      "step": 1322
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 8.325991189427312e-05,
+      "loss": 0.1971,
+      "step": 1323
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 8.299559471365639e-05,
+      "loss": 0.0618,
+      "step": 1324
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 8.273127753303965e-05,
+      "loss": 0.1676,
+      "step": 1325
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 8.246696035242289e-05,
+      "loss": 0.1725,
+      "step": 1326
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 8.220264317180616e-05,
+      "loss": 0.2023,
+      "step": 1327
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 8.193832599118942e-05,
+      "loss": 0.0302,
+      "step": 1328
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 8.167400881057268e-05,
+      "loss": 0.037,
+      "step": 1329
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 8.140969162995594e-05,
+      "loss": 0.0396,
+      "step": 1330
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 8.114537444933921e-05,
+      "loss": 0.0934,
+      "step": 1331
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 8.088105726872247e-05,
+      "loss": 0.0224,
+      "step": 1332
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 8.061674008810571e-05,
+      "loss": 0.1356,
+      "step": 1333
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 8.035242290748898e-05,
+      "loss": 0.0457,
+      "step": 1334
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 8.008810572687224e-05,
+      "loss": 0.0445,
+      "step": 1335
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 7.982378854625551e-05,
+      "loss": 0.0594,
+      "step": 1336
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 7.955947136563876e-05,
+      "loss": 0.0998,
+      "step": 1337
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 7.929515418502201e-05,
+      "loss": 0.258,
+      "step": 1338
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 7.903083700440529e-05,
+      "loss": 0.0352,
+      "step": 1339
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 7.876651982378853e-05,
+      "loss": 0.0766,
+      "step": 1340
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 7.85022026431718e-05,
+      "loss": 0.0328,
+      "step": 1341
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 7.823788546255506e-05,
+      "loss": 0.0662,
+      "step": 1342
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 7.797356828193832e-05,
+      "loss": 0.08,
+      "step": 1343
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 7.770925110132158e-05,
+      "loss": 0.0508,
+      "step": 1344
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 7.744493392070484e-05,
+      "loss": 0.0737,
+      "step": 1345
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 7.718061674008811e-05,
+      "loss": 0.039,
+      "step": 1346
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 7.691629955947137e-05,
+      "loss": 0.0318,
+      "step": 1347
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 7.665198237885461e-05,
+      "loss": 0.066,
+      "step": 1348
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 7.638766519823788e-05,
+      "loss": 0.0291,
+      "step": 1349
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 7.612334801762114e-05,
+      "loss": 0.064,
+      "step": 1350
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 7.58590308370044e-05,
+      "loss": 0.0654,
+      "step": 1351
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 7.559471365638766e-05,
+      "loss": 0.0151,
+      "step": 1352
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 7.533039647577093e-05,
+      "loss": 0.0845,
+      "step": 1353
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 7.506607929515419e-05,
+      "loss": 0.0576,
+      "step": 1354
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 7.480176211453744e-05,
+      "loss": 0.0852,
+      "step": 1355
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 7.45374449339207e-05,
+      "loss": 0.0649,
+      "step": 1356
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 7.427312775330396e-05,
+      "loss": 0.0671,
+      "step": 1357
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 7.400881057268722e-05,
+      "loss": 0.0943,
+      "step": 1358
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 7.374449339207048e-05,
+      "loss": 0.0206,
+      "step": 1359
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 7.348017621145373e-05,
+      "loss": 0.1085,
+      "step": 1360
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 7.321585903083699e-05,
+      "loss": 0.045,
+      "step": 1361
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 7.295154185022026e-05,
+      "loss": 0.1404,
+      "step": 1362
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 7.268722466960352e-05,
+      "loss": 0.0697,
+      "step": 1363
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 7.242290748898678e-05,
+      "loss": 0.0727,
+      "step": 1364
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 7.215859030837004e-05,
+      "loss": 0.0079,
+      "step": 1365
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 7.18942731277533e-05,
+      "loss": 0.0439,
+      "step": 1366
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 7.162995594713656e-05,
+      "loss": 0.0307,
+      "step": 1367
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 7.136563876651981e-05,
+      "loss": 0.0148,
+      "step": 1368
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 7.110132158590308e-05,
+      "loss": 0.1295,
+      "step": 1369
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 7.083700440528633e-05,
+      "loss": 0.1301,
+      "step": 1370
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 7.05726872246696e-05,
+      "loss": 0.0271,
+      "step": 1371
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 7.030837004405286e-05,
+      "loss": 0.0861,
+      "step": 1372
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 7.004405286343612e-05,
+      "loss": 0.0903,
+      "step": 1373
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 6.977973568281938e-05,
+      "loss": 0.1827,
+      "step": 1374
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 6.951541850220265e-05,
+      "loss": 0.1098,
+      "step": 1375
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 6.925110132158589e-05,
+      "loss": 0.0518,
+      "step": 1376
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 6.898678414096915e-05,
+      "loss": 0.0255,
+      "step": 1377
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 6.872246696035242e-05,
+      "loss": 0.0553,
+      "step": 1378
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 6.845814977973568e-05,
+      "loss": 0.0155,
+      "step": 1379
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 6.819383259911894e-05,
+      "loss": 0.0659,
+      "step": 1380
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 6.79295154185022e-05,
+      "loss": 0.0241,
+      "step": 1381
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 6.766519823788545e-05,
+      "loss": 0.091,
+      "step": 1382
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 6.740088105726871e-05,
+      "loss": 0.1952,
+      "step": 1383
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 6.713656387665198e-05,
+      "loss": 0.0116,
+      "step": 1384
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 6.687224669603524e-05,
+      "loss": 0.1266,
+      "step": 1385
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 6.660792951541849e-05,
+      "loss": 0.1285,
+      "step": 1386
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 6.634361233480176e-05,
+      "loss": 0.106,
+      "step": 1387
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 6.607929515418502e-05,
+      "loss": 0.0258,
+      "step": 1388
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 6.581497797356827e-05,
+      "loss": 0.0717,
+      "step": 1389
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 6.555066079295153e-05,
+      "loss": 0.0158,
+      "step": 1390
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 6.52863436123348e-05,
+      "loss": 0.0057,
+      "step": 1391
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 6.502202643171805e-05,
+      "loss": 0.0964,
+      "step": 1392
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 6.475770925110132e-05,
+      "loss": 0.0066,
+      "step": 1393
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 6.449339207048458e-05,
+      "loss": 0.2254,
+      "step": 1394
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 6.422907488986784e-05,
+      "loss": 0.265,
+      "step": 1395
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 6.39647577092511e-05,
+      "loss": 0.0476,
+      "step": 1396
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 6.370044052863435e-05,
+      "loss": 0.0181,
+      "step": 1397
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 6.343612334801761e-05,
+      "loss": 0.0784,
+      "step": 1398
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 6.317180616740087e-05,
+      "loss": 0.0093,
+      "step": 1399
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 6.290748898678414e-05,
+      "loss": 0.022,
+      "step": 1400
+    },
+    {
+      "epoch": 12.84,
+      "eval_loss": 0.3993262052536011,
+      "eval_runtime": 20.6855,
+      "eval_samples_per_second": 79.621,
+      "eval_steps_per_second": 1.257,
+      "eval_wer": 0.3656419160453478,
+      "step": 1400
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 6.26431718061674e-05,
+      "loss": 0.1132,
+      "step": 1401
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 6.237885462555066e-05,
+      "loss": 0.0405,
+      "step": 1402
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 6.211453744493392e-05,
+      "loss": 0.0542,
+      "step": 1403
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 6.185022026431717e-05,
+      "loss": 0.074,
+      "step": 1404
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 6.158590308370043e-05,
+      "loss": 0.0092,
+      "step": 1405
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 6.132158590308369e-05,
+      "loss": 0.046,
+      "step": 1406
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 6.105726872246696e-05,
+      "loss": 0.1066,
+      "step": 1407
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 6.0792951541850214e-05,
+      "loss": 0.1222,
+      "step": 1408
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 6.052863436123348e-05,
+      "loss": 0.0262,
+      "step": 1409
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 6.0264317180616737e-05,
+      "loss": 0.0154,
+      "step": 1410
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 0.0333,
+      "step": 1411
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 5.973568281938325e-05,
+      "loss": 0.0984,
+      "step": 1412
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 5.947136563876651e-05,
+      "loss": 0.0441,
+      "step": 1413
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 5.9207048458149776e-05,
+      "loss": 0.0819,
+      "step": 1414
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 5.8942731277533034e-05,
+      "loss": 0.0035,
+      "step": 1415
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 5.86784140969163e-05,
+      "loss": 0.0966,
+      "step": 1416
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 5.841409691629955e-05,
+      "loss": 0.0501,
+      "step": 1417
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 5.8149779735682815e-05,
+      "loss": 0.0588,
+      "step": 1418
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 5.7885462555066073e-05,
+      "loss": 0.0637,
+      "step": 1419
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 5.762114537444934e-05,
+      "loss": 0.0689,
+      "step": 1420
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 5.7356828193832597e-05,
+      "loss": 0.1554,
+      "step": 1421
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 5.709251101321586e-05,
+      "loss": 0.0498,
+      "step": 1422
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 5.682819383259911e-05,
+      "loss": 0.1015,
+      "step": 1423
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 5.656387665198237e-05,
+      "loss": 0.0721,
+      "step": 1424
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 5.6299559471365636e-05,
+      "loss": 0.1501,
+      "step": 1425
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 5.6035242290748894e-05,
+      "loss": 0.0046,
+      "step": 1426
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 5.577092511013216e-05,
+      "loss": 0.197,
+      "step": 1427
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 5.550660792951541e-05,
+      "loss": 0.1334,
+      "step": 1428
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 5.5242290748898675e-05,
+      "loss": 0.203,
+      "step": 1429
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 5.497797356828193e-05,
+      "loss": 0.0178,
+      "step": 1430
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 5.47136563876652e-05,
+      "loss": 0.0683,
+      "step": 1431
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 5.4449339207048456e-05,
+      "loss": 0.112,
+      "step": 1432
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 5.418502202643171e-05,
+      "loss": 0.0952,
+      "step": 1433
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 5.392070484581497e-05,
+      "loss": 0.0914,
+      "step": 1434
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 5.365638766519823e-05,
+      "loss": 0.1275,
+      "step": 1435
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 5.3392070484581496e-05,
+      "loss": 0.0124,
+      "step": 1436
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 5.3127753303964754e-05,
+      "loss": 0.0498,
+      "step": 1437
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 5.286343612334802e-05,
+      "loss": 0.0736,
+      "step": 1438
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 5.259911894273127e-05,
+      "loss": 0.0186,
+      "step": 1439
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 5.2334801762114535e-05,
+      "loss": 0.0923,
+      "step": 1440
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 5.207048458149779e-05,
+      "loss": 0.083,
+      "step": 1441
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 5.180616740088105e-05,
+      "loss": 0.8404,
+      "step": 1442
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 5.1541850220264316e-05,
+      "loss": 0.0767,
+      "step": 1443
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 5.127753303964757e-05,
+      "loss": 0.0284,
+      "step": 1444
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 5.101321585903083e-05,
+      "loss": 0.0326,
+      "step": 1445
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 5.074889867841409e-05,
+      "loss": 0.1548,
+      "step": 1446
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 5.0484581497797356e-05,
+      "loss": 0.1346,
+      "step": 1447
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 5.0220264317180614e-05,
+      "loss": 0.0294,
+      "step": 1448
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.995594713656388e-05,
+      "loss": 0.0173,
+      "step": 1449
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 4.969162995594713e-05,
+      "loss": 0.0799,
+      "step": 1450
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 4.942731277533039e-05,
+      "loss": 0.0841,
+      "step": 1451
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 4.916299559471365e-05,
+      "loss": 0.0599,
+      "step": 1452
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 4.889867841409691e-05,
+      "loss": 0.1058,
+      "step": 1453
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 4.8634361233480176e-05,
+      "loss": 0.1809,
+      "step": 1454
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 4.837004405286343e-05,
+      "loss": 0.0034,
+      "step": 1455
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 4.810572687224669e-05,
+      "loss": 0.0163,
+      "step": 1456
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 4.784140969162995e-05,
+      "loss": 0.1942,
+      "step": 1457
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.7577092511013216e-05,
+      "loss": 0.0579,
+      "step": 1458
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 4.7312775330396474e-05,
+      "loss": 0.0828,
+      "step": 1459
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 4.7048458149779725e-05,
+      "loss": 0.3206,
+      "step": 1460
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 4.678414096916299e-05,
+      "loss": 0.0206,
+      "step": 1461
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 4.651982378854625e-05,
+      "loss": 0.0082,
+      "step": 1462
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 4.625550660792951e-05,
+      "loss": 0.0189,
+      "step": 1463
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 4.599118942731277e-05,
+      "loss": 0.0811,
+      "step": 1464
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 4.5726872246696036e-05,
+      "loss": 0.0118,
+      "step": 1465
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 4.546255506607929e-05,
+      "loss": 0.0462,
+      "step": 1466
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.519823788546255e-05,
+      "loss": 0.1703,
+      "step": 1467
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 4.493392070484581e-05,
+      "loss": 0.0369,
+      "step": 1468
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 4.4669603524229076e-05,
+      "loss": 0.0274,
+      "step": 1469
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.4405286343612334e-05,
+      "loss": 0.2661,
+      "step": 1470
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 4.4140969162995585e-05,
+      "loss": 0.0475,
+      "step": 1471
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 4.387665198237885e-05,
+      "loss": 0.1279,
+      "step": 1472
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.361233480176211e-05,
+      "loss": 0.196,
+      "step": 1473
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 4.334801762114537e-05,
+      "loss": 0.0381,
+      "step": 1474
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 4.308370044052863e-05,
+      "loss": 0.1028,
+      "step": 1475
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.2819383259911896e-05,
+      "loss": 0.1709,
+      "step": 1476
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 4.255506607929515e-05,
+      "loss": 0.156,
+      "step": 1477
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 4.229074889867841e-05,
+      "loss": 0.1543,
+      "step": 1478
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 4.202643171806167e-05,
+      "loss": 0.1417,
+      "step": 1479
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 4.176211453744493e-05,
+      "loss": 0.0376,
+      "step": 1480
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 4.1497797356828194e-05,
+      "loss": 0.0148,
+      "step": 1481
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 4.1233480176211445e-05,
+      "loss": 0.1573,
+      "step": 1482
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 4.096916299559471e-05,
+      "loss": 0.1412,
+      "step": 1483
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 4.070484581497797e-05,
+      "loss": 0.0097,
+      "step": 1484
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.044052863436123e-05,
+      "loss": 0.1147,
+      "step": 1485
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 4.017621145374449e-05,
+      "loss": 0.0044,
+      "step": 1486
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 3.9911894273127756e-05,
+      "loss": 0.0529,
+      "step": 1487
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 3.964757709251101e-05,
+      "loss": 0.1298,
+      "step": 1488
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 3.9383259911894265e-05,
+      "loss": 0.0488,
+      "step": 1489
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 3.911894273127753e-05,
+      "loss": 0.0664,
+      "step": 1490
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 3.885462555066079e-05,
+      "loss": 0.0961,
+      "step": 1491
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 3.8590308370044053e-05,
+      "loss": 0.0562,
+      "step": 1492
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 3.8325991189427305e-05,
+      "loss": 0.0781,
+      "step": 1493
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 3.806167400881057e-05,
+      "loss": 0.1106,
+      "step": 1494
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 3.779735682819383e-05,
+      "loss": 0.0093,
+      "step": 1495
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 3.753303964757709e-05,
+      "loss": 0.0888,
+      "step": 1496
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 3.726872246696035e-05,
+      "loss": 0.0139,
+      "step": 1497
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 3.700440528634361e-05,
+      "loss": 0.0402,
+      "step": 1498
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 3.674008810572687e-05,
+      "loss": 0.0231,
+      "step": 1499
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 3.647577092511013e-05,
+      "loss": 0.1871,
+      "step": 1500
+    },
+    {
+      "epoch": 13.76,
+      "eval_loss": 0.4036968946456909,
+      "eval_runtime": 20.5937,
+      "eval_samples_per_second": 79.976,
+      "eval_steps_per_second": 1.263,
+      "eval_wer": 0.3656419160453478,
+      "step": 1500
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 3.621145374449339e-05,
+      "loss": 0.158,
+      "step": 1501
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 3.594713656387665e-05,
+      "loss": 0.0098,
+      "step": 1502
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 3.568281938325991e-05,
+      "loss": 0.0414,
+      "step": 1503
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 3.5418502202643165e-05,
+      "loss": 0.1187,
+      "step": 1504
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 3.515418502202643e-05,
+      "loss": 0.1032,
+      "step": 1505
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 3.488986784140969e-05,
+      "loss": 0.1317,
+      "step": 1506
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 3.4625550660792946e-05,
+      "loss": 0.0996,
+      "step": 1507
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 3.436123348017621e-05,
+      "loss": 0.0236,
+      "step": 1508
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 3.409691629955947e-05,
+      "loss": 0.0329,
+      "step": 1509
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 3.383259911894273e-05,
+      "loss": 0.0813,
+      "step": 1510
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 3.356828193832599e-05,
+      "loss": 0.1203,
+      "step": 1511
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 3.3303964757709243e-05,
+      "loss": 0.2639,
+      "step": 1512
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 3.303964757709251e-05,
+      "loss": 0.0087,
+      "step": 1513
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 3.2775330396475767e-05,
+      "loss": 0.0399,
+      "step": 1514
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 3.2511013215859025e-05,
+      "loss": 0.0502,
+      "step": 1515
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 3.224669603524229e-05,
+      "loss": 0.1111,
+      "step": 1516
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 3.198237885462555e-05,
+      "loss": 0.1912,
+      "step": 1517
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 3.1718061674008806e-05,
+      "loss": 0.1152,
+      "step": 1518
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 3.145374449339207e-05,
+      "loss": 0.0832,
+      "step": 1519
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 3.118942731277533e-05,
+      "loss": 0.0357,
+      "step": 1520
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 3.092511013215859e-05,
+      "loss": 0.1363,
+      "step": 1521
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 3.0660792951541845e-05,
+      "loss": 0.0419,
+      "step": 1522
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 3.0396475770925107e-05,
+      "loss": 0.0503,
+      "step": 1523
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 3.0132158590308368e-05,
+      "loss": 0.0033,
+      "step": 1524
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.9867841409691626e-05,
+      "loss": 0.0506,
+      "step": 1525
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.9603524229074888e-05,
+      "loss": 0.1933,
+      "step": 1526
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.933920704845815e-05,
+      "loss": 0.0025,
+      "step": 1527
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.9074889867841408e-05,
+      "loss": 0.0882,
+      "step": 1528
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.881057268722467e-05,
+      "loss": 0.0378,
+      "step": 1529
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.854625550660793e-05,
+      "loss": 0.0722,
+      "step": 1530
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.8281938325991185e-05,
+      "loss": 0.0952,
+      "step": 1531
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.8017621145374447e-05,
+      "loss": 0.1082,
+      "step": 1532
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.7753303964757705e-05,
+      "loss": 0.0461,
+      "step": 1533
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.7488986784140967e-05,
+      "loss": 0.034,
+      "step": 1534
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.7224669603524228e-05,
+      "loss": 0.0689,
+      "step": 1535
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.6960352422907486e-05,
+      "loss": 0.035,
+      "step": 1536
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.6696035242290748e-05,
+      "loss": 0.0459,
+      "step": 1537
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.643171806167401e-05,
+      "loss": 0.1002,
+      "step": 1538
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.6167400881057268e-05,
+      "loss": 0.0232,
+      "step": 1539
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.5903083700440526e-05,
+      "loss": 0.0337,
+      "step": 1540
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.5638766519823784e-05,
+      "loss": 0.1067,
+      "step": 1541
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.5374449339207045e-05,
+      "loss": 0.0337,
+      "step": 1542
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.5110132158590307e-05,
+      "loss": 0.0262,
+      "step": 1543
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.4845814977973565e-05,
+      "loss": 0.0371,
+      "step": 1544
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.4581497797356827e-05,
+      "loss": 0.0445,
+      "step": 1545
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.4317180616740088e-05,
+      "loss": 0.02,
+      "step": 1546
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.4052863436123346e-05,
+      "loss": 0.005,
+      "step": 1547
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.3788546255506608e-05,
+      "loss": 0.0223,
+      "step": 1548
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.3524229074889863e-05,
+      "loss": 0.1106,
+      "step": 1549
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.3259911894273124e-05,
+      "loss": 0.0934,
+      "step": 1550
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.2995594713656386e-05,
+      "loss": 0.1036,
+      "step": 1551
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.2731277533039644e-05,
+      "loss": 0.0335,
+      "step": 1552
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.2466960352422905e-05,
+      "loss": 0.0534,
+      "step": 1553
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.2202643171806167e-05,
+      "loss": 0.0529,
+      "step": 1554
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1938325991189425e-05,
+      "loss": 0.0389,
+      "step": 1555
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1674008810572687e-05,
+      "loss": 0.0417,
+      "step": 1556
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1409691629955948e-05,
+      "loss": 0.0386,
+      "step": 1557
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1145374449339206e-05,
+      "loss": 0.0049,
+      "step": 1558
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.0881057268722464e-05,
+      "loss": 0.0209,
+      "step": 1559
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.0616740088105722e-05,
+      "loss": 0.0941,
+      "step": 1560
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.0352422907488984e-05,
+      "loss": 0.0336,
+      "step": 1561
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.0088105726872246e-05,
+      "loss": 0.1241,
+      "step": 1562
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.9823788546255504e-05,
+      "loss": 0.0914,
+      "step": 1563
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.9559471365638765e-05,
+      "loss": 0.0167,
+      "step": 1564
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.9295154185022027e-05,
+      "loss": 0.1179,
+      "step": 1565
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.9030837004405285e-05,
+      "loss": 0.0201,
+      "step": 1566
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.8766519823788546e-05,
+      "loss": 0.0158,
+      "step": 1567
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.8502202643171805e-05,
+      "loss": 0.0146,
+      "step": 1568
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.8237885462555066e-05,
+      "loss": 0.1066,
+      "step": 1569
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.7973568281938324e-05,
+      "loss": 0.062,
+      "step": 1570
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.7709251101321582e-05,
+      "loss": 0.0054,
+      "step": 1571
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.7444933920704844e-05,
+      "loss": 0.1336,
+      "step": 1572
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.7180616740088105e-05,
+      "loss": 0.0829,
+      "step": 1573
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.6916299559471364e-05,
+      "loss": 0.0597,
+      "step": 1574
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.6651982378854622e-05,
+      "loss": 0.0492,
+      "step": 1575
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.6387665198237883e-05,
+      "loss": 0.1838,
+      "step": 1576
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.6123348017621145e-05,
+      "loss": 0.0363,
+      "step": 1577
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.5859030837004403e-05,
+      "loss": 0.166,
+      "step": 1578
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.5594713656387664e-05,
+      "loss": 0.1688,
+      "step": 1579
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.5330396475770923e-05,
+      "loss": 0.0853,
+      "step": 1580
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.5066079295154184e-05,
+      "loss": 0.0364,
+      "step": 1581
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.4801762114537444e-05,
+      "loss": 0.042,
+      "step": 1582
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.4537444933920704e-05,
+      "loss": 0.1183,
+      "step": 1583
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.4273127753303965e-05,
+      "loss": 0.0312,
+      "step": 1584
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.4008810572687224e-05,
+      "loss": 0.0842,
+      "step": 1585
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3744493392070483e-05,
+      "loss": 0.2848,
+      "step": 1586
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3480176211453743e-05,
+      "loss": 0.087,
+      "step": 1587
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3215859030837005e-05,
+      "loss": 0.1918,
+      "step": 1588
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.2951541850220263e-05,
+      "loss": 0.0386,
+      "step": 1589
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.2687224669603523e-05,
+      "loss": 0.024,
+      "step": 1590
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.2422907488986783e-05,
+      "loss": 0.0622,
+      "step": 1591
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.2158590308370044e-05,
+      "loss": 0.1429,
+      "step": 1592
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.1894273127753304e-05,
+      "loss": 0.1573,
+      "step": 1593
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.1629955947136562e-05,
+      "loss": 0.0564,
+      "step": 1594
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.1365638766519822e-05,
+      "loss": 0.0118,
+      "step": 1595
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.1101321585903083e-05,
+      "loss": 0.0912,
+      "step": 1596
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.0837004405286343e-05,
+      "loss": 0.0192,
+      "step": 1597
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.0572687224669603e-05,
+      "loss": 0.0225,
+      "step": 1598
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.0308370044052861e-05,
+      "loss": 0.0153,
+      "step": 1599
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.0044052863436123e-05,
+      "loss": 0.2273,
+      "step": 1600
+    },
+    {
+      "epoch": 14.68,
+      "eval_loss": 0.39717164635658264,
+      "eval_runtime": 20.6414,
+      "eval_samples_per_second": 79.791,
+      "eval_steps_per_second": 1.26,
+      "eval_wer": 0.3589010315595956,
+      "step": 1600
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 9.779735682819383e-06,
+      "loss": 0.1062,
+      "step": 1601
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 9.515418502202642e-06,
+      "loss": 0.0487,
+      "step": 1602
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 9.251101321585902e-06,
+      "loss": 0.0587,
+      "step": 1603
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.986784140969162e-06,
+      "loss": 0.0118,
+      "step": 1604
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.722466960352422e-06,
+      "loss": 0.0244,
+      "step": 1605
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.458149779735682e-06,
+      "loss": 0.0486,
+      "step": 1606
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.193832599118942e-06,
+      "loss": 0.0769,
+      "step": 1607
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 7.929515418502201e-06,
+      "loss": 0.0035,
+      "step": 1608
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 7.665198237885461e-06,
+      "loss": 0.1179,
+      "step": 1609
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 7.400881057268722e-06,
+      "loss": 0.0492,
+      "step": 1610
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 7.136563876651983e-06,
+      "loss": 0.0489,
+      "step": 1611
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 6.872246696035242e-06,
+      "loss": 0.0699,
+      "step": 1612
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 6.607929515418502e-06,
+      "loss": 0.0845,
+      "step": 1613
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 6.343612334801761e-06,
+      "loss": 0.01,
+      "step": 1614
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 6.079295154185022e-06,
+      "loss": 0.0203,
+      "step": 1615
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 5.814977973568281e-06,
+      "loss": 0.016,
+      "step": 1616
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 5.550660792951542e-06,
+      "loss": 0.159,
+      "step": 1617
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 5.2863436123348015e-06,
+      "loss": 0.067,
+      "step": 1618
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 5.022026431718061e-06,
+      "loss": 0.135,
+      "step": 1619
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 4.757709251101321e-06,
+      "loss": 0.0018,
+      "step": 1620
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 4.493392070484581e-06,
+      "loss": 0.0624,
+      "step": 1621
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.229074889867841e-06,
+      "loss": 0.0381,
+      "step": 1622
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 3.964757709251101e-06,
+      "loss": 0.0171,
+      "step": 1623
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 3.700440528634361e-06,
+      "loss": 0.0886,
+      "step": 1624
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 3.436123348017621e-06,
+      "loss": 0.0969,
+      "step": 1625
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 3.1718061674008807e-06,
+      "loss": 0.1242,
+      "step": 1626
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.9074889867841405e-06,
+      "loss": 0.1019,
+      "step": 1627
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.6431718061674008e-06,
+      "loss": 0.0034,
+      "step": 1628
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.3788546255506606e-06,
+      "loss": 0.015,
+      "step": 1629
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.1145374449339204e-06,
+      "loss": 0.0473,
+      "step": 1630
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.8502202643171805e-06,
+      "loss": 0.0464,
+      "step": 1631
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.5859030837004403e-06,
+      "loss": 0.0576,
+      "step": 1632
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.3215859030837004e-06,
+      "loss": 0.0443,
+      "step": 1633
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.0572687224669602e-06,
+      "loss": 0.0393,
+      "step": 1634
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.929515418502202e-07,
+      "loss": 0.0359,
+      "step": 1635
+    },
+    {
+      "epoch": 15.0,
+      "step": 1635,
+      "total_flos": 7.077752219977646e+18,
+      "train_loss": 1.0806824558201624,
+      "train_runtime": 1173.0236,
+      "train_samples_per_second": 44.475,
+      "train_steps_per_second": 1.394
+    }
+  ],
+  "max_steps": 1635,
+  "num_train_epochs": 15,
+  "total_flos": 7.077752219977646e+18,
+  "trial_name": null,
+  "trial_params": null
+}