diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,19753 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 29.99310344827586,
+  "global_step": 3240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-08,
+      "loss": 15.6125,
+      "step": 1
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4e-08,
+      "loss": 14.3834,
+      "step": 2
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 6.000000000000001e-08,
+      "loss": 14.1192,
+      "step": 3
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 8e-08,
+      "loss": 14.6255,
+      "step": 4
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 13.9643,
+      "step": 5
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 17.6492,
+      "step": 6
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 16.1044,
+      "step": 7
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.2000000000000002e-07,
+      "loss": 14.9581,
+      "step": 8
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.4e-07,
+      "loss": 13.5407,
+      "step": 9
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.6e-07,
+      "loss": 14.5327,
+      "step": 10
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8e-07,
+      "loss": 15.2553,
+      "step": 11
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 16.276,
+      "step": 12
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2e-07,
+      "loss": 14.025,
+      "step": 13
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4000000000000003e-07,
+      "loss": 14.4253,
+      "step": 14
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.6e-07,
+      "loss": 13.596,
+      "step": 15
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8e-07,
+      "loss": 14.7193,
+      "step": 16
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 14.0049,
+      "step": 17
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.2e-07,
+      "loss": 12.7994,
+      "step": 18
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.4000000000000003e-07,
+      "loss": 13.474,
+      "step": 19
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.6e-07,
+      "loss": 14.6847,
+      "step": 20
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 3.8e-07,
+      "loss": 12.7952,
+      "step": 21
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 12.9517,
+      "step": 22
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.2000000000000006e-07,
+      "loss": 15.2301,
+      "step": 23
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.2000000000000006e-07,
+      "loss": 14.3598,
+      "step": 24
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4e-07,
+      "loss": 13.0184,
+      "step": 25
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.6000000000000004e-07,
+      "loss": 13.9059,
+      "step": 26
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.800000000000001e-07,
+      "loss": 13.244,
+      "step": 27
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 12.498,
+      "step": 28
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 5.2e-07,
+      "loss": 12.5503,
+      "step": 29
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 5.4e-07,
+      "loss": 13.2954,
+      "step": 30
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 5.6e-07,
+      "loss": 10.6686,
+      "step": 31
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 5.800000000000001e-07,
+      "loss": 13.5655,
+      "step": 32
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 13.5897,
+      "step": 33
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.200000000000001e-07,
+      "loss": 10.7764,
+      "step": 34
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.4e-07,
+      "loss": 11.7895,
+      "step": 35
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.6e-07,
+      "loss": 11.029,
+      "step": 36
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.6e-07,
+      "loss": 9.8552,
+      "step": 37
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.800000000000001e-07,
+      "loss": 12.1996,
+      "step": 38
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 10.6448,
+      "step": 39
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2e-07,
+      "loss": 9.2266,
+      "step": 40
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4e-07,
+      "loss": 9.3842,
+      "step": 41
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.6e-07,
+      "loss": 9.9854,
+      "step": 42
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.8e-07,
+      "loss": 8.8267,
+      "step": 43
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 9.5824,
+      "step": 44
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.200000000000001e-07,
+      "loss": 8.3395,
+      "step": 45
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 7.8379,
+      "step": 46
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.6e-07,
+      "loss": 8.333,
+      "step": 47
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.8e-07,
+      "loss": 7.5948,
+      "step": 48
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 6.6517,
+      "step": 49
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.200000000000001e-07,
+      "loss": 6.71,
+      "step": 50
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.400000000000001e-07,
+      "loss": 6.2793,
+      "step": 51
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.600000000000001e-07,
+      "loss": 5.7408,
+      "step": 52
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.800000000000001e-07,
+      "loss": 5.4254,
+      "step": 53
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 5.1803,
+      "step": 54
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02e-06,
+      "loss": 4.5298,
+      "step": 55
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.04e-06,
+      "loss": 4.3988,
+      "step": 56
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.06e-06,
+      "loss": 4.29,
+      "step": 57
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.08e-06,
+      "loss": 4.0272,
+      "step": 58
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.1e-06,
+      "loss": 3.9508,
+      "step": 59
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.12e-06,
+      "loss": 4.084,
+      "step": 60
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.14e-06,
+      "loss": 3.8152,
+      "step": 61
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.1600000000000001e-06,
+      "loss": 3.8503,
+      "step": 62
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1800000000000001e-06,
+      "loss": 3.8557,
+      "step": 63
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 3.8566,
+      "step": 64
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2200000000000002e-06,
+      "loss": 3.8701,
+      "step": 65
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2400000000000002e-06,
+      "loss": 3.8901,
+      "step": 66
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.26e-06,
+      "loss": 3.7997,
+      "step": 67
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.28e-06,
+      "loss": 3.7898,
+      "step": 68
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.3e-06,
+      "loss": 3.7657,
+      "step": 69
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.32e-06,
+      "loss": 3.7088,
+      "step": 70
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.34e-06,
+      "loss": 3.7135,
+      "step": 71
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3600000000000001e-06,
+      "loss": 3.6795,
+      "step": 72
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3800000000000001e-06,
+      "loss": 3.6537,
+      "step": 73
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 3.5713,
+      "step": 74
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.42e-06,
+      "loss": 3.5568,
+      "step": 75
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.44e-06,
+      "loss": 3.4995,
+      "step": 76
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.46e-06,
+      "loss": 3.5318,
+      "step": 77
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.48e-06,
+      "loss": 3.5824,
+      "step": 78
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5e-06,
+      "loss": 3.4513,
+      "step": 79
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.52e-06,
+      "loss": 3.435,
+      "step": 80
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.54e-06,
+      "loss": 3.4863,
+      "step": 81
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.56e-06,
+      "loss": 3.4667,
+      "step": 82
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 3.3752,
+      "step": 83
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 3.3746,
+      "step": 84
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6200000000000002e-06,
+      "loss": 3.6092,
+      "step": 85
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6400000000000002e-06,
+      "loss": 3.3079,
+      "step": 86
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6600000000000002e-06,
+      "loss": 3.3401,
+      "step": 87
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6800000000000002e-06,
+      "loss": 3.3808,
+      "step": 88
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 3.3068,
+      "step": 89
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.72e-06,
+      "loss": 3.2597,
+      "step": 90
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.74e-06,
+      "loss": 3.2623,
+      "step": 91
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.76e-06,
+      "loss": 3.2299,
+      "step": 92
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.7800000000000001e-06,
+      "loss": 3.2305,
+      "step": 93
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 3.3822,
+      "step": 94
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8200000000000002e-06,
+      "loss": 3.1923,
+      "step": 95
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8400000000000002e-06,
+      "loss": 3.2007,
+      "step": 96
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.8600000000000002e-06,
+      "loss": 3.1704,
+      "step": 97
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.8800000000000002e-06,
+      "loss": 3.1524,
+      "step": 98
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 3.1479,
+      "step": 99
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9200000000000003e-06,
+      "loss": 3.1646,
+      "step": 100
+    },
+    {
+      "epoch": 0.92,
+      "eval_loss": 3.210620641708374,
+      "eval_runtime": 65.7809,
+      "eval_samples_per_second": 25.038,
+      "eval_steps_per_second": 1.566,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.94e-06,
+      "loss": 3.1444,
+      "step": 101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9600000000000003e-06,
+      "loss": 3.1609,
+      "step": 102
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.98e-06,
+      "loss": 3.1686,
+      "step": 103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 3.1571,
+      "step": 104
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.02e-06,
+      "loss": 3.156,
+      "step": 105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.04e-06,
+      "loss": 3.1405,
+      "step": 106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.06e-06,
+      "loss": 3.1422,
+      "step": 107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.08e-06,
+      "loss": 3.1269,
+      "step": 108
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 5.4893,
+      "step": 109
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.12e-06,
+      "loss": 3.0968,
+      "step": 110
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.1400000000000003e-06,
+      "loss": 3.0483,
+      "step": 111
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.16e-06,
+      "loss": 3.1096,
+      "step": 112
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.1800000000000003e-06,
+      "loss": 3.0971,
+      "step": 113
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.2e-06,
+      "loss": 3.0985,
+      "step": 114
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.2200000000000003e-06,
+      "loss": 3.1238,
+      "step": 115
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.24e-06,
+      "loss": 3.0668,
+      "step": 116
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.2600000000000004e-06,
+      "loss": 3.0742,
+      "step": 117
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.28e-06,
+      "loss": 3.0523,
+      "step": 118
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 3.0594,
+      "step": 119
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.3200000000000002e-06,
+      "loss": 2.9866,
+      "step": 120
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.3400000000000005e-06,
+      "loss": 2.9892,
+      "step": 121
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.3600000000000003e-06,
+      "loss": 2.9748,
+      "step": 122
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.38e-06,
+      "loss": 2.984,
+      "step": 123
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 3.0392,
+      "step": 124
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.42e-06,
+      "loss": 2.9758,
+      "step": 125
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.4400000000000004e-06,
+      "loss": 2.9587,
+      "step": 126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.46e-06,
+      "loss": 3.0037,
+      "step": 127
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.4800000000000004e-06,
+      "loss": 2.9908,
+      "step": 128
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.5e-06,
+      "loss": 2.8789,
+      "step": 129
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.52e-06,
+      "loss": 2.9244,
+      "step": 130
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.5400000000000002e-06,
+      "loss": 2.844,
+      "step": 131
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.56e-06,
+      "loss": 2.8837,
+      "step": 132
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 2.8248,
+      "step": 133
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6e-06,
+      "loss": 2.8488,
+      "step": 134
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6200000000000003e-06,
+      "loss": 2.8064,
+      "step": 135
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.64e-06,
+      "loss": 2.752,
+      "step": 136
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6600000000000004e-06,
+      "loss": 2.7703,
+      "step": 137
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.68e-06,
+      "loss": 2.7082,
+      "step": 138
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 2.632,
+      "step": 139
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.7200000000000002e-06,
+      "loss": 2.6731,
+      "step": 140
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.7400000000000004e-06,
+      "loss": 2.6164,
+      "step": 141
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.7600000000000003e-06,
+      "loss": 2.5367,
+      "step": 142
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.7800000000000005e-06,
+      "loss": 2.4815,
+      "step": 143
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 2.4263,
+      "step": 144
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.82e-06,
+      "loss": 2.3901,
+      "step": 145
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.84e-06,
+      "loss": 2.3535,
+      "step": 146
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.86e-06,
+      "loss": 2.2753,
+      "step": 147
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.88e-06,
+      "loss": 2.1824,
+      "step": 148
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.9e-06,
+      "loss": 2.1984,
+      "step": 149
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.92e-06,
+      "loss": 2.0474,
+      "step": 150
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.9400000000000002e-06,
+      "loss": 2.0605,
+      "step": 151
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.96e-06,
+      "loss": 1.9639,
+      "step": 152
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9800000000000003e-06,
+      "loss": 1.9454,
+      "step": 153
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 3e-06,
+      "loss": 1.8509,
+      "step": 154
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.0200000000000003e-06,
+      "loss": 1.8947,
+      "step": 155
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.04e-06,
+      "loss": 1.6702,
+      "step": 156
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.0600000000000003e-06,
+      "loss": 1.6755,
+      "step": 157
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.08e-06,
+      "loss": 1.6647,
+      "step": 158
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 1.4576,
+      "step": 159
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.12e-06,
+      "loss": 1.4793,
+      "step": 160
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.1400000000000004e-06,
+      "loss": 1.3493,
+      "step": 161
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.1600000000000002e-06,
+      "loss": 1.2696,
+      "step": 162
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.1800000000000005e-06,
+      "loss": 1.2714,
+      "step": 163
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 1.1438,
+      "step": 164
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.2200000000000005e-06,
+      "loss": 1.0983,
+      "step": 165
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.2400000000000003e-06,
+      "loss": 1.1143,
+      "step": 166
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.2600000000000006e-06,
+      "loss": 1.0719,
+      "step": 167
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.2800000000000004e-06,
+      "loss": 0.9655,
+      "step": 168
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.962,
+      "step": 169
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.3200000000000004e-06,
+      "loss": 0.9304,
+      "step": 170
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.3400000000000006e-06,
+      "loss": 0.8993,
+      "step": 171
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.3600000000000004e-06,
+      "loss": 0.8029,
+      "step": 172
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.3800000000000007e-06,
+      "loss": 0.7713,
+      "step": 173
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.8287,
+      "step": 174
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.4200000000000007e-06,
+      "loss": 0.7307,
+      "step": 175
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.44e-06,
+      "loss": 0.7621,
+      "step": 176
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.46e-06,
+      "loss": 0.7228,
+      "step": 177
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.48e-06,
+      "loss": 0.5989,
+      "step": 178
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5e-06,
+      "loss": 0.6095,
+      "step": 179
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.52e-06,
+      "loss": 0.6489,
+      "step": 180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.54e-06,
+      "loss": 0.6002,
+      "step": 181
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.5600000000000002e-06,
+      "loss": 0.5754,
+      "step": 182
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.58e-06,
+      "loss": 0.6392,
+      "step": 183
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.5334,
+      "step": 184
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.62e-06,
+      "loss": 0.5087,
+      "step": 185
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.6400000000000003e-06,
+      "loss": 0.5398,
+      "step": 186
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.66e-06,
+      "loss": 0.5603,
+      "step": 187
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.6800000000000003e-06,
+      "loss": 0.4915,
+      "step": 188
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.7e-06,
+      "loss": 0.6469,
+      "step": 189
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7200000000000004e-06,
+      "loss": 0.4197,
+      "step": 190
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.74e-06,
+      "loss": 0.5177,
+      "step": 191
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.7600000000000004e-06,
+      "loss": 0.4551,
+      "step": 192
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.7800000000000002e-06,
+      "loss": 0.5005,
+      "step": 193
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.4125,
+      "step": 194
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.820000000000001e-06,
+      "loss": 0.443,
+      "step": 195
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.8400000000000005e-06,
+      "loss": 0.4782,
+      "step": 196
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.86e-06,
+      "loss": 0.3756,
+      "step": 197
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.88e-06,
+      "loss": 0.3613,
+      "step": 198
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.4457,
+      "step": 199
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.920000000000001e-06,
+      "loss": 0.368,
+      "step": 200
+    },
+    {
+      "epoch": 1.85,
+      "eval_loss": 2.9963252544403076,
+      "eval_runtime": 65.1826,
+      "eval_samples_per_second": 25.267,
+      "eval_steps_per_second": 1.58,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.94e-06,
+      "loss": 0.4106,
+      "step": 201
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.96e-06,
+      "loss": 0.4839,
+      "step": 202
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.980000000000001e-06,
+      "loss": 0.3475,
+      "step": 203
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.3767,
+      "step": 204
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.0200000000000005e-06,
+      "loss": 0.471,
+      "step": 205
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.04e-06,
+      "loss": 0.3526,
+      "step": 206
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.060000000000001e-06,
+      "loss": 0.3153,
+      "step": 207
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.08e-06,
+      "loss": 0.4261,
+      "step": 208
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.1e-06,
+      "loss": 0.2649,
+      "step": 209
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.12e-06,
+      "loss": 0.3328,
+      "step": 210
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.14e-06,
+      "loss": 0.4615,
+      "step": 211
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.16e-06,
+      "loss": 0.3119,
+      "step": 212
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.18e-06,
+      "loss": 0.2961,
+      "step": 213
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.3503,
+      "step": 214
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.22e-06,
+      "loss": 0.3237,
+      "step": 215
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.24e-06,
+      "loss": 0.3468,
+      "step": 216
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.26e-06,
+      "loss": 0.5931,
+      "step": 217
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.2800000000000005e-06,
+      "loss": 0.2666,
+      "step": 218
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.3e-06,
+      "loss": 0.3831,
+      "step": 219
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.32e-06,
+      "loss": 0.319,
+      "step": 220
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.34e-06,
+      "loss": 0.282,
+      "step": 221
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.360000000000001e-06,
+      "loss": 0.43,
+      "step": 222
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.38e-06,
+      "loss": 0.2975,
+      "step": 223
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4e-06,
+      "loss": 0.248,
+      "step": 224
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.42e-06,
+      "loss": 0.2405,
+      "step": 225
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.440000000000001e-06,
+      "loss": 0.3203,
+      "step": 226
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.4600000000000005e-06,
+      "loss": 0.2599,
+      "step": 227
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.48e-06,
+      "loss": 0.3237,
+      "step": 228
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.5e-06,
+      "loss": 0.326,
+      "step": 229
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.520000000000001e-06,
+      "loss": 0.2547,
+      "step": 230
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.540000000000001e-06,
+      "loss": 0.3276,
+      "step": 231
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.56e-06,
+      "loss": 0.2848,
+      "step": 232
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.58e-06,
+      "loss": 0.2744,
+      "step": 233
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.3478,
+      "step": 234
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.620000000000001e-06,
+      "loss": 0.367,
+      "step": 235
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.6400000000000005e-06,
+      "loss": 0.2781,
+      "step": 236
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.66e-06,
+      "loss": 0.293,
+      "step": 237
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.680000000000001e-06,
+      "loss": 0.3633,
+      "step": 238
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.7e-06,
+      "loss": 0.2703,
+      "step": 239
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.7200000000000005e-06,
+      "loss": 0.3267,
+      "step": 240
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.74e-06,
+      "loss": 0.3551,
+      "step": 241
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.76e-06,
+      "loss": 0.1663,
+      "step": 242
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.78e-06,
+      "loss": 0.2626,
+      "step": 243
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.2863,
+      "step": 244
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8200000000000004e-06,
+      "loss": 0.2602,
+      "step": 245
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.84e-06,
+      "loss": 0.2992,
+      "step": 246
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.86e-06,
+      "loss": 0.2214,
+      "step": 247
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.880000000000001e-06,
+      "loss": 0.2545,
+      "step": 248
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.2337,
+      "step": 249
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.92e-06,
+      "loss": 0.2238,
+      "step": 250
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.94e-06,
+      "loss": 0.2593,
+      "step": 251
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.960000000000001e-06,
+      "loss": 0.2872,
+      "step": 252
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.980000000000001e-06,
+      "loss": 0.2478,
+      "step": 253
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 5e-06,
+      "loss": 0.3048,
+      "step": 254
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 5.02e-06,
+      "loss": 0.2917,
+      "step": 255
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 5.04e-06,
+      "loss": 0.2456,
+      "step": 256
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 5.060000000000001e-06,
+      "loss": 0.3209,
+      "step": 257
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 5.0800000000000005e-06,
+      "loss": 0.2766,
+      "step": 258
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 5.1e-06,
+      "loss": 0.3059,
+      "step": 259
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 5.12e-06,
+      "loss": 0.317,
+      "step": 260
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 5.140000000000001e-06,
+      "loss": 0.2092,
+      "step": 261
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 5.1600000000000006e-06,
+      "loss": 0.1998,
+      "step": 262
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 5.18e-06,
+      "loss": 0.4046,
+      "step": 263
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 5.2e-06,
+      "loss": 0.1854,
+      "step": 264
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 5.220000000000001e-06,
+      "loss": 0.2286,
+      "step": 265
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 5.240000000000001e-06,
+      "loss": 0.267,
+      "step": 266
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 5.2600000000000005e-06,
+      "loss": 0.234,
+      "step": 267
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 5.28e-06,
+      "loss": 0.1687,
+      "step": 268
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 5.300000000000001e-06,
+      "loss": 0.3434,
+      "step": 269
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 5.320000000000001e-06,
+      "loss": 0.1994,
+      "step": 270
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 5.3400000000000005e-06,
+      "loss": 0.2512,
+      "step": 271
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 5.36e-06,
+      "loss": 0.2494,
+      "step": 272
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 5.380000000000001e-06,
+      "loss": 0.3121,
+      "step": 273
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 0.2294,
+      "step": 274
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 5.420000000000001e-06,
+      "loss": 0.2397,
+      "step": 275
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 5.4400000000000004e-06,
+      "loss": 0.2391,
+      "step": 276
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 5.460000000000001e-06,
+      "loss": 0.1838,
+      "step": 277
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 5.480000000000001e-06,
+      "loss": 0.2328,
+      "step": 278
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.2325,
+      "step": 279
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 5.5200000000000005e-06,
+      "loss": 0.1432,
+      "step": 280
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 5.540000000000001e-06,
+      "loss": 0.2279,
+      "step": 281
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 5.560000000000001e-06,
+      "loss": 0.3173,
+      "step": 282
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 5.580000000000001e-06,
+      "loss": 0.2146,
+      "step": 283
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 0.2216,
+      "step": 284
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 5.620000000000001e-06,
+      "loss": 0.2518,
+      "step": 285
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 5.64e-06,
+      "loss": 0.2037,
+      "step": 286
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 5.66e-06,
+      "loss": 0.1749,
+      "step": 287
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 5.68e-06,
+      "loss": 0.1666,
+      "step": 288
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.7e-06,
+      "loss": 0.2107,
+      "step": 289
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.72e-06,
+      "loss": 0.2696,
+      "step": 290
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.74e-06,
+      "loss": 0.2655,
+      "step": 291
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.76e-06,
+      "loss": 0.1744,
+      "step": 292
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 5.78e-06,
+      "loss": 0.1801,
+      "step": 293
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 5.8e-06,
+      "loss": 0.2649,
+      "step": 294
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 5.82e-06,
+      "loss": 0.2587,
+      "step": 295
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 5.84e-06,
+      "loss": 0.254,
+      "step": 296
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 5.86e-06,
+      "loss": 0.2889,
+      "step": 297
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 5.8800000000000005e-06,
+      "loss": 0.1717,
+      "step": 298
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 5.9e-06,
+      "loss": 0.2457,
+      "step": 299
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 5.92e-06,
+      "loss": 0.2252,
+      "step": 300
+    },
+    {
+      "epoch": 2.77,
+      "eval_loss": 2.8077850341796875,
+      "eval_runtime": 64.8994,
+      "eval_samples_per_second": 25.378,
+      "eval_steps_per_second": 1.587,
+      "eval_wer": 0.9998978653865795,
+      "step": 300
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 5.94e-06,
+      "loss": 0.1672,
+      "step": 301
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 5.9600000000000005e-06,
+      "loss": 0.2313,
+      "step": 302
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 5.98e-06,
+      "loss": 0.2185,
+      "step": 303
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 6e-06,
+      "loss": 0.2135,
+      "step": 304
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 6.02e-06,
+      "loss": 0.142,
+      "step": 305
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 6.040000000000001e-06,
+      "loss": 0.2611,
+      "step": 306
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 6.0600000000000004e-06,
+      "loss": 0.3132,
+      "step": 307
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 6.08e-06,
+      "loss": 0.1803,
+      "step": 308
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.1e-06,
+      "loss": 0.1884,
+      "step": 309
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 6.120000000000001e-06,
+      "loss": 0.2324,
+      "step": 310
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 6.1400000000000005e-06,
+      "loss": 0.1765,
+      "step": 311
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 6.16e-06,
+      "loss": 0.1979,
+      "step": 312
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.18e-06,
+      "loss": 0.3315,
+      "step": 313
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 6.200000000000001e-06,
+      "loss": 0.2124,
+      "step": 314
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 6.220000000000001e-06,
+      "loss": 0.2556,
+      "step": 315
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 6.24e-06,
+      "loss": 0.4031,
+      "step": 316
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 6.26e-06,
+      "loss": 0.1566,
+      "step": 317
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 6.280000000000001e-06,
+      "loss": 0.2123,
+      "step": 318
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 6.300000000000001e-06,
+      "loss": 0.228,
+      "step": 319
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.3200000000000005e-06,
+      "loss": 0.1984,
+      "step": 320
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.34e-06,
+      "loss": 0.1997,
+      "step": 321
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.360000000000001e-06,
+      "loss": 0.2141,
+      "step": 322
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 6.380000000000001e-06,
+      "loss": 0.269,
+      "step": 323
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 0.2609,
+      "step": 324
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.42e-06,
+      "loss": 0.3147,
+      "step": 325
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.440000000000001e-06,
+      "loss": 0.1577,
+      "step": 326
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.460000000000001e-06,
+      "loss": 0.1512,
+      "step": 327
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.480000000000001e-06,
+      "loss": 0.1501,
+      "step": 328
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.1725,
+      "step": 329
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.520000000000001e-06,
+      "loss": 0.1883,
+      "step": 330
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.540000000000001e-06,
+      "loss": 0.2035,
+      "step": 331
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.560000000000001e-06,
+      "loss": 0.1721,
+      "step": 332
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 0.205,
+      "step": 333
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 0.2412,
+      "step": 334
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.620000000000001e-06,
+      "loss": 0.1638,
+      "step": 335
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.640000000000001e-06,
+      "loss": 0.1782,
+      "step": 336
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.660000000000001e-06,
+      "loss": 0.1918,
+      "step": 337
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.680000000000001e-06,
+      "loss": 0.1434,
+      "step": 338
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.700000000000001e-06,
+      "loss": 0.146,
+      "step": 339
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.720000000000001e-06,
+      "loss": 0.176,
+      "step": 340
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.740000000000001e-06,
+      "loss": 0.1586,
+      "step": 341
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.760000000000001e-06,
+      "loss": 0.1255,
+      "step": 342
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.780000000000001e-06,
+      "loss": 0.1837,
+      "step": 343
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 0.2158,
+      "step": 344
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.820000000000001e-06,
+      "loss": 0.1843,
+      "step": 345
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.8400000000000014e-06,
+      "loss": 0.15,
+      "step": 346
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.860000000000001e-06,
+      "loss": 0.1068,
+      "step": 347
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.88e-06,
+      "loss": 0.1362,
+      "step": 348
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.9e-06,
+      "loss": 0.1877,
+      "step": 349
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.92e-06,
+      "loss": 0.1177,
+      "step": 350
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.9400000000000005e-06,
+      "loss": 0.1268,
+      "step": 351
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.96e-06,
+      "loss": 0.2077,
+      "step": 352
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.98e-06,
+      "loss": 0.1724,
+      "step": 353
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 7e-06,
+      "loss": 0.1851,
+      "step": 354
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 7.0200000000000006e-06,
+      "loss": 0.1669,
+      "step": 355
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 7.04e-06,
+      "loss": 0.1602,
+      "step": 356
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 7.06e-06,
+      "loss": 0.1362,
+      "step": 357
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 7.08e-06,
+      "loss": 0.2499,
+      "step": 358
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 7.100000000000001e-06,
+      "loss": 0.1769,
+      "step": 359
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 7.1200000000000004e-06,
+      "loss": 0.2172,
+      "step": 360
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 7.14e-06,
+      "loss": 0.0999,
+      "step": 361
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 7.16e-06,
+      "loss": 0.2302,
+      "step": 362
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 7.180000000000001e-06,
+      "loss": 0.1166,
+      "step": 363
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 0.1406,
+      "step": 364
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 7.22e-06,
+      "loss": 0.1341,
+      "step": 365
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 7.24e-06,
+      "loss": 0.1469,
+      "step": 366
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 7.260000000000001e-06,
+      "loss": 0.1347,
+      "step": 367
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 7.280000000000001e-06,
+      "loss": 0.2344,
+      "step": 368
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 7.3e-06,
+      "loss": 0.192,
+      "step": 369
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 7.32e-06,
+      "loss": 0.1173,
+      "step": 370
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 7.340000000000001e-06,
+      "loss": 0.1799,
+      "step": 371
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 7.360000000000001e-06,
+      "loss": 0.1231,
+      "step": 372
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 7.3800000000000005e-06,
+      "loss": 0.1413,
+      "step": 373
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 7.4e-06,
+      "loss": 0.1925,
+      "step": 374
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 7.420000000000001e-06,
+      "loss": 0.128,
+      "step": 375
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 7.440000000000001e-06,
+      "loss": 0.1296,
+      "step": 376
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 7.4600000000000006e-06,
+      "loss": 0.1771,
+      "step": 377
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 7.48e-06,
+      "loss": 0.2118,
+      "step": 378
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.1379,
+      "step": 379
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 7.520000000000001e-06,
+      "loss": 0.2187,
+      "step": 380
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 7.540000000000001e-06,
+      "loss": 0.1757,
+      "step": 381
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 7.5600000000000005e-06,
+      "loss": 0.1432,
+      "step": 382
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 7.58e-06,
+      "loss": 0.2369,
+      "step": 383
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 0.1435,
+      "step": 384
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 7.620000000000001e-06,
+      "loss": 0.1604,
+      "step": 385
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 7.640000000000001e-06,
+      "loss": 0.1676,
+      "step": 386
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 7.660000000000001e-06,
+      "loss": 0.2103,
+      "step": 387
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 7.680000000000001e-06,
+      "loss": 0.1243,
+      "step": 388
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 7.7e-06,
+      "loss": 0.2333,
+      "step": 389
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 7.72e-06,
+      "loss": 0.2131,
+      "step": 390
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 7.74e-06,
+      "loss": 0.1776,
+      "step": 391
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 7.76e-06,
+      "loss": 0.1968,
+      "step": 392
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 7.78e-06,
+      "loss": 0.195,
+      "step": 393
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 0.1526,
+      "step": 394
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 7.820000000000001e-06,
+      "loss": 0.1798,
+      "step": 395
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 7.840000000000001e-06,
+      "loss": 0.1734,
+      "step": 396
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 7.860000000000001e-06,
+      "loss": 0.1206,
+      "step": 397
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 7.88e-06,
+      "loss": 0.1241,
+      "step": 398
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 7.9e-06,
+      "loss": 0.1714,
+      "step": 399
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 7.92e-06,
+      "loss": 0.1546,
+      "step": 400
+    },
+    {
+      "epoch": 3.7,
+      "eval_loss": 2.3457818031311035,
+      "eval_runtime": 64.9252,
+      "eval_samples_per_second": 25.368,
+      "eval_steps_per_second": 1.586,
+      "eval_wer": 0.9995914615463181,
+      "step": 400
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 7.94e-06,
+      "loss": 0.2012,
+      "step": 401
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 7.960000000000002e-06,
+      "loss": 0.2384,
+      "step": 402
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 7.980000000000002e-06,
+      "loss": 0.2397,
+      "step": 403
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1064,
+      "step": 404
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 8.020000000000001e-06,
+      "loss": 0.1589,
+      "step": 405
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 8.040000000000001e-06,
+      "loss": 0.2228,
+      "step": 406
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 8.06e-06,
+      "loss": 0.1624,
+      "step": 407
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 8.08e-06,
+      "loss": 0.147,
+      "step": 408
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 8.1e-06,
+      "loss": 0.204,
+      "step": 409
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 8.120000000000002e-06,
+      "loss": 0.1709,
+      "step": 410
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 8.14e-06,
+      "loss": 0.218,
+      "step": 411
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 8.16e-06,
+      "loss": 0.1203,
+      "step": 412
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 8.18e-06,
+      "loss": 0.1651,
+      "step": 413
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 8.2e-06,
+      "loss": 0.1769,
+      "step": 414
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 8.220000000000001e-06,
+      "loss": 0.2021,
+      "step": 415
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 8.24e-06,
+      "loss": 0.1315,
+      "step": 416
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 8.26e-06,
+      "loss": 0.1554,
+      "step": 417
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 8.28e-06,
+      "loss": 0.1442,
+      "step": 418
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 8.3e-06,
+      "loss": 0.1078,
+      "step": 419
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 8.32e-06,
+      "loss": 0.1799,
+      "step": 420
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 8.34e-06,
+      "loss": 0.1387,
+      "step": 421
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 8.36e-06,
+      "loss": 0.1051,
+      "step": 422
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 8.380000000000001e-06,
+      "loss": 0.1963,
+      "step": 423
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.2608,
+      "step": 424
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 8.42e-06,
+      "loss": 0.1296,
+      "step": 425
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 8.44e-06,
+      "loss": 0.1472,
+      "step": 426
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 8.46e-06,
+      "loss": 0.2663,
+      "step": 427
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 8.48e-06,
+      "loss": 0.1362,
+      "step": 428
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 8.5e-06,
+      "loss": 0.1464,
+      "step": 429
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 8.52e-06,
+      "loss": 0.178,
+      "step": 430
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 8.540000000000001e-06,
+      "loss": 0.2135,
+      "step": 431
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 8.560000000000001e-06,
+      "loss": 0.1421,
+      "step": 432
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 8.580000000000001e-06,
+      "loss": 0.2515,
+      "step": 433
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 8.6e-06,
+      "loss": 0.1618,
+      "step": 434
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 8.62e-06,
+      "loss": 0.1387,
+      "step": 435
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.64e-06,
+      "loss": 0.0955,
+      "step": 436
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 8.66e-06,
+      "loss": 0.1575,
+      "step": 437
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 8.68e-06,
+      "loss": 0.2457,
+      "step": 438
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 8.700000000000001e-06,
+      "loss": 0.1236,
+      "step": 439
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 8.720000000000001e-06,
+      "loss": 0.1447,
+      "step": 440
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 8.740000000000001e-06,
+      "loss": 0.1535,
+      "step": 441
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 8.76e-06,
+      "loss": 0.213,
+      "step": 442
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 8.78e-06,
+      "loss": 0.1313,
+      "step": 443
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 8.8e-06,
+      "loss": 0.1265,
+      "step": 444
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 8.82e-06,
+      "loss": 0.1951,
+      "step": 445
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 8.84e-06,
+      "loss": 0.1277,
+      "step": 446
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 8.860000000000002e-06,
+      "loss": 0.152,
+      "step": 447
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 8.880000000000001e-06,
+      "loss": 0.1727,
+      "step": 448
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 8.900000000000001e-06,
+      "loss": 0.1027,
+      "step": 449
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.920000000000001e-06,
+      "loss": 0.1104,
+      "step": 450
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.94e-06,
+      "loss": 0.1766,
+      "step": 451
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 8.96e-06,
+      "loss": 0.1276,
+      "step": 452
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 8.98e-06,
+      "loss": 0.0728,
+      "step": 453
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 9e-06,
+      "loss": 0.1569,
+      "step": 454
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 9.020000000000002e-06,
+      "loss": 0.1092,
+      "step": 455
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 9.040000000000002e-06,
+      "loss": 0.2114,
+      "step": 456
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 9.060000000000001e-06,
+      "loss": 0.2531,
+      "step": 457
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 9.080000000000001e-06,
+      "loss": 0.1212,
+      "step": 458
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 9.100000000000001e-06,
+      "loss": 0.1435,
+      "step": 459
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 9.12e-06,
+      "loss": 0.1827,
+      "step": 460
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 9.14e-06,
+      "loss": 0.0738,
+      "step": 461
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 9.16e-06,
+      "loss": 0.1331,
+      "step": 462
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 9.180000000000002e-06,
+      "loss": 0.1241,
+      "step": 463
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 0.157,
+      "step": 464
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 9.220000000000002e-06,
+      "loss": 0.0962,
+      "step": 465
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 9.240000000000001e-06,
+      "loss": 0.1807,
+      "step": 466
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 9.260000000000001e-06,
+      "loss": 0.1664,
+      "step": 467
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 9.280000000000001e-06,
+      "loss": 0.1323,
+      "step": 468
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 9.3e-06,
+      "loss": 0.1317,
+      "step": 469
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.32e-06,
+      "loss": 0.1718,
+      "step": 470
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.340000000000002e-06,
+      "loss": 0.1102,
+      "step": 471
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.360000000000002e-06,
+      "loss": 0.1173,
+      "step": 472
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.38e-06,
+      "loss": 0.1248,
+      "step": 473
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.4e-06,
+      "loss": 0.1309,
+      "step": 474
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.42e-06,
+      "loss": 0.1414,
+      "step": 475
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.440000000000001e-06,
+      "loss": 0.1001,
+      "step": 476
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 9.460000000000001e-06,
+      "loss": 0.1669,
+      "step": 477
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 9.48e-06,
+      "loss": 0.0927,
+      "step": 478
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 9.5e-06,
+      "loss": 0.2741,
+      "step": 479
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 9.52e-06,
+      "loss": 0.0995,
+      "step": 480
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 9.54e-06,
+      "loss": 0.1056,
+      "step": 481
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 9.56e-06,
+      "loss": 0.2083,
+      "step": 482
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 9.58e-06,
+      "loss": 0.1214,
+      "step": 483
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 0.1562,
+      "step": 484
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 9.620000000000001e-06,
+      "loss": 0.201,
+      "step": 485
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 9.640000000000001e-06,
+      "loss": 0.1287,
+      "step": 486
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 9.66e-06,
+      "loss": 0.156,
+      "step": 487
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 9.68e-06,
+      "loss": 0.1904,
+      "step": 488
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 9.7e-06,
+      "loss": 0.216,
+      "step": 489
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 9.72e-06,
+      "loss": 0.0923,
+      "step": 490
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 9.74e-06,
+      "loss": 0.1765,
+      "step": 491
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 9.760000000000001e-06,
+      "loss": 0.1185,
+      "step": 492
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 9.780000000000001e-06,
+      "loss": 0.1077,
+      "step": 493
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 0.222,
+      "step": 494
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 9.820000000000001e-06,
+      "loss": 0.2063,
+      "step": 495
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 9.84e-06,
+      "loss": 0.1036,
+      "step": 496
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 9.86e-06,
+      "loss": 0.1516,
+      "step": 497
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 9.88e-06,
+      "loss": 0.1464,
+      "step": 498
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 9.9e-06,
+      "loss": 0.1354,
+      "step": 499
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 9.920000000000002e-06,
+      "loss": 0.1468,
+      "step": 500
+    },
+    {
+      "epoch": 4.63,
+      "eval_loss": 2.0086076259613037,
+      "eval_runtime": 65.2984,
+      "eval_samples_per_second": 25.223,
+      "eval_steps_per_second": 1.577,
+      "eval_wer": 0.9985701154121132,
+      "step": 500
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 9.940000000000001e-06,
+      "loss": 0.1973,
+      "step": 501
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 9.960000000000001e-06,
+      "loss": 0.181,
+      "step": 502
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 9.980000000000001e-06,
+      "loss": 0.1595,
+      "step": 503
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1e-05,
+      "loss": 0.1544,
+      "step": 504
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 9.996350364963505e-06,
+      "loss": 0.1222,
+      "step": 505
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 9.992700729927009e-06,
+      "loss": 0.1163,
+      "step": 506
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 9.989051094890512e-06,
+      "loss": 0.1731,
+      "step": 507
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 9.985401459854015e-06,
+      "loss": 0.1484,
+      "step": 508
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 9.981751824817519e-06,
+      "loss": 0.1385,
+      "step": 509
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 9.978102189781022e-06,
+      "loss": 0.1602,
+      "step": 510
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 9.974452554744526e-06,
+      "loss": 0.1202,
+      "step": 511
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 9.97080291970803e-06,
+      "loss": 0.1131,
+      "step": 512
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 9.967153284671534e-06,
+      "loss": 0.1442,
+      "step": 513
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 9.963503649635036e-06,
+      "loss": 0.141,
+      "step": 514
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 9.959854014598542e-06,
+      "loss": 0.1265,
+      "step": 515
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 9.956204379562044e-06,
+      "loss": 0.1415,
+      "step": 516
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 9.952554744525548e-06,
+      "loss": 0.1398,
+      "step": 517
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 9.948905109489052e-06,
+      "loss": 0.1325,
+      "step": 518
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 9.945255474452556e-06,
+      "loss": 0.2012,
+      "step": 519
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 9.94160583941606e-06,
+      "loss": 0.0914,
+      "step": 520
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 9.937956204379563e-06,
+      "loss": 0.0747,
+      "step": 521
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 9.934306569343066e-06,
+      "loss": 0.1368,
+      "step": 522
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 9.930656934306571e-06,
+      "loss": 0.1267,
+      "step": 523
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 9.927007299270073e-06,
+      "loss": 0.1077,
+      "step": 524
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 9.923357664233577e-06,
+      "loss": 0.1235,
+      "step": 525
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 9.919708029197081e-06,
+      "loss": 0.1163,
+      "step": 526
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 9.916058394160585e-06,
+      "loss": 0.1371,
+      "step": 527
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 9.912408759124087e-06,
+      "loss": 0.1184,
+      "step": 528
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 9.908759124087593e-06,
+      "loss": 0.1019,
+      "step": 529
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 9.905109489051095e-06,
+      "loss": 0.1846,
+      "step": 530
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 9.901459854014599e-06,
+      "loss": 0.1297,
+      "step": 531
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 9.897810218978103e-06,
+      "loss": 0.3406,
+      "step": 532
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 9.894160583941607e-06,
+      "loss": 0.0868,
+      "step": 533
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 9.89051094890511e-06,
+      "loss": 0.1636,
+      "step": 534
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 9.886861313868614e-06,
+      "loss": 0.1496,
+      "step": 535
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 9.883211678832118e-06,
+      "loss": 0.144,
+      "step": 536
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 9.87956204379562e-06,
+      "loss": 0.1355,
+      "step": 537
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 9.875912408759124e-06,
+      "loss": 0.1208,
+      "step": 538
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 9.872262773722628e-06,
+      "loss": 0.1012,
+      "step": 539
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 9.868613138686132e-06,
+      "loss": 0.1445,
+      "step": 540
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 9.864963503649636e-06,
+      "loss": 0.2461,
+      "step": 541
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 9.86131386861314e-06,
+      "loss": 0.1342,
+      "step": 542
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 9.857664233576642e-06,
+      "loss": 0.1429,
+      "step": 543
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 9.854014598540148e-06,
+      "loss": 0.123,
+      "step": 544
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 9.85036496350365e-06,
+      "loss": 0.0723,
+      "step": 545
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 9.846715328467155e-06,
+      "loss": 0.0988,
+      "step": 546
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 9.843065693430658e-06,
+      "loss": 0.1,
+      "step": 547
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 9.839416058394161e-06,
+      "loss": 0.0695,
+      "step": 548
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 9.835766423357665e-06,
+      "loss": 0.1792,
+      "step": 549
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 9.83211678832117e-06,
+      "loss": 0.0863,
+      "step": 550
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 9.828467153284671e-06,
+      "loss": 0.0839,
+      "step": 551
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 9.824817518248177e-06,
+      "loss": 0.1152,
+      "step": 552
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 9.82116788321168e-06,
+      "loss": 0.1549,
+      "step": 553
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 9.817518248175183e-06,
+      "loss": 0.0673,
+      "step": 554
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 9.813868613138687e-06,
+      "loss": 0.0964,
+      "step": 555
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 9.81021897810219e-06,
+      "loss": 0.1415,
+      "step": 556
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 9.806569343065693e-06,
+      "loss": 0.1033,
+      "step": 557
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 9.802919708029199e-06,
+      "loss": 0.1004,
+      "step": 558
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 9.7992700729927e-06,
+      "loss": 0.0655,
+      "step": 559
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 9.795620437956205e-06,
+      "loss": 0.1013,
+      "step": 560
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 9.791970802919709e-06,
+      "loss": 0.1463,
+      "step": 561
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 9.788321167883212e-06,
+      "loss": 0.0961,
+      "step": 562
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 9.784671532846716e-06,
+      "loss": 0.0626,
+      "step": 563
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 9.78102189781022e-06,
+      "loss": 0.1187,
+      "step": 564
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 9.777372262773724e-06,
+      "loss": 0.1438,
+      "step": 565
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 9.773722627737226e-06,
+      "loss": 0.0808,
+      "step": 566
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 9.77007299270073e-06,
+      "loss": 0.1157,
+      "step": 567
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 9.766423357664234e-06,
+      "loss": 0.1638,
+      "step": 568
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 9.762773722627738e-06,
+      "loss": 0.1341,
+      "step": 569
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 9.759124087591242e-06,
+      "loss": 0.1141,
+      "step": 570
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 9.755474452554746e-06,
+      "loss": 0.1425,
+      "step": 571
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 9.75182481751825e-06,
+      "loss": 0.0764,
+      "step": 572
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 9.748175182481753e-06,
+      "loss": 0.167,
+      "step": 573
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 9.744525547445256e-06,
+      "loss": 0.1363,
+      "step": 574
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 9.740875912408761e-06,
+      "loss": 0.1487,
+      "step": 575
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 9.737226277372263e-06,
+      "loss": 0.0709,
+      "step": 576
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 9.733576642335767e-06,
+      "loss": 0.1037,
+      "step": 577
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 9.729927007299271e-06,
+      "loss": 0.121,
+      "step": 578
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 9.726277372262775e-06,
+      "loss": 0.1094,
+      "step": 579
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 9.722627737226277e-06,
+      "loss": 0.1066,
+      "step": 580
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 9.718978102189783e-06,
+      "loss": 0.063,
+      "step": 581
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 9.715328467153285e-06,
+      "loss": 0.0978,
+      "step": 582
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 9.711678832116789e-06,
+      "loss": 0.1209,
+      "step": 583
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 9.708029197080293e-06,
+      "loss": 0.1373,
+      "step": 584
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 9.704379562043797e-06,
+      "loss": 0.0926,
+      "step": 585
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 9.700729927007299e-06,
+      "loss": 0.1004,
+      "step": 586
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 9.697080291970804e-06,
+      "loss": 0.1561,
+      "step": 587
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 9.693430656934307e-06,
+      "loss": 0.0824,
+      "step": 588
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 9.68978102189781e-06,
+      "loss": 0.1049,
+      "step": 589
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 9.686131386861314e-06,
+      "loss": 0.1371,
+      "step": 590
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.682481751824818e-06,
+      "loss": 0.0972,
+      "step": 591
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 9.678832116788322e-06,
+      "loss": 0.1546,
+      "step": 592
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 9.675182481751826e-06,
+      "loss": 0.1979,
+      "step": 593
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 9.67153284671533e-06,
+      "loss": 0.1292,
+      "step": 594
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 9.667883211678832e-06,
+      "loss": 0.1414,
+      "step": 595
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 9.664233576642336e-06,
+      "loss": 0.1481,
+      "step": 596
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 9.66058394160584e-06,
+      "loss": 0.138,
+      "step": 597
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 9.656934306569344e-06,
+      "loss": 0.0957,
+      "step": 598
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 9.653284671532848e-06,
+      "loss": 0.0828,
+      "step": 599
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 9.649635036496352e-06,
+      "loss": 0.1261,
+      "step": 600
+    },
+    {
+      "epoch": 5.55,
+      "eval_loss": 1.826940655708313,
+      "eval_runtime": 64.971,
+      "eval_samples_per_second": 25.35,
+      "eval_steps_per_second": 1.585,
+      "eval_wer": 0.9984679807986927,
+      "step": 600
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 9.645985401459855e-06,
+      "loss": 0.1319,
+      "step": 601
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 9.64233576642336e-06,
+      "loss": 0.0995,
+      "step": 602
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 9.638686131386861e-06,
+      "loss": 0.1321,
+      "step": 603
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 9.635036496350367e-06,
+      "loss": 0.0964,
+      "step": 604
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 9.63138686131387e-06,
+      "loss": 0.122,
+      "step": 605
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 9.627737226277373e-06,
+      "loss": 0.1394,
+      "step": 606
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 9.624087591240877e-06,
+      "loss": 0.1147,
+      "step": 607
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 9.620437956204381e-06,
+      "loss": 0.2518,
+      "step": 608
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 9.616788321167883e-06,
+      "loss": 0.1096,
+      "step": 609
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 9.613138686131389e-06,
+      "loss": 0.122,
+      "step": 610
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 9.60948905109489e-06,
+      "loss": 0.1172,
+      "step": 611
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 9.605839416058395e-06,
+      "loss": 0.1823,
+      "step": 612
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 9.602189781021899e-06,
+      "loss": 0.0926,
+      "step": 613
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 9.598540145985402e-06,
+      "loss": 0.0702,
+      "step": 614
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 9.594890510948905e-06,
+      "loss": 0.1528,
+      "step": 615
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 9.59124087591241e-06,
+      "loss": 0.0844,
+      "step": 616
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 9.587591240875912e-06,
+      "loss": 0.0603,
+      "step": 617
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 9.583941605839416e-06,
+      "loss": 0.1212,
+      "step": 618
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 9.58029197080292e-06,
+      "loss": 0.073,
+      "step": 619
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 9.576642335766424e-06,
+      "loss": 0.1155,
+      "step": 620
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 9.572992700729928e-06,
+      "loss": 0.1236,
+      "step": 621
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 9.569343065693432e-06,
+      "loss": 0.1008,
+      "step": 622
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 9.565693430656936e-06,
+      "loss": 0.0912,
+      "step": 623
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 9.56204379562044e-06,
+      "loss": 0.0897,
+      "step": 624
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 9.558394160583942e-06,
+      "loss": 0.1408,
+      "step": 625
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 9.554744525547446e-06,
+      "loss": 0.1619,
+      "step": 626
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 9.55109489051095e-06,
+      "loss": 0.1845,
+      "step": 627
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 9.547445255474453e-06,
+      "loss": 0.1048,
+      "step": 628
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 9.543795620437957e-06,
+      "loss": 0.0844,
+      "step": 629
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 9.540145985401461e-06,
+      "loss": 0.1434,
+      "step": 630
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 9.536496350364965e-06,
+      "loss": 0.1594,
+      "step": 631
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 9.532846715328467e-06,
+      "loss": 0.0782,
+      "step": 632
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 9.529197080291971e-06,
+      "loss": 0.1405,
+      "step": 633
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 9.525547445255475e-06,
+      "loss": 0.221,
+      "step": 634
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 9.521897810218979e-06,
+      "loss": 0.1235,
+      "step": 635
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 9.518248175182483e-06,
+      "loss": 0.1071,
+      "step": 636
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 9.514598540145987e-06,
+      "loss": 0.1288,
+      "step": 637
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 9.510948905109489e-06,
+      "loss": 0.1124,
+      "step": 638
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 9.507299270072994e-06,
+      "loss": 0.1092,
+      "step": 639
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 9.503649635036497e-06,
+      "loss": 0.0925,
+      "step": 640
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 9.5e-06,
+      "loss": 0.0911,
+      "step": 641
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 9.496350364963504e-06,
+      "loss": 0.1316,
+      "step": 642
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 9.492700729927008e-06,
+      "loss": 0.1489,
+      "step": 643
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 9.48905109489051e-06,
+      "loss": 0.1183,
+      "step": 644
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 9.485401459854016e-06,
+      "loss": 0.0807,
+      "step": 645
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 9.481751824817518e-06,
+      "loss": 0.127,
+      "step": 646
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 9.478102189781022e-06,
+      "loss": 0.0855,
+      "step": 647
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 9.474452554744526e-06,
+      "loss": 0.1027,
+      "step": 648
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 9.47080291970803e-06,
+      "loss": 0.1266,
+      "step": 649
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 9.467153284671534e-06,
+      "loss": 0.0924,
+      "step": 650
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 9.463503649635038e-06,
+      "loss": 0.1717,
+      "step": 651
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 9.459854014598542e-06,
+      "loss": 0.116,
+      "step": 652
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 9.456204379562045e-06,
+      "loss": 0.0947,
+      "step": 653
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 9.452554744525548e-06,
+      "loss": 0.0929,
+      "step": 654
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 9.448905109489051e-06,
+      "loss": 0.0819,
+      "step": 655
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 9.445255474452555e-06,
+      "loss": 0.0832,
+      "step": 656
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 9.44160583941606e-06,
+      "loss": 0.1008,
+      "step": 657
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 9.437956204379563e-06,
+      "loss": 0.0565,
+      "step": 658
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 9.434306569343067e-06,
+      "loss": 0.0752,
+      "step": 659
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 9.430656934306571e-06,
+      "loss": 0.0792,
+      "step": 660
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 9.427007299270073e-06,
+      "loss": 0.0645,
+      "step": 661
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 9.423357664233577e-06,
+      "loss": 0.1156,
+      "step": 662
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 9.419708029197081e-06,
+      "loss": 0.0867,
+      "step": 663
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 9.416058394160585e-06,
+      "loss": 0.0721,
+      "step": 664
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 9.412408759124089e-06,
+      "loss": 0.0829,
+      "step": 665
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 9.408759124087593e-06,
+      "loss": 0.0938,
+      "step": 666
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 9.405109489051095e-06,
+      "loss": 0.2426,
+      "step": 667
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 9.4014598540146e-06,
+      "loss": 0.0923,
+      "step": 668
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 9.397810218978102e-06,
+      "loss": 0.0955,
+      "step": 669
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 9.394160583941606e-06,
+      "loss": 0.1593,
+      "step": 670
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 9.39051094890511e-06,
+      "loss": 0.0822,
+      "step": 671
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 9.386861313868614e-06,
+      "loss": 0.0896,
+      "step": 672
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 9.383211678832116e-06,
+      "loss": 0.0661,
+      "step": 673
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 9.379562043795622e-06,
+      "loss": 0.0607,
+      "step": 674
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 9.375912408759124e-06,
+      "loss": 0.1071,
+      "step": 675
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 9.37226277372263e-06,
+      "loss": 0.1424,
+      "step": 676
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 9.368613138686132e-06,
+      "loss": 0.11,
+      "step": 677
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 9.364963503649636e-06,
+      "loss": 0.1081,
+      "step": 678
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 9.36131386861314e-06,
+      "loss": 0.0932,
+      "step": 679
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 9.357664233576643e-06,
+      "loss": 0.1023,
+      "step": 680
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 9.354014598540146e-06,
+      "loss": 0.0897,
+      "step": 681
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 9.350364963503651e-06,
+      "loss": 0.0925,
+      "step": 682
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 9.346715328467153e-06,
+      "loss": 0.1167,
+      "step": 683
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 9.343065693430657e-06,
+      "loss": 0.0844,
+      "step": 684
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 9.339416058394161e-06,
+      "loss": 0.1301,
+      "step": 685
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 9.335766423357665e-06,
+      "loss": 0.0848,
+      "step": 686
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 9.332116788321169e-06,
+      "loss": 0.0639,
+      "step": 687
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 9.328467153284673e-06,
+      "loss": 0.0818,
+      "step": 688
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 9.324817518248177e-06,
+      "loss": 0.2025,
+      "step": 689
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 9.321167883211679e-06,
+      "loss": 0.0504,
+      "step": 690
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 9.317518248175183e-06,
+      "loss": 0.0586,
+      "step": 691
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 9.313868613138687e-06,
+      "loss": 0.1024,
+      "step": 692
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 9.31021897810219e-06,
+      "loss": 0.0661,
+      "step": 693
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 9.306569343065694e-06,
+      "loss": 0.1393,
+      "step": 694
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 9.302919708029198e-06,
+      "loss": 0.1388,
+      "step": 695
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 9.2992700729927e-06,
+      "loss": 0.0859,
+      "step": 696
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 9.295620437956206e-06,
+      "loss": 0.1003,
+      "step": 697
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 9.291970802919708e-06,
+      "loss": 0.1233,
+      "step": 698
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 9.288321167883212e-06,
+      "loss": 0.0643,
+      "step": 699
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 9.284671532846716e-06,
+      "loss": 0.1206,
+      "step": 700
+    },
+    {
+      "epoch": 6.48,
+      "eval_loss": 1.7346789836883545,
+      "eval_runtime": 65.6764,
+      "eval_samples_per_second": 25.077,
+      "eval_steps_per_second": 1.568,
+      "eval_wer": 0.995608211622919,
+      "step": 700
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 9.28102189781022e-06,
+      "loss": 0.1072,
+      "step": 701
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 9.277372262773724e-06,
+      "loss": 0.1008,
+      "step": 702
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 9.273722627737228e-06,
+      "loss": 0.1181,
+      "step": 703
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 9.27007299270073e-06,
+      "loss": 0.1647,
+      "step": 704
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 9.266423357664235e-06,
+      "loss": 0.1707,
+      "step": 705
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 9.262773722627738e-06,
+      "loss": 0.088,
+      "step": 706
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 9.259124087591242e-06,
+      "loss": 0.1318,
+      "step": 707
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 9.255474452554745e-06,
+      "loss": 0.0797,
+      "step": 708
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 9.25182481751825e-06,
+      "loss": 0.0896,
+      "step": 709
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 9.248175182481751e-06,
+      "loss": 0.1446,
+      "step": 710
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 9.244525547445257e-06,
+      "loss": 0.1235,
+      "step": 711
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 9.24087591240876e-06,
+      "loss": 0.0685,
+      "step": 712
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 9.237226277372263e-06,
+      "loss": 0.0975,
+      "step": 713
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 9.233576642335767e-06,
+      "loss": 0.1656,
+      "step": 714
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 9.229927007299271e-06,
+      "loss": 0.0882,
+      "step": 715
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 9.226277372262775e-06,
+      "loss": 0.1002,
+      "step": 716
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 9.222627737226279e-06,
+      "loss": 0.0961,
+      "step": 717
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 9.218978102189783e-06,
+      "loss": 0.0932,
+      "step": 718
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 9.215328467153285e-06,
+      "loss": 0.0899,
+      "step": 719
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 9.211678832116789e-06,
+      "loss": 0.1378,
+      "step": 720
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.208029197080292e-06,
+      "loss": 0.1192,
+      "step": 721
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.204379562043796e-06,
+      "loss": 0.0714,
+      "step": 722
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.2007299270073e-06,
+      "loss": 0.1715,
+      "step": 723
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.197080291970804e-06,
+      "loss": 0.0517,
+      "step": 724
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.193430656934306e-06,
+      "loss": 0.057,
+      "step": 725
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.189781021897812e-06,
+      "loss": 0.1847,
+      "step": 726
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.186131386861314e-06,
+      "loss": 0.1212,
+      "step": 727
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.18248175182482e-06,
+      "loss": 0.0737,
+      "step": 728
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.178832116788322e-06,
+      "loss": 0.1491,
+      "step": 729
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.175182481751826e-06,
+      "loss": 0.1115,
+      "step": 730
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.17153284671533e-06,
+      "loss": 0.0765,
+      "step": 731
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.167883211678833e-06,
+      "loss": 0.1125,
+      "step": 732
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.164233576642336e-06,
+      "loss": 0.1158,
+      "step": 733
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.160583941605841e-06,
+      "loss": 0.0761,
+      "step": 734
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.156934306569343e-06,
+      "loss": 0.0852,
+      "step": 735
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.153284671532847e-06,
+      "loss": 0.0918,
+      "step": 736
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.149635036496351e-06,
+      "loss": 0.0412,
+      "step": 737
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.145985401459855e-06,
+      "loss": 0.1541,
+      "step": 738
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.142335766423357e-06,
+      "loss": 0.1362,
+      "step": 739
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.138686131386863e-06,
+      "loss": 0.0935,
+      "step": 740
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.135036496350365e-06,
+      "loss": 0.0839,
+      "step": 741
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.131386861313869e-06,
+      "loss": 0.0489,
+      "step": 742
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.127737226277373e-06,
+      "loss": 0.0608,
+      "step": 743
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.124087591240877e-06,
+      "loss": 0.0775,
+      "step": 744
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.12043795620438e-06,
+      "loss": 0.1019,
+      "step": 745
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.116788321167884e-06,
+      "loss": 0.0917,
+      "step": 746
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.113138686131388e-06,
+      "loss": 0.1164,
+      "step": 747
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.10948905109489e-06,
+      "loss": 0.1318,
+      "step": 748
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.105839416058394e-06,
+      "loss": 0.1025,
+      "step": 749
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.102189781021898e-06,
+      "loss": 0.0751,
+      "step": 750
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.098540145985402e-06,
+      "loss": 0.1362,
+      "step": 751
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.094890510948906e-06,
+      "loss": 0.0838,
+      "step": 752
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.09124087591241e-06,
+      "loss": 0.0614,
+      "step": 753
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.087591240875912e-06,
+      "loss": 0.119,
+      "step": 754
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.083941605839418e-06,
+      "loss": 0.084,
+      "step": 755
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.08029197080292e-06,
+      "loss": 0.0982,
+      "step": 756
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 9.076642335766424e-06,
+      "loss": 0.1618,
+      "step": 757
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 9.072992700729928e-06,
+      "loss": 0.0884,
+      "step": 758
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 9.069343065693432e-06,
+      "loss": 0.1251,
+      "step": 759
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 9.065693430656935e-06,
+      "loss": 0.0759,
+      "step": 760
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 9.06204379562044e-06,
+      "loss": 0.0776,
+      "step": 761
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 9.058394160583941e-06,
+      "loss": 0.0999,
+      "step": 762
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 9.054744525547447e-06,
+      "loss": 0.1244,
+      "step": 763
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 9.05109489051095e-06,
+      "loss": 0.1486,
+      "step": 764
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 9.047445255474453e-06,
+      "loss": 0.0934,
+      "step": 765
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 9.043795620437957e-06,
+      "loss": 0.076,
+      "step": 766
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 9.040145985401461e-06,
+      "loss": 0.0852,
+      "step": 767
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 9.036496350364963e-06,
+      "loss": 0.1117,
+      "step": 768
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 9.032846715328469e-06,
+      "loss": 0.0913,
+      "step": 769
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 9.029197080291971e-06,
+      "loss": 0.0741,
+      "step": 770
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 9.025547445255475e-06,
+      "loss": 0.109,
+      "step": 771
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 9.021897810218979e-06,
+      "loss": 0.1141,
+      "step": 772
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 9.018248175182483e-06,
+      "loss": 0.0462,
+      "step": 773
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.014598540145986e-06,
+      "loss": 0.0666,
+      "step": 774
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.01094890510949e-06,
+      "loss": 0.1031,
+      "step": 775
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 9.007299270072994e-06,
+      "loss": 0.0876,
+      "step": 776
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 9.003649635036496e-06,
+      "loss": 0.0527,
+      "step": 777
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 9e-06,
+      "loss": 0.0973,
+      "step": 778
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.996350364963504e-06,
+      "loss": 0.1045,
+      "step": 779
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.992700729927008e-06,
+      "loss": 0.0568,
+      "step": 780
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.989051094890512e-06,
+      "loss": 0.1186,
+      "step": 781
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.985401459854016e-06,
+      "loss": 0.0819,
+      "step": 782
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.98175182481752e-06,
+      "loss": 0.0648,
+      "step": 783
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.978102189781024e-06,
+      "loss": 0.1254,
+      "step": 784
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.974452554744526e-06,
+      "loss": 0.0909,
+      "step": 785
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.97080291970803e-06,
+      "loss": 0.0954,
+      "step": 786
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.967153284671533e-06,
+      "loss": 0.1388,
+      "step": 787
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.963503649635037e-06,
+      "loss": 0.1022,
+      "step": 788
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.959854014598541e-06,
+      "loss": 0.0479,
+      "step": 789
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.956204379562045e-06,
+      "loss": 0.0832,
+      "step": 790
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.952554744525547e-06,
+      "loss": 0.0621,
+      "step": 791
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.948905109489053e-06,
+      "loss": 0.0683,
+      "step": 792
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 8.945255474452555e-06,
+      "loss": 0.1174,
+      "step": 793
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 8.941605839416059e-06,
+      "loss": 0.1344,
+      "step": 794
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 8.937956204379563e-06,
+      "loss": 0.0812,
+      "step": 795
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 8.934306569343067e-06,
+      "loss": 0.1199,
+      "step": 796
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 8.930656934306569e-06,
+      "loss": 0.0931,
+      "step": 797
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 8.927007299270074e-06,
+      "loss": 0.0576,
+      "step": 798
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 8.923357664233577e-06,
+      "loss": 0.0828,
+      "step": 799
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 8.91970802919708e-06,
+      "loss": 0.1959,
+      "step": 800
+    },
+    {
+      "epoch": 7.4,
+      "eval_loss": 1.6818538904190063,
+      "eval_runtime": 64.821,
+      "eval_samples_per_second": 25.408,
+      "eval_steps_per_second": 1.589,
+      "eval_wer": 0.9955060770094986,
+      "step": 800
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 8.916058394160584e-06,
+      "loss": 0.1095,
+      "step": 801
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 8.912408759124088e-06,
+      "loss": 0.0729,
+      "step": 802
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 8.908759124087592e-06,
+      "loss": 0.0852,
+      "step": 803
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 8.905109489051096e-06,
+      "loss": 0.0599,
+      "step": 804
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 8.901459854014598e-06,
+      "loss": 0.0665,
+      "step": 805
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 8.897810218978102e-06,
+      "loss": 0.0869,
+      "step": 806
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 8.894160583941606e-06,
+      "loss": 0.0318,
+      "step": 807
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 8.89051094890511e-06,
+      "loss": 0.0571,
+      "step": 808
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 8.886861313868614e-06,
+      "loss": 0.1026,
+      "step": 809
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 8.883211678832118e-06,
+      "loss": 0.0598,
+      "step": 810
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 8.879562043795622e-06,
+      "loss": 0.075,
+      "step": 811
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 8.875912408759125e-06,
+      "loss": 0.0861,
+      "step": 812
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 8.87226277372263e-06,
+      "loss": 0.0741,
+      "step": 813
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 8.868613138686132e-06,
+      "loss": 0.0558,
+      "step": 814
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 8.864963503649635e-06,
+      "loss": 0.0787,
+      "step": 815
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 8.86131386861314e-06,
+      "loss": 0.1013,
+      "step": 816
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 8.857664233576643e-06,
+      "loss": 0.0463,
+      "step": 817
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 8.854014598540147e-06,
+      "loss": 0.0988,
+      "step": 818
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 8.850364963503651e-06,
+      "loss": 0.1498,
+      "step": 819
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 8.846715328467153e-06,
+      "loss": 0.0895,
+      "step": 820
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 8.843065693430659e-06,
+      "loss": 0.1315,
+      "step": 821
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 8.839416058394161e-06,
+      "loss": 0.0487,
+      "step": 822
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 8.835766423357665e-06,
+      "loss": 0.0581,
+      "step": 823
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 8.832116788321169e-06,
+      "loss": 0.0952,
+      "step": 824
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 8.828467153284673e-06,
+      "loss": 0.0946,
+      "step": 825
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 8.824817518248175e-06,
+      "loss": 0.0745,
+      "step": 826
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 8.82116788321168e-06,
+      "loss": 0.1071,
+      "step": 827
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 8.817518248175182e-06,
+      "loss": 0.1084,
+      "step": 828
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 8.813868613138686e-06,
+      "loss": 0.0598,
+      "step": 829
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 8.81021897810219e-06,
+      "loss": 0.0838,
+      "step": 830
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 8.806569343065694e-06,
+      "loss": 0.0944,
+      "step": 831
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 8.802919708029198e-06,
+      "loss": 0.047,
+      "step": 832
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 8.799270072992702e-06,
+      "loss": 0.0979,
+      "step": 833
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 8.795620437956204e-06,
+      "loss": 0.0671,
+      "step": 834
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 8.79197080291971e-06,
+      "loss": 0.0475,
+      "step": 835
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 8.788321167883212e-06,
+      "loss": 0.0629,
+      "step": 836
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 8.784671532846716e-06,
+      "loss": 0.1038,
+      "step": 837
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 8.78102189781022e-06,
+      "loss": 0.082,
+      "step": 838
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 8.777372262773724e-06,
+      "loss": 0.0729,
+      "step": 839
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 8.773722627737227e-06,
+      "loss": 0.1469,
+      "step": 840
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 8.770072992700731e-06,
+      "loss": 0.1673,
+      "step": 841
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 8.766423357664235e-06,
+      "loss": 0.0653,
+      "step": 842
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 8.762773722627737e-06,
+      "loss": 0.139,
+      "step": 843
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 8.759124087591241e-06,
+      "loss": 0.1067,
+      "step": 844
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 8.755474452554745e-06,
+      "loss": 0.0734,
+      "step": 845
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 8.751824817518249e-06,
+      "loss": 0.0464,
+      "step": 846
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 8.748175182481753e-06,
+      "loss": 0.0719,
+      "step": 847
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 8.744525547445257e-06,
+      "loss": 0.0638,
+      "step": 848
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 8.740875912408759e-06,
+      "loss": 0.0958,
+      "step": 849
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 8.737226277372265e-06,
+      "loss": 0.1166,
+      "step": 850
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 8.733576642335767e-06,
+      "loss": 0.0975,
+      "step": 851
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 8.72992700729927e-06,
+      "loss": 0.0584,
+      "step": 852
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 8.726277372262774e-06,
+      "loss": 0.0766,
+      "step": 853
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 8.722627737226278e-06,
+      "loss": 0.0666,
+      "step": 854
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 8.71897810218978e-06,
+      "loss": 0.0749,
+      "step": 855
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 8.715328467153286e-06,
+      "loss": 0.1479,
+      "step": 856
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 8.711678832116788e-06,
+      "loss": 0.052,
+      "step": 857
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 8.708029197080292e-06,
+      "loss": 0.0389,
+      "step": 858
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 8.704379562043796e-06,
+      "loss": 0.0803,
+      "step": 859
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 8.7007299270073e-06,
+      "loss": 0.0983,
+      "step": 860
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 8.697080291970804e-06,
+      "loss": 0.0507,
+      "step": 861
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 8.693430656934308e-06,
+      "loss": 0.1299,
+      "step": 862
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 8.68978102189781e-06,
+      "loss": 0.1417,
+      "step": 863
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 8.686131386861315e-06,
+      "loss": 0.0539,
+      "step": 864
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 8.682481751824818e-06,
+      "loss": 0.1468,
+      "step": 865
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 8.678832116788322e-06,
+      "loss": 0.0296,
+      "step": 866
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 8.675182481751825e-06,
+      "loss": 0.0815,
+      "step": 867
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 8.67153284671533e-06,
+      "loss": 0.0591,
+      "step": 868
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 8.667883211678833e-06,
+      "loss": 0.0766,
+      "step": 869
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 8.664233576642337e-06,
+      "loss": 0.0928,
+      "step": 870
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 8.660583941605841e-06,
+      "loss": 0.1465,
+      "step": 871
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 8.656934306569343e-06,
+      "loss": 0.0706,
+      "step": 872
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 8.653284671532847e-06,
+      "loss": 0.0915,
+      "step": 873
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 8.649635036496351e-06,
+      "loss": 0.0814,
+      "step": 874
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 8.645985401459855e-06,
+      "loss": 0.0731,
+      "step": 875
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 8.642335766423359e-06,
+      "loss": 0.1264,
+      "step": 876
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 8.638686131386863e-06,
+      "loss": 0.1084,
+      "step": 877
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 8.635036496350365e-06,
+      "loss": 0.0425,
+      "step": 878
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 8.63138686131387e-06,
+      "loss": 0.0594,
+      "step": 879
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 8.627737226277373e-06,
+      "loss": 0.0898,
+      "step": 880
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 8.624087591240876e-06,
+      "loss": 0.0436,
+      "step": 881
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 8.62043795620438e-06,
+      "loss": 0.0521,
+      "step": 882
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 8.616788321167884e-06,
+      "loss": 0.0967,
+      "step": 883
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 8.613138686131386e-06,
+      "loss": 0.0664,
+      "step": 884
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 8.609489051094892e-06,
+      "loss": 0.1049,
+      "step": 885
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 8.605839416058394e-06,
+      "loss": 0.1063,
+      "step": 886
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 8.6021897810219e-06,
+      "loss": 0.0398,
+      "step": 887
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 8.598540145985402e-06,
+      "loss": 0.063,
+      "step": 888
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 8.594890510948906e-06,
+      "loss": 0.1001,
+      "step": 889
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 8.59124087591241e-06,
+      "loss": 0.0613,
+      "step": 890
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 8.587591240875914e-06,
+      "loss": 0.0746,
+      "step": 891
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 8.583941605839416e-06,
+      "loss": 0.1067,
+      "step": 892
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 8.580291970802921e-06,
+      "loss": 0.0637,
+      "step": 893
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 8.576642335766423e-06,
+      "loss": 0.0815,
+      "step": 894
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 8.572992700729927e-06,
+      "loss": 0.0779,
+      "step": 895
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 8.569343065693431e-06,
+      "loss": 0.0737,
+      "step": 896
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 8.565693430656935e-06,
+      "loss": 0.0582,
+      "step": 897
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 8.562043795620439e-06,
+      "loss": 0.0872,
+      "step": 898
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 8.558394160583943e-06,
+      "loss": 0.0732,
+      "step": 899
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 8.554744525547447e-06,
+      "loss": 0.0502,
+      "step": 900
+    },
+    {
+      "epoch": 8.33,
+      "eval_loss": 1.6809108257293701,
+      "eval_runtime": 65.0229,
+      "eval_samples_per_second": 25.33,
+      "eval_steps_per_second": 1.584,
+      "eval_wer": 0.9965274231437034,
+      "step": 900
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 8.551094890510949e-06,
+      "loss": 0.0473,
+      "step": 901
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 8.547445255474453e-06,
+      "loss": 0.0677,
+      "step": 902
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 8.543795620437957e-06,
+      "loss": 0.0891,
+      "step": 903
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 8.54014598540146e-06,
+      "loss": 0.0348,
+      "step": 904
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 8.536496350364964e-06,
+      "loss": 0.1448,
+      "step": 905
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 8.532846715328468e-06,
+      "loss": 0.0596,
+      "step": 906
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 8.52919708029197e-06,
+      "loss": 0.1381,
+      "step": 907
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 8.525547445255476e-06,
+      "loss": 0.141,
+      "step": 908
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 8.521897810218978e-06,
+      "loss": 0.0611,
+      "step": 909
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 8.518248175182482e-06,
+      "loss": 0.1102,
+      "step": 910
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 8.514598540145986e-06,
+      "loss": 0.1001,
+      "step": 911
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 8.51094890510949e-06,
+      "loss": 0.0972,
+      "step": 912
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 8.507299270072994e-06,
+      "loss": 0.062,
+      "step": 913
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 8.503649635036498e-06,
+      "loss": 0.0733,
+      "step": 914
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 8.5e-06,
+      "loss": 0.0369,
+      "step": 915
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 8.496350364963506e-06,
+      "loss": 0.0898,
+      "step": 916
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 8.492700729927008e-06,
+      "loss": 0.0765,
+      "step": 917
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 8.489051094890512e-06,
+      "loss": 0.066,
+      "step": 918
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 8.485401459854015e-06,
+      "loss": 0.0877,
+      "step": 919
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 8.48175182481752e-06,
+      "loss": 0.131,
+      "step": 920
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 8.478102189781022e-06,
+      "loss": 0.076,
+      "step": 921
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 8.474452554744527e-06,
+      "loss": 0.0719,
+      "step": 922
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 8.47080291970803e-06,
+      "loss": 0.0989,
+      "step": 923
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 8.467153284671533e-06,
+      "loss": 0.0621,
+      "step": 924
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 8.463503649635037e-06,
+      "loss": 0.0834,
+      "step": 925
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 8.459854014598541e-06,
+      "loss": 0.0907,
+      "step": 926
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 8.456204379562045e-06,
+      "loss": 0.0665,
+      "step": 927
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 8.452554744525549e-06,
+      "loss": 0.0606,
+      "step": 928
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 8.448905109489051e-06,
+      "loss": 0.0749,
+      "step": 929
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 8.445255474452555e-06,
+      "loss": 0.1378,
+      "step": 930
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 8.441605839416059e-06,
+      "loss": 0.0317,
+      "step": 931
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 8.437956204379563e-06,
+      "loss": 0.0605,
+      "step": 932
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 8.434306569343066e-06,
+      "loss": 0.1206,
+      "step": 933
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 8.43065693430657e-06,
+      "loss": 0.0749,
+      "step": 934
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 8.427007299270074e-06,
+      "loss": 0.0651,
+      "step": 935
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 8.423357664233576e-06,
+      "loss": 0.1073,
+      "step": 936
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 8.419708029197082e-06,
+      "loss": 0.0611,
+      "step": 937
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 8.416058394160584e-06,
+      "loss": 0.0708,
+      "step": 938
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 8.412408759124088e-06,
+      "loss": 0.0743,
+      "step": 939
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 8.408759124087592e-06,
+      "loss": 0.047,
+      "step": 940
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 8.405109489051096e-06,
+      "loss": 0.0788,
+      "step": 941
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 8.4014598540146e-06,
+      "loss": 0.1528,
+      "step": 942
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 8.397810218978104e-06,
+      "loss": 0.0849,
+      "step": 943
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 8.394160583941606e-06,
+      "loss": 0.0485,
+      "step": 944
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 8.390510948905111e-06,
+      "loss": 0.1241,
+      "step": 945
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 8.386861313868614e-06,
+      "loss": 0.0889,
+      "step": 946
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 8.383211678832117e-06,
+      "loss": 0.0237,
+      "step": 947
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 8.379562043795621e-06,
+      "loss": 0.0573,
+      "step": 948
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 8.375912408759125e-06,
+      "loss": 0.0786,
+      "step": 949
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 8.372262773722627e-06,
+      "loss": 0.0587,
+      "step": 950
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 8.368613138686133e-06,
+      "loss": 0.1135,
+      "step": 951
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 8.364963503649635e-06,
+      "loss": 0.066,
+      "step": 952
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 8.361313868613139e-06,
+      "loss": 0.0866,
+      "step": 953
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 8.357664233576643e-06,
+      "loss": 0.0794,
+      "step": 954
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 8.354014598540147e-06,
+      "loss": 0.079,
+      "step": 955
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 8.35036496350365e-06,
+      "loss": 0.054,
+      "step": 956
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 8.346715328467155e-06,
+      "loss": 0.0729,
+      "step": 957
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 8.343065693430657e-06,
+      "loss": 0.0899,
+      "step": 958
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 8.33941605839416e-06,
+      "loss": 0.0552,
+      "step": 959
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 8.335766423357664e-06,
+      "loss": 0.1034,
+      "step": 960
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 8.332116788321168e-06,
+      "loss": 0.1576,
+      "step": 961
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 8.328467153284672e-06,
+      "loss": 0.0899,
+      "step": 962
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 8.324817518248176e-06,
+      "loss": 0.0482,
+      "step": 963
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 8.32116788321168e-06,
+      "loss": 0.0989,
+      "step": 964
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 8.317518248175182e-06,
+      "loss": 0.0728,
+      "step": 965
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 8.313868613138688e-06,
+      "loss": 0.0893,
+      "step": 966
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 8.31021897810219e-06,
+      "loss": 0.0577,
+      "step": 967
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 8.306569343065694e-06,
+      "loss": 0.0519,
+      "step": 968
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 8.302919708029198e-06,
+      "loss": 0.0829,
+      "step": 969
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 8.299270072992702e-06,
+      "loss": 0.1227,
+      "step": 970
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 8.295620437956205e-06,
+      "loss": 0.1193,
+      "step": 971
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 8.29197080291971e-06,
+      "loss": 0.0828,
+      "step": 972
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 8.288321167883212e-06,
+      "loss": 0.1809,
+      "step": 973
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 8.284671532846717e-06,
+      "loss": 0.0614,
+      "step": 974
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 8.28102189781022e-06,
+      "loss": 0.1069,
+      "step": 975
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 8.277372262773723e-06,
+      "loss": 0.0823,
+      "step": 976
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 8.273722627737227e-06,
+      "loss": 0.0416,
+      "step": 977
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 8.270072992700731e-06,
+      "loss": 0.0783,
+      "step": 978
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 8.266423357664233e-06,
+      "loss": 0.0545,
+      "step": 979
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 8.262773722627739e-06,
+      "loss": 0.0722,
+      "step": 980
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 8.259124087591241e-06,
+      "loss": 0.1115,
+      "step": 981
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 8.255474452554745e-06,
+      "loss": 0.0689,
+      "step": 982
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 8.251824817518249e-06,
+      "loss": 0.0336,
+      "step": 983
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.248175182481753e-06,
+      "loss": 0.0733,
+      "step": 984
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 8.244525547445256e-06,
+      "loss": 0.0946,
+      "step": 985
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.24087591240876e-06,
+      "loss": 0.0598,
+      "step": 986
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.237226277372263e-06,
+      "loss": 0.0614,
+      "step": 987
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.233576642335766e-06,
+      "loss": 0.1032,
+      "step": 988
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 8.22992700729927e-06,
+      "loss": 0.0339,
+      "step": 989
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.226277372262774e-06,
+      "loss": 0.0503,
+      "step": 990
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.222627737226278e-06,
+      "loss": 0.0987,
+      "step": 991
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.218978102189782e-06,
+      "loss": 0.0431,
+      "step": 992
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 8.215328467153286e-06,
+      "loss": 0.08,
+      "step": 993
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 8.21167883211679e-06,
+      "loss": 0.0799,
+      "step": 994
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 8.208029197080294e-06,
+      "loss": 0.0534,
+      "step": 995
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 8.204379562043796e-06,
+      "loss": 0.0256,
+      "step": 996
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 8.2007299270073e-06,
+      "loss": 0.0733,
+      "step": 997
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 8.197080291970804e-06,
+      "loss": 0.0439,
+      "step": 998
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 8.193430656934307e-06,
+      "loss": 0.0752,
+      "step": 999
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 8.189781021897811e-06,
+      "loss": 0.0811,
+      "step": 1000
+    },
+    {
+      "epoch": 9.26,
+      "eval_loss": 1.6673643589019775,
+      "eval_runtime": 65.4446,
+      "eval_samples_per_second": 25.166,
+      "eval_steps_per_second": 1.574,
+      "eval_wer": 0.99162496169952,
+      "step": 1000
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 8.186131386861315e-06,
+      "loss": 0.026,
+      "step": 1001
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 8.182481751824817e-06,
+      "loss": 0.0972,
+      "step": 1002
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 8.178832116788323e-06,
+      "loss": 0.1107,
+      "step": 1003
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 8.175182481751825e-06,
+      "loss": 0.058,
+      "step": 1004
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 8.171532846715329e-06,
+      "loss": 0.0243,
+      "step": 1005
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 8.167883211678833e-06,
+      "loss": 0.0915,
+      "step": 1006
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 8.164233576642337e-06,
+      "loss": 0.1198,
+      "step": 1007
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 8.160583941605839e-06,
+      "loss": 0.0492,
+      "step": 1008
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 8.156934306569345e-06,
+      "loss": 0.0779,
+      "step": 1009
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 8.153284671532847e-06,
+      "loss": 0.0616,
+      "step": 1010
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 8.14963503649635e-06,
+      "loss": 0.0278,
+      "step": 1011
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 8.145985401459855e-06,
+      "loss": 0.0687,
+      "step": 1012
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 8.142335766423358e-06,
+      "loss": 0.0511,
+      "step": 1013
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 8.138686131386862e-06,
+      "loss": 0.0556,
+      "step": 1014
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 8.135036496350366e-06,
+      "loss": 0.0847,
+      "step": 1015
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 8.131386861313868e-06,
+      "loss": 0.1104,
+      "step": 1016
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 8.127737226277372e-06,
+      "loss": 0.041,
+      "step": 1017
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 8.124087591240876e-06,
+      "loss": 0.0899,
+      "step": 1018
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 8.12043795620438e-06,
+      "loss": 0.1109,
+      "step": 1019
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 8.116788321167884e-06,
+      "loss": 0.0523,
+      "step": 1020
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 8.113138686131388e-06,
+      "loss": 0.0577,
+      "step": 1021
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 8.109489051094892e-06,
+      "loss": 0.1438,
+      "step": 1022
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 8.105839416058396e-06,
+      "loss": 0.0485,
+      "step": 1023
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 8.1021897810219e-06,
+      "loss": 0.0355,
+      "step": 1024
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 8.098540145985402e-06,
+      "loss": 0.0652,
+      "step": 1025
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 8.094890510948905e-06,
+      "loss": 0.0312,
+      "step": 1026
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 8.09124087591241e-06,
+      "loss": 0.046,
+      "step": 1027
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 8.087591240875913e-06,
+      "loss": 0.0998,
+      "step": 1028
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 8.083941605839417e-06,
+      "loss": 0.0604,
+      "step": 1029
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 8.080291970802921e-06,
+      "loss": 0.0387,
+      "step": 1030
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 8.076642335766423e-06,
+      "loss": 0.0669,
+      "step": 1031
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 8.072992700729929e-06,
+      "loss": 0.061,
+      "step": 1032
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 8.069343065693431e-06,
+      "loss": 0.042,
+      "step": 1033
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 8.065693430656935e-06,
+      "loss": 0.0494,
+      "step": 1034
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 8.062043795620439e-06,
+      "loss": 0.1188,
+      "step": 1035
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 8.058394160583943e-06,
+      "loss": 0.0678,
+      "step": 1036
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 8.054744525547445e-06,
+      "loss": 0.0757,
+      "step": 1037
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 8.05109489051095e-06,
+      "loss": 0.0562,
+      "step": 1038
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 8.047445255474453e-06,
+      "loss": 0.0544,
+      "step": 1039
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 8.043795620437956e-06,
+      "loss": 0.0426,
+      "step": 1040
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 8.04014598540146e-06,
+      "loss": 0.0755,
+      "step": 1041
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 8.036496350364964e-06,
+      "loss": 0.038,
+      "step": 1042
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 8.032846715328468e-06,
+      "loss": 0.118,
+      "step": 1043
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 8.029197080291972e-06,
+      "loss": 0.0636,
+      "step": 1044
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 8.025547445255474e-06,
+      "loss": 0.0368,
+      "step": 1045
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 8.02189781021898e-06,
+      "loss": 0.0406,
+      "step": 1046
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 8.018248175182482e-06,
+      "loss": 0.1383,
+      "step": 1047
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.014598540145986e-06,
+      "loss": 0.0571,
+      "step": 1048
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.01094890510949e-06,
+      "loss": 0.0957,
+      "step": 1049
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.007299270072994e-06,
+      "loss": 0.0969,
+      "step": 1050
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.003649635036497e-06,
+      "loss": 0.0604,
+      "step": 1051
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.0549,
+      "step": 1052
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.996350364963504e-06,
+      "loss": 0.0783,
+      "step": 1053
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.992700729927007e-06,
+      "loss": 0.0783,
+      "step": 1054
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.989051094890511e-06,
+      "loss": 0.0382,
+      "step": 1055
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.985401459854015e-06,
+      "loss": 0.0892,
+      "step": 1056
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 7.981751824817519e-06,
+      "loss": 0.085,
+      "step": 1057
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 7.978102189781023e-06,
+      "loss": 0.0438,
+      "step": 1058
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 7.974452554744527e-06,
+      "loss": 0.1017,
+      "step": 1059
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 7.970802919708029e-06,
+      "loss": 0.0737,
+      "step": 1060
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 7.967153284671535e-06,
+      "loss": 0.0768,
+      "step": 1061
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 7.963503649635037e-06,
+      "loss": 0.0648,
+      "step": 1062
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 7.95985401459854e-06,
+      "loss": 0.0579,
+      "step": 1063
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 7.956204379562045e-06,
+      "loss": 0.058,
+      "step": 1064
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 7.952554744525548e-06,
+      "loss": 0.0753,
+      "step": 1065
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 7.94890510948905e-06,
+      "loss": 0.0932,
+      "step": 1066
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 7.945255474452556e-06,
+      "loss": 0.0552,
+      "step": 1067
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 7.941605839416058e-06,
+      "loss": 0.0937,
+      "step": 1068
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 7.937956204379562e-06,
+      "loss": 0.1135,
+      "step": 1069
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 7.934306569343066e-06,
+      "loss": 0.0364,
+      "step": 1070
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 7.93065693430657e-06,
+      "loss": 0.0528,
+      "step": 1071
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 7.927007299270074e-06,
+      "loss": 0.0723,
+      "step": 1072
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 7.923357664233578e-06,
+      "loss": 0.041,
+      "step": 1073
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 7.91970802919708e-06,
+      "loss": 0.0624,
+      "step": 1074
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 7.916058394160586e-06,
+      "loss": 0.0262,
+      "step": 1075
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 7.912408759124088e-06,
+      "loss": 0.0672,
+      "step": 1076
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.908759124087592e-06,
+      "loss": 0.066,
+      "step": 1077
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.905109489051095e-06,
+      "loss": 0.0715,
+      "step": 1078
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.9014598540146e-06,
+      "loss": 0.0544,
+      "step": 1079
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 7.897810218978103e-06,
+      "loss": 0.0464,
+      "step": 1080
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 7.894160583941607e-06,
+      "loss": 0.1154,
+      "step": 1081
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 7.89051094890511e-06,
+      "loss": 0.0532,
+      "step": 1082
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 7.886861313868613e-06,
+      "loss": 0.1262,
+      "step": 1083
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 7.883211678832117e-06,
+      "loss": 0.0679,
+      "step": 1084
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 7.879562043795621e-06,
+      "loss": 0.0617,
+      "step": 1085
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 7.875912408759125e-06,
+      "loss": 0.067,
+      "step": 1086
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 7.872262773722629e-06,
+      "loss": 0.0552,
+      "step": 1087
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 7.868613138686133e-06,
+      "loss": 0.0493,
+      "step": 1088
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 7.864963503649635e-06,
+      "loss": 0.0307,
+      "step": 1089
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 7.86131386861314e-06,
+      "loss": 0.0771,
+      "step": 1090
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 7.857664233576643e-06,
+      "loss": 0.0292,
+      "step": 1091
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 7.854014598540146e-06,
+      "loss": 0.0672,
+      "step": 1092
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 7.85036496350365e-06,
+      "loss": 0.0225,
+      "step": 1093
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 7.846715328467154e-06,
+      "loss": 0.0508,
+      "step": 1094
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 7.843065693430656e-06,
+      "loss": 0.0625,
+      "step": 1095
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 7.839416058394162e-06,
+      "loss": 0.0579,
+      "step": 1096
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 7.835766423357664e-06,
+      "loss": 0.0461,
+      "step": 1097
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 7.83211678832117e-06,
+      "loss": 0.0553,
+      "step": 1098
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 7.828467153284672e-06,
+      "loss": 0.0662,
+      "step": 1099
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 7.824817518248176e-06,
+      "loss": 0.0534,
+      "step": 1100
+    },
+    {
+      "epoch": 10.18,
+      "eval_loss": 1.5718997716903687,
+      "eval_runtime": 65.3076,
+      "eval_samples_per_second": 25.219,
+      "eval_steps_per_second": 1.577,
+      "eval_wer": 0.9897865386579512,
+      "step": 1100
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 7.82116788321168e-06,
+      "loss": 0.0671,
+      "step": 1101
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 7.817518248175184e-06,
+      "loss": 0.0903,
+      "step": 1102
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 7.813868613138686e-06,
+      "loss": 0.0426,
+      "step": 1103
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 7.810218978102191e-06,
+      "loss": 0.0603,
+      "step": 1104
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 7.806569343065694e-06,
+      "loss": 0.094,
+      "step": 1105
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 7.802919708029197e-06,
+      "loss": 0.036,
+      "step": 1106
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 7.799270072992701e-06,
+      "loss": 0.0828,
+      "step": 1107
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 7.795620437956205e-06,
+      "loss": 0.1342,
+      "step": 1108
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 7.791970802919709e-06,
+      "loss": 0.0588,
+      "step": 1109
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 7.788321167883213e-06,
+      "loss": 0.0465,
+      "step": 1110
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 7.784671532846715e-06,
+      "loss": 0.1089,
+      "step": 1111
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 7.781021897810219e-06,
+      "loss": 0.0403,
+      "step": 1112
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 7.777372262773723e-06,
+      "loss": 0.0557,
+      "step": 1113
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 7.773722627737227e-06,
+      "loss": 0.1062,
+      "step": 1114
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 7.77007299270073e-06,
+      "loss": 0.0565,
+      "step": 1115
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 7.766423357664235e-06,
+      "loss": 0.0328,
+      "step": 1116
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 7.762773722627738e-06,
+      "loss": 0.0546,
+      "step": 1117
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 7.75912408759124e-06,
+      "loss": 0.0635,
+      "step": 1118
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 7.755474452554746e-06,
+      "loss": 0.046,
+      "step": 1119
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 7.751824817518248e-06,
+      "loss": 0.0576,
+      "step": 1120
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 7.748175182481752e-06,
+      "loss": 0.0344,
+      "step": 1121
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 7.744525547445256e-06,
+      "loss": 0.0656,
+      "step": 1122
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 7.74087591240876e-06,
+      "loss": 0.0294,
+      "step": 1123
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 7.737226277372264e-06,
+      "loss": 0.1109,
+      "step": 1124
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 7.733576642335768e-06,
+      "loss": 0.0648,
+      "step": 1125
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 7.72992700729927e-06,
+      "loss": 0.0727,
+      "step": 1126
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 7.726277372262776e-06,
+      "loss": 0.0897,
+      "step": 1127
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 7.722627737226278e-06,
+      "loss": 0.0741,
+      "step": 1128
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 7.718978102189782e-06,
+      "loss": 0.0284,
+      "step": 1129
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 7.715328467153286e-06,
+      "loss": 0.0766,
+      "step": 1130
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 7.71167883211679e-06,
+      "loss": 0.0582,
+      "step": 1131
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 7.708029197080292e-06,
+      "loss": 0.0512,
+      "step": 1132
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 7.704379562043797e-06,
+      "loss": 0.1094,
+      "step": 1133
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 7.7007299270073e-06,
+      "loss": 0.0832,
+      "step": 1134
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 7.697080291970803e-06,
+      "loss": 0.0527,
+      "step": 1135
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 7.693430656934307e-06,
+      "loss": 0.0517,
+      "step": 1136
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 7.689781021897811e-06,
+      "loss": 0.0474,
+      "step": 1137
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 7.686131386861315e-06,
+      "loss": 0.0418,
+      "step": 1138
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 7.682481751824819e-06,
+      "loss": 0.0523,
+      "step": 1139
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 7.678832116788321e-06,
+      "loss": 0.066,
+      "step": 1140
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 7.675182481751825e-06,
+      "loss": 0.0418,
+      "step": 1141
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 7.671532846715329e-06,
+      "loss": 0.064,
+      "step": 1142
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 7.667883211678833e-06,
+      "loss": 0.0387,
+      "step": 1143
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 7.664233576642336e-06,
+      "loss": 0.0473,
+      "step": 1144
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 7.66058394160584e-06,
+      "loss": 0.0923,
+      "step": 1145
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 7.656934306569344e-06,
+      "loss": 0.0668,
+      "step": 1146
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 7.653284671532846e-06,
+      "loss": 0.0557,
+      "step": 1147
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 7.649635036496352e-06,
+      "loss": 0.0347,
+      "step": 1148
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 7.645985401459854e-06,
+      "loss": 0.1046,
+      "step": 1149
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 7.642335766423358e-06,
+      "loss": 0.0415,
+      "step": 1150
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 7.638686131386862e-06,
+      "loss": 0.0541,
+      "step": 1151
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 7.635036496350366e-06,
+      "loss": 0.0672,
+      "step": 1152
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 7.63138686131387e-06,
+      "loss": 0.0459,
+      "step": 1153
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 7.627737226277373e-06,
+      "loss": 0.0529,
+      "step": 1154
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 7.624087591240877e-06,
+      "loss": 0.0709,
+      "step": 1155
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 7.6204379562043805e-06,
+      "loss": 0.0199,
+      "step": 1156
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 7.6167883211678836e-06,
+      "loss": 0.0528,
+      "step": 1157
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 7.6131386861313875e-06,
+      "loss": 0.1152,
+      "step": 1158
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 7.609489051094891e-06,
+      "loss": 0.0464,
+      "step": 1159
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 7.605839416058395e-06,
+      "loss": 0.0475,
+      "step": 1160
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 7.602189781021898e-06,
+      "loss": 0.0849,
+      "step": 1161
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 7.598540145985402e-06,
+      "loss": 0.0768,
+      "step": 1162
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 7.594890510948906e-06,
+      "loss": 0.0693,
+      "step": 1163
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 7.591240875912409e-06,
+      "loss": 0.0937,
+      "step": 1164
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 7.587591240875914e-06,
+      "loss": 0.0592,
+      "step": 1165
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 7.583941605839417e-06,
+      "loss": 0.0437,
+      "step": 1166
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 7.58029197080292e-06,
+      "loss": 0.0725,
+      "step": 1167
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 7.576642335766425e-06,
+      "loss": 0.0977,
+      "step": 1168
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 7.572992700729928e-06,
+      "loss": 0.0429,
+      "step": 1169
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 7.569343065693431e-06,
+      "loss": 0.0384,
+      "step": 1170
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 7.565693430656935e-06,
+      "loss": 0.1211,
+      "step": 1171
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 7.5620437956204384e-06,
+      "loss": 0.0405,
+      "step": 1172
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 7.5583941605839415e-06,
+      "loss": 0.0514,
+      "step": 1173
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 7.554744525547446e-06,
+      "loss": 0.0605,
+      "step": 1174
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 7.551094890510949e-06,
+      "loss": 0.0609,
+      "step": 1175
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 7.547445255474454e-06,
+      "loss": 0.1047,
+      "step": 1176
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 7.543795620437957e-06,
+      "loss": 0.1114,
+      "step": 1177
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 7.54014598540146e-06,
+      "loss": 0.0384,
+      "step": 1178
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 7.536496350364965e-06,
+      "loss": 0.1056,
+      "step": 1179
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 7.532846715328468e-06,
+      "loss": 0.0561,
+      "step": 1180
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 7.529197080291971e-06,
+      "loss": 0.0546,
+      "step": 1181
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 7.5255474452554756e-06,
+      "loss": 0.053,
+      "step": 1182
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 7.521897810218979e-06,
+      "loss": 0.1487,
+      "step": 1183
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 7.5182481751824825e-06,
+      "loss": 0.061,
+      "step": 1184
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 7.514598540145986e-06,
+      "loss": 0.0411,
+      "step": 1185
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 7.510948905109489e-06,
+      "loss": 0.1118,
+      "step": 1186
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 7.507299270072993e-06,
+      "loss": 0.0357,
+      "step": 1187
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 7.503649635036497e-06,
+      "loss": 0.0581,
+      "step": 1188
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.0793,
+      "step": 1189
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 7.496350364963504e-06,
+      "loss": 0.0325,
+      "step": 1190
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 7.492700729927008e-06,
+      "loss": 0.0985,
+      "step": 1191
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 7.489051094890512e-06,
+      "loss": 0.0489,
+      "step": 1192
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 7.485401459854015e-06,
+      "loss": 0.0483,
+      "step": 1193
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 7.481751824817519e-06,
+      "loss": 0.0674,
+      "step": 1194
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 7.478102189781023e-06,
+      "loss": 0.0144,
+      "step": 1195
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 7.474452554744526e-06,
+      "loss": 0.0573,
+      "step": 1196
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 7.47080291970803e-06,
+      "loss": 0.0959,
+      "step": 1197
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 7.4671532846715334e-06,
+      "loss": 0.0411,
+      "step": 1198
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 7.4635036496350365e-06,
+      "loss": 0.0568,
+      "step": 1199
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 7.459854014598541e-06,
+      "loss": 0.0402,
+      "step": 1200
+    },
+    {
+      "epoch": 11.11,
+      "eval_loss": 1.461952567100525,
+      "eval_runtime": 65.1642,
+      "eval_samples_per_second": 25.275,
+      "eval_steps_per_second": 1.581,
+      "eval_wer": 0.9821264426514146,
+      "step": 1200
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 7.456204379562044e-06,
+      "loss": 0.0226,
+      "step": 1201
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 7.452554744525547e-06,
+      "loss": 0.0539,
+      "step": 1202
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 7.448905109489052e-06,
+      "loss": 0.0534,
+      "step": 1203
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 7.445255474452555e-06,
+      "loss": 0.0476,
+      "step": 1204
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 7.44160583941606e-06,
+      "loss": 0.0565,
+      "step": 1205
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 7.437956204379563e-06,
+      "loss": 0.0515,
+      "step": 1206
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 7.434306569343066e-06,
+      "loss": 0.0584,
+      "step": 1207
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 7.4306569343065706e-06,
+      "loss": 0.0399,
+      "step": 1208
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 7.427007299270074e-06,
+      "loss": 0.0529,
+      "step": 1209
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 7.423357664233577e-06,
+      "loss": 0.0861,
+      "step": 1210
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 7.419708029197081e-06,
+      "loss": 0.0528,
+      "step": 1211
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 7.416058394160584e-06,
+      "loss": 0.0542,
+      "step": 1212
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 7.4124087591240874e-06,
+      "loss": 0.1104,
+      "step": 1213
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 7.408759124087592e-06,
+      "loss": 0.0281,
+      "step": 1214
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 7.405109489051095e-06,
+      "loss": 0.0418,
+      "step": 1215
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 7.401459854014599e-06,
+      "loss": 0.0785,
+      "step": 1216
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 7.397810218978103e-06,
+      "loss": 0.0727,
+      "step": 1217
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 7.394160583941606e-06,
+      "loss": 0.0196,
+      "step": 1218
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 7.39051094890511e-06,
+      "loss": 0.0313,
+      "step": 1219
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 7.386861313868614e-06,
+      "loss": 0.0601,
+      "step": 1220
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 7.383211678832118e-06,
+      "loss": 0.054,
+      "step": 1221
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 7.379562043795621e-06,
+      "loss": 0.0416,
+      "step": 1222
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 7.3759124087591246e-06,
+      "loss": 0.0451,
+      "step": 1223
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 7.3722627737226285e-06,
+      "loss": 0.0707,
+      "step": 1224
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 7.3686131386861315e-06,
+      "loss": 0.0708,
+      "step": 1225
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 7.364963503649636e-06,
+      "loss": 0.0342,
+      "step": 1226
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 7.361313868613139e-06,
+      "loss": 0.047,
+      "step": 1227
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 7.357664233576642e-06,
+      "loss": 0.047,
+      "step": 1228
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 7.354014598540147e-06,
+      "loss": 0.0829,
+      "step": 1229
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 7.35036496350365e-06,
+      "loss": 0.0434,
+      "step": 1230
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 7.346715328467155e-06,
+      "loss": 0.0791,
+      "step": 1231
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 7.343065693430658e-06,
+      "loss": 0.1062,
+      "step": 1232
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 7.339416058394161e-06,
+      "loss": 0.0301,
+      "step": 1233
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 7.335766423357666e-06,
+      "loss": 0.0797,
+      "step": 1234
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 7.332116788321169e-06,
+      "loss": 0.1074,
+      "step": 1235
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 7.328467153284672e-06,
+      "loss": 0.0574,
+      "step": 1236
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 7.324817518248176e-06,
+      "loss": 0.057,
+      "step": 1237
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 7.321167883211679e-06,
+      "loss": 0.0475,
+      "step": 1238
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 7.3175182481751825e-06,
+      "loss": 0.0774,
+      "step": 1239
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 7.313868613138687e-06,
+      "loss": 0.0658,
+      "step": 1240
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 7.31021897810219e-06,
+      "loss": 0.091,
+      "step": 1241
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 7.306569343065693e-06,
+      "loss": 0.029,
+      "step": 1242
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 7.302919708029198e-06,
+      "loss": 0.0292,
+      "step": 1243
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 7.299270072992701e-06,
+      "loss": 0.0665,
+      "step": 1244
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 7.295620437956205e-06,
+      "loss": 0.074,
+      "step": 1245
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 7.291970802919709e-06,
+      "loss": 0.0319,
+      "step": 1246
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 7.288321167883212e-06,
+      "loss": 0.0395,
+      "step": 1247
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 7.284671532846716e-06,
+      "loss": 0.0694,
+      "step": 1248
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 7.28102189781022e-06,
+      "loss": 0.0422,
+      "step": 1249
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 7.2773722627737235e-06,
+      "loss": 0.0448,
+      "step": 1250
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 7.2737226277372265e-06,
+      "loss": 0.056,
+      "step": 1251
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 7.27007299270073e-06,
+      "loss": 0.0568,
+      "step": 1252
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 7.266423357664234e-06,
+      "loss": 0.0413,
+      "step": 1253
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 7.262773722627737e-06,
+      "loss": 0.0642,
+      "step": 1254
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 7.259124087591242e-06,
+      "loss": 0.0532,
+      "step": 1255
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 7.255474452554745e-06,
+      "loss": 0.0773,
+      "step": 1256
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 7.251824817518249e-06,
+      "loss": 0.0487,
+      "step": 1257
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 7.248175182481753e-06,
+      "loss": 0.0657,
+      "step": 1258
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 7.244525547445256e-06,
+      "loss": 0.0798,
+      "step": 1259
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 7.240875912408761e-06,
+      "loss": 0.0612,
+      "step": 1260
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 7.237226277372264e-06,
+      "loss": 0.0678,
+      "step": 1261
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 7.233576642335767e-06,
+      "loss": 0.0607,
+      "step": 1262
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 7.229927007299271e-06,
+      "loss": 0.0936,
+      "step": 1263
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 7.2262773722627744e-06,
+      "loss": 0.0697,
+      "step": 1264
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 7.2226277372262775e-06,
+      "loss": 0.0292,
+      "step": 1265
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 7.218978102189782e-06,
+      "loss": 0.0937,
+      "step": 1266
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 7.215328467153285e-06,
+      "loss": 0.0248,
+      "step": 1267
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 7.211678832116788e-06,
+      "loss": 0.072,
+      "step": 1268
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 7.208029197080293e-06,
+      "loss": 0.0386,
+      "step": 1269
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 7.204379562043796e-06,
+      "loss": 0.0464,
+      "step": 1270
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 7.200729927007299e-06,
+      "loss": 0.0519,
+      "step": 1271
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 7.197080291970804e-06,
+      "loss": 0.0864,
+      "step": 1272
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 7.193430656934307e-06,
+      "loss": 0.0388,
+      "step": 1273
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 7.189781021897811e-06,
+      "loss": 0.0672,
+      "step": 1274
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 7.186131386861315e-06,
+      "loss": 0.0411,
+      "step": 1275
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 7.182481751824818e-06,
+      "loss": 0.0355,
+      "step": 1276
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 7.1788321167883215e-06,
+      "loss": 0.0792,
+      "step": 1277
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 7.175182481751825e-06,
+      "loss": 0.0652,
+      "step": 1278
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 7.171532846715329e-06,
+      "loss": 0.089,
+      "step": 1279
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 7.167883211678832e-06,
+      "loss": 0.043,
+      "step": 1280
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 7.164233576642336e-06,
+      "loss": 0.0936,
+      "step": 1281
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 7.16058394160584e-06,
+      "loss": 0.0728,
+      "step": 1282
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 7.156934306569344e-06,
+      "loss": 0.0384,
+      "step": 1283
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 7.153284671532848e-06,
+      "loss": 0.0648,
+      "step": 1284
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 7.149635036496351e-06,
+      "loss": 0.0959,
+      "step": 1285
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 7.145985401459855e-06,
+      "loss": 0.0302,
+      "step": 1286
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 7.142335766423359e-06,
+      "loss": 0.0424,
+      "step": 1287
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 7.138686131386862e-06,
+      "loss": 0.049,
+      "step": 1288
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 7.135036496350366e-06,
+      "loss": 0.0452,
+      "step": 1289
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 7.1313868613138694e-06,
+      "loss": 0.0405,
+      "step": 1290
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 7.1277372262773725e-06,
+      "loss": 0.0857,
+      "step": 1291
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 7.124087591240877e-06,
+      "loss": 0.0527,
+      "step": 1292
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 7.12043795620438e-06,
+      "loss": 0.0492,
+      "step": 1293
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 7.116788321167883e-06,
+      "loss": 0.1476,
+      "step": 1294
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 7.113138686131388e-06,
+      "loss": 0.1168,
+      "step": 1295
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 7.109489051094891e-06,
+      "loss": 0.0229,
+      "step": 1296
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 7.105839416058394e-06,
+      "loss": 0.0646,
+      "step": 1297
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 7.102189781021899e-06,
+      "loss": 0.0458,
+      "step": 1298
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 7.098540145985402e-06,
+      "loss": 0.0337,
+      "step": 1299
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 7.094890510948905e-06,
+      "loss": 0.057,
+      "step": 1300
+    },
+    {
+      "epoch": 12.04,
+      "eval_loss": 1.301539421081543,
+      "eval_runtime": 64.9749,
+      "eval_samples_per_second": 25.348,
+      "eval_steps_per_second": 1.585,
+      "eval_wer": 0.9553671739352466,
+      "step": 1300
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 7.09124087591241e-06,
+      "loss": 0.0415,
+      "step": 1301
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 7.087591240875913e-06,
+      "loss": 0.1137,
+      "step": 1302
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 7.0839416058394165e-06,
+      "loss": 0.1132,
+      "step": 1303
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 7.08029197080292e-06,
+      "loss": 0.0319,
+      "step": 1304
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 7.0766423357664234e-06,
+      "loss": 0.0473,
+      "step": 1305
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 7.072992700729927e-06,
+      "loss": 0.0541,
+      "step": 1306
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 7.069343065693431e-06,
+      "loss": 0.015,
+      "step": 1307
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 7.065693430656935e-06,
+      "loss": 0.0337,
+      "step": 1308
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 7.062043795620439e-06,
+      "loss": 0.0376,
+      "step": 1309
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 7.058394160583942e-06,
+      "loss": 0.0636,
+      "step": 1310
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 7.054744525547446e-06,
+      "loss": 0.0481,
+      "step": 1311
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 7.05109489051095e-06,
+      "loss": 0.0708,
+      "step": 1312
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 7.047445255474454e-06,
+      "loss": 0.0283,
+      "step": 1313
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 7.043795620437957e-06,
+      "loss": 0.0326,
+      "step": 1314
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 7.0401459854014606e-06,
+      "loss": 0.0662,
+      "step": 1315
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 7.0364963503649645e-06,
+      "loss": 0.0255,
+      "step": 1316
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 7.0328467153284675e-06,
+      "loss": 0.0523,
+      "step": 1317
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 7.029197080291971e-06,
+      "loss": 0.1185,
+      "step": 1318
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 7.025547445255475e-06,
+      "loss": 0.0537,
+      "step": 1319
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 7.021897810218978e-06,
+      "loss": 0.0416,
+      "step": 1320
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 7.018248175182483e-06,
+      "loss": 0.0467,
+      "step": 1321
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 7.014598540145986e-06,
+      "loss": 0.045,
+      "step": 1322
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 7.010948905109489e-06,
+      "loss": 0.0212,
+      "step": 1323
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 7.007299270072994e-06,
+      "loss": 0.0874,
+      "step": 1324
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 7.003649635036497e-06,
+      "loss": 0.0278,
+      "step": 1325
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 7e-06,
+      "loss": 0.0534,
+      "step": 1326
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 6.996350364963505e-06,
+      "loss": 0.1063,
+      "step": 1327
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 6.992700729927008e-06,
+      "loss": 0.038,
+      "step": 1328
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 6.989051094890511e-06,
+      "loss": 0.029,
+      "step": 1329
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 6.9854014598540154e-06,
+      "loss": 0.0375,
+      "step": 1330
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 6.9817518248175185e-06,
+      "loss": 0.0425,
+      "step": 1331
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 6.978102189781022e-06,
+      "loss": 0.0229,
+      "step": 1332
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 6.974452554744526e-06,
+      "loss": 0.0346,
+      "step": 1333
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 6.970802919708029e-06,
+      "loss": 0.0861,
+      "step": 1334
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 6.967153284671534e-06,
+      "loss": 0.0692,
+      "step": 1335
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 6.963503649635037e-06,
+      "loss": 0.0542,
+      "step": 1336
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 6.959854014598541e-06,
+      "loss": 0.0586,
+      "step": 1337
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 6.956204379562045e-06,
+      "loss": 0.0816,
+      "step": 1338
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 6.952554744525548e-06,
+      "loss": 0.0492,
+      "step": 1339
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 6.948905109489052e-06,
+      "loss": 0.0552,
+      "step": 1340
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 6.945255474452556e-06,
+      "loss": 0.0414,
+      "step": 1341
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 6.941605839416059e-06,
+      "loss": 0.0486,
+      "step": 1342
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 6.9379562043795625e-06,
+      "loss": 0.0659,
+      "step": 1343
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 6.934306569343066e-06,
+      "loss": 0.0338,
+      "step": 1344
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 6.93065693430657e-06,
+      "loss": 0.0609,
+      "step": 1345
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 6.927007299270073e-06,
+      "loss": 0.164,
+      "step": 1346
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 6.923357664233577e-06,
+      "loss": 0.0505,
+      "step": 1347
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 6.919708029197081e-06,
+      "loss": 0.0611,
+      "step": 1348
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 6.916058394160584e-06,
+      "loss": 0.0996,
+      "step": 1349
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 6.912408759124089e-06,
+      "loss": 0.0398,
+      "step": 1350
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 6.908759124087592e-06,
+      "loss": 0.0275,
+      "step": 1351
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 6.905109489051095e-06,
+      "loss": 0.0284,
+      "step": 1352
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 6.9014598540146e-06,
+      "loss": 0.0811,
+      "step": 1353
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 6.897810218978103e-06,
+      "loss": 0.0346,
+      "step": 1354
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 6.894160583941606e-06,
+      "loss": 0.0306,
+      "step": 1355
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 6.8905109489051104e-06,
+      "loss": 0.0519,
+      "step": 1356
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 6.8868613138686135e-06,
+      "loss": 0.0423,
+      "step": 1357
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 6.8832116788321165e-06,
+      "loss": 0.034,
+      "step": 1358
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 6.879562043795621e-06,
+      "loss": 0.086,
+      "step": 1359
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 6.875912408759124e-06,
+      "loss": 0.0202,
+      "step": 1360
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 6.872262773722629e-06,
+      "loss": 0.0415,
+      "step": 1361
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 6.868613138686132e-06,
+      "loss": 0.0536,
+      "step": 1362
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 6.864963503649635e-06,
+      "loss": 0.0644,
+      "step": 1363
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 6.86131386861314e-06,
+      "loss": 0.0709,
+      "step": 1364
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 6.857664233576643e-06,
+      "loss": 0.035,
+      "step": 1365
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 6.854014598540146e-06,
+      "loss": 0.0152,
+      "step": 1366
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 6.850364963503651e-06,
+      "loss": 0.0468,
+      "step": 1367
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 6.846715328467154e-06,
+      "loss": 0.1359,
+      "step": 1368
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 6.8430656934306575e-06,
+      "loss": 0.0414,
+      "step": 1369
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 6.839416058394161e-06,
+      "loss": 0.035,
+      "step": 1370
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 6.8357664233576644e-06,
+      "loss": 0.0365,
+      "step": 1371
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 6.832116788321168e-06,
+      "loss": 0.0422,
+      "step": 1372
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 6.828467153284672e-06,
+      "loss": 0.0612,
+      "step": 1373
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 6.824817518248176e-06,
+      "loss": 0.0935,
+      "step": 1374
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 6.821167883211679e-06,
+      "loss": 0.0731,
+      "step": 1375
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 6.817518248175183e-06,
+      "loss": 0.0377,
+      "step": 1376
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 6.813868613138687e-06,
+      "loss": 0.141,
+      "step": 1377
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 6.81021897810219e-06,
+      "loss": 0.0186,
+      "step": 1378
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 6.806569343065695e-06,
+      "loss": 0.0755,
+      "step": 1379
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 6.802919708029198e-06,
+      "loss": 0.0731,
+      "step": 1380
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 6.799270072992701e-06,
+      "loss": 0.0293,
+      "step": 1381
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 6.7956204379562055e-06,
+      "loss": 0.0451,
+      "step": 1382
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 6.7919708029197085e-06,
+      "loss": 0.0514,
+      "step": 1383
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 6.7883211678832115e-06,
+      "loss": 0.0436,
+      "step": 1384
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 6.784671532846716e-06,
+      "loss": 0.0312,
+      "step": 1385
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 6.781021897810219e-06,
+      "loss": 0.0584,
+      "step": 1386
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 6.777372262773724e-06,
+      "loss": 0.0621,
+      "step": 1387
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 6.773722627737227e-06,
+      "loss": 0.0255,
+      "step": 1388
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 6.77007299270073e-06,
+      "loss": 0.0367,
+      "step": 1389
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 6.766423357664235e-06,
+      "loss": 0.0168,
+      "step": 1390
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 6.762773722627738e-06,
+      "loss": 0.0553,
+      "step": 1391
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 6.759124087591241e-06,
+      "loss": 0.0515,
+      "step": 1392
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 6.755474452554746e-06,
+      "loss": 0.1135,
+      "step": 1393
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 6.751824817518249e-06,
+      "loss": 0.0335,
+      "step": 1394
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 6.748175182481752e-06,
+      "loss": 0.0397,
+      "step": 1395
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 6.744525547445256e-06,
+      "loss": 0.0847,
+      "step": 1396
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 6.7408759124087595e-06,
+      "loss": 0.0159,
+      "step": 1397
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 6.737226277372263e-06,
+      "loss": 0.035,
+      "step": 1398
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 6.733576642335767e-06,
+      "loss": 0.037,
+      "step": 1399
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 6.72992700729927e-06,
+      "loss": 0.0385,
+      "step": 1400
+    },
+    {
+      "epoch": 12.96,
+      "eval_loss": 1.3798296451568604,
+      "eval_runtime": 65.227,
+      "eval_samples_per_second": 25.25,
+      "eval_steps_per_second": 1.579,
+      "eval_wer": 0.9599632315391686,
+      "step": 1400
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 6.726277372262774e-06,
+      "loss": 0.0464,
+      "step": 1401
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 6.722627737226278e-06,
+      "loss": 0.0612,
+      "step": 1402
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 6.718978102189782e-06,
+      "loss": 0.0372,
+      "step": 1403
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 6.715328467153285e-06,
+      "loss": 0.0398,
+      "step": 1404
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 6.711678832116789e-06,
+      "loss": 0.0394,
+      "step": 1405
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 6.708029197080293e-06,
+      "loss": 0.034,
+      "step": 1406
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 6.704379562043796e-06,
+      "loss": 0.049,
+      "step": 1407
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 6.7007299270073005e-06,
+      "loss": 0.0329,
+      "step": 1408
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 6.6970802919708035e-06,
+      "loss": 0.0255,
+      "step": 1409
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 6.6934306569343065e-06,
+      "loss": 0.0559,
+      "step": 1410
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 6.689781021897811e-06,
+      "loss": 0.0366,
+      "step": 1411
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 6.686131386861314e-06,
+      "loss": 0.0348,
+      "step": 1412
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 6.682481751824819e-06,
+      "loss": 0.0784,
+      "step": 1413
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 6.678832116788322e-06,
+      "loss": 0.0708,
+      "step": 1414
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 6.675182481751825e-06,
+      "loss": 0.0285,
+      "step": 1415
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 6.67153284671533e-06,
+      "loss": 0.0632,
+      "step": 1416
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 6.667883211678833e-06,
+      "loss": 0.0405,
+      "step": 1417
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 6.664233576642336e-06,
+      "loss": 0.03,
+      "step": 1418
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 6.660583941605841e-06,
+      "loss": 0.0379,
+      "step": 1419
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 6.656934306569344e-06,
+      "loss": 0.0342,
+      "step": 1420
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 6.653284671532847e-06,
+      "loss": 0.0344,
+      "step": 1421
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 6.6496350364963514e-06,
+      "loss": 0.0403,
+      "step": 1422
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 6.6459854014598545e-06,
+      "loss": 0.0743,
+      "step": 1423
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 6.6423357664233575e-06,
+      "loss": 0.029,
+      "step": 1424
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 6.638686131386862e-06,
+      "loss": 0.0405,
+      "step": 1425
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 6.635036496350365e-06,
+      "loss": 0.0442,
+      "step": 1426
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 6.631386861313869e-06,
+      "loss": 0.0409,
+      "step": 1427
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 6.627737226277373e-06,
+      "loss": 0.0656,
+      "step": 1428
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 6.624087591240876e-06,
+      "loss": 0.0849,
+      "step": 1429
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 6.62043795620438e-06,
+      "loss": 0.0587,
+      "step": 1430
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 6.616788321167884e-06,
+      "loss": 0.021,
+      "step": 1431
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 6.613138686131388e-06,
+      "loss": 0.0635,
+      "step": 1432
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 6.609489051094891e-06,
+      "loss": 0.0417,
+      "step": 1433
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 6.605839416058395e-06,
+      "loss": 0.0786,
+      "step": 1434
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 6.6021897810218985e-06,
+      "loss": 0.0708,
+      "step": 1435
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 6.5985401459854016e-06,
+      "loss": 0.0485,
+      "step": 1436
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 6.594890510948906e-06,
+      "loss": 0.0682,
+      "step": 1437
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 6.591240875912409e-06,
+      "loss": 0.067,
+      "step": 1438
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 6.587591240875912e-06,
+      "loss": 0.04,
+      "step": 1439
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 6.583941605839417e-06,
+      "loss": 0.0361,
+      "step": 1440
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 6.58029197080292e-06,
+      "loss": 0.0681,
+      "step": 1441
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 6.576642335766424e-06,
+      "loss": 0.0604,
+      "step": 1442
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 6.572992700729928e-06,
+      "loss": 0.0764,
+      "step": 1443
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 6.569343065693431e-06,
+      "loss": 0.0609,
+      "step": 1444
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 6.565693430656936e-06,
+      "loss": 0.0837,
+      "step": 1445
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 6.562043795620439e-06,
+      "loss": 0.0348,
+      "step": 1446
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 6.558394160583942e-06,
+      "loss": 0.0277,
+      "step": 1447
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 6.5547445255474464e-06,
+      "loss": 0.0953,
+      "step": 1448
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 6.5510948905109495e-06,
+      "loss": 0.0446,
+      "step": 1449
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 6.5474452554744525e-06,
+      "loss": 0.0582,
+      "step": 1450
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 6.543795620437957e-06,
+      "loss": 0.0819,
+      "step": 1451
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 6.54014598540146e-06,
+      "loss": 0.0407,
+      "step": 1452
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 6.536496350364963e-06,
+      "loss": 0.0367,
+      "step": 1453
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 6.532846715328468e-06,
+      "loss": 0.0987,
+      "step": 1454
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 6.529197080291971e-06,
+      "loss": 0.0186,
+      "step": 1455
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 6.525547445255475e-06,
+      "loss": 0.036,
+      "step": 1456
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 6.521897810218979e-06,
+      "loss": 0.0938,
+      "step": 1457
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 6.518248175182482e-06,
+      "loss": 0.047,
+      "step": 1458
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 6.514598540145986e-06,
+      "loss": 0.0886,
+      "step": 1459
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 6.51094890510949e-06,
+      "loss": 0.1613,
+      "step": 1460
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 6.5072992700729935e-06,
+      "loss": 0.0484,
+      "step": 1461
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 6.5036496350364966e-06,
+      "loss": 0.0451,
+      "step": 1462
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.0378,
+      "step": 1463
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 6.496350364963504e-06,
+      "loss": 0.0361,
+      "step": 1464
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 6.492700729927007e-06,
+      "loss": 0.0366,
+      "step": 1465
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 6.489051094890511e-06,
+      "loss": 0.0494,
+      "step": 1466
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 6.485401459854015e-06,
+      "loss": 0.0277,
+      "step": 1467
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 6.481751824817519e-06,
+      "loss": 0.0455,
+      "step": 1468
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 6.478102189781023e-06,
+      "loss": 0.0198,
+      "step": 1469
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 6.474452554744526e-06,
+      "loss": 0.0852,
+      "step": 1470
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 6.47080291970803e-06,
+      "loss": 0.0238,
+      "step": 1471
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 6.467153284671534e-06,
+      "loss": 0.0581,
+      "step": 1472
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 6.463503649635037e-06,
+      "loss": 0.0773,
+      "step": 1473
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 6.4598540145985415e-06,
+      "loss": 0.0148,
+      "step": 1474
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 6.4562043795620445e-06,
+      "loss": 0.0378,
+      "step": 1475
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 6.4525547445255475e-06,
+      "loss": 0.0554,
+      "step": 1476
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 6.448905109489052e-06,
+      "loss": 0.0418,
+      "step": 1477
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 6.445255474452555e-06,
+      "loss": 0.0647,
+      "step": 1478
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 6.441605839416058e-06,
+      "loss": 0.0507,
+      "step": 1479
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 6.437956204379563e-06,
+      "loss": 0.0324,
+      "step": 1480
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 6.434306569343066e-06,
+      "loss": 0.111,
+      "step": 1481
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 6.430656934306569e-06,
+      "loss": 0.0376,
+      "step": 1482
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 6.427007299270074e-06,
+      "loss": 0.0567,
+      "step": 1483
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 6.423357664233577e-06,
+      "loss": 0.0313,
+      "step": 1484
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 6.419708029197081e-06,
+      "loss": 0.0881,
+      "step": 1485
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 6.416058394160585e-06,
+      "loss": 0.0355,
+      "step": 1486
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 6.412408759124088e-06,
+      "loss": 0.027,
+      "step": 1487
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 6.408759124087592e-06,
+      "loss": 0.0354,
+      "step": 1488
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 6.4051094890510955e-06,
+      "loss": 0.0533,
+      "step": 1489
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 6.4014598540145985e-06,
+      "loss": 0.0368,
+      "step": 1490
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 6.397810218978102e-06,
+      "loss": 0.0667,
+      "step": 1491
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 6.394160583941606e-06,
+      "loss": 0.0473,
+      "step": 1492
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 6.39051094890511e-06,
+      "loss": 0.0521,
+      "step": 1493
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 6.386861313868614e-06,
+      "loss": 0.0457,
+      "step": 1494
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 6.383211678832117e-06,
+      "loss": 0.0795,
+      "step": 1495
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 6.379562043795621e-06,
+      "loss": 0.0466,
+      "step": 1496
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 6.375912408759125e-06,
+      "loss": 0.0741,
+      "step": 1497
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 6.372262773722629e-06,
+      "loss": 0.0654,
+      "step": 1498
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 6.368613138686132e-06,
+      "loss": 0.0477,
+      "step": 1499
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 6.364963503649636e-06,
+      "loss": 0.0422,
+      "step": 1500
+    },
+    {
+      "epoch": 13.88,
+      "eval_loss": 1.3538329601287842,
+      "eval_runtime": 65.1363,
+      "eval_samples_per_second": 25.285,
+      "eval_steps_per_second": 1.581,
+      "eval_wer": 0.969870289040956,
+      "step": 1500
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 6.3613138686131395e-06,
+      "loss": 0.0341,
+      "step": 1501
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 6.3576642335766425e-06,
+      "loss": 0.0416,
+      "step": 1502
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 6.354014598540147e-06,
+      "loss": 0.0425,
+      "step": 1503
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 6.35036496350365e-06,
+      "loss": 0.0708,
+      "step": 1504
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 6.346715328467153e-06,
+      "loss": 0.0308,
+      "step": 1505
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 6.343065693430658e-06,
+      "loss": 0.0622,
+      "step": 1506
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 6.339416058394161e-06,
+      "loss": 0.0581,
+      "step": 1507
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 6.335766423357664e-06,
+      "loss": 0.0521,
+      "step": 1508
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 6.332116788321169e-06,
+      "loss": 0.0503,
+      "step": 1509
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 6.328467153284672e-06,
+      "loss": 0.0288,
+      "step": 1510
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 6.324817518248175e-06,
+      "loss": 0.0552,
+      "step": 1511
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 6.32116788321168e-06,
+      "loss": 0.0708,
+      "step": 1512
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 6.317518248175183e-06,
+      "loss": 0.109,
+      "step": 1513
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 6.313868613138686e-06,
+      "loss": 0.0506,
+      "step": 1514
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 6.3102189781021905e-06,
+      "loss": 0.0379,
+      "step": 1515
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 6.3065693430656935e-06,
+      "loss": 0.016,
+      "step": 1516
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 6.302919708029197e-06,
+      "loss": 0.0183,
+      "step": 1517
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 6.299270072992701e-06,
+      "loss": 0.0537,
+      "step": 1518
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 6.295620437956204e-06,
+      "loss": 0.0806,
+      "step": 1519
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 6.291970802919709e-06,
+      "loss": 0.0468,
+      "step": 1520
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 6.288321167883212e-06,
+      "loss": 0.0297,
+      "step": 1521
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 6.284671532846716e-06,
+      "loss": 0.023,
+      "step": 1522
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 6.28102189781022e-06,
+      "loss": 0.0497,
+      "step": 1523
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 6.277372262773723e-06,
+      "loss": 0.0525,
+      "step": 1524
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 6.273722627737227e-06,
+      "loss": 0.0516,
+      "step": 1525
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 6.270072992700731e-06,
+      "loss": 0.0397,
+      "step": 1526
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 6.2664233576642345e-06,
+      "loss": 0.033,
+      "step": 1527
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 6.2627737226277376e-06,
+      "loss": 0.0827,
+      "step": 1528
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 6.2591240875912414e-06,
+      "loss": 0.0235,
+      "step": 1529
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 6.255474452554745e-06,
+      "loss": 0.0848,
+      "step": 1530
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 6.251824817518248e-06,
+      "loss": 0.041,
+      "step": 1531
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 6.248175182481753e-06,
+      "loss": 0.0627,
+      "step": 1532
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 6.244525547445256e-06,
+      "loss": 0.0444,
+      "step": 1533
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 6.240875912408759e-06,
+      "loss": 0.0428,
+      "step": 1534
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 6.237226277372264e-06,
+      "loss": 0.0833,
+      "step": 1535
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 6.233576642335767e-06,
+      "loss": 0.0308,
+      "step": 1536
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 6.22992700729927e-06,
+      "loss": 0.067,
+      "step": 1537
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 6.226277372262775e-06,
+      "loss": 0.0164,
+      "step": 1538
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 6.222627737226278e-06,
+      "loss": 0.0221,
+      "step": 1539
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 6.218978102189781e-06,
+      "loss": 0.0609,
+      "step": 1540
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 6.2153284671532855e-06,
+      "loss": 0.0303,
+      "step": 1541
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 6.2116788321167885e-06,
+      "loss": 0.0278,
+      "step": 1542
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 6.2080291970802916e-06,
+      "loss": 0.0571,
+      "step": 1543
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 6.204379562043796e-06,
+      "loss": 0.0474,
+      "step": 1544
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 6.200729927007299e-06,
+      "loss": 0.0322,
+      "step": 1545
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 6.197080291970804e-06,
+      "loss": 0.0385,
+      "step": 1546
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 6.193430656934307e-06,
+      "loss": 0.0689,
+      "step": 1547
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 6.18978102189781e-06,
+      "loss": 0.0398,
+      "step": 1548
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 6.186131386861315e-06,
+      "loss": 0.0448,
+      "step": 1549
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 6.182481751824818e-06,
+      "loss": 0.0248,
+      "step": 1550
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 6.178832116788322e-06,
+      "loss": 0.0343,
+      "step": 1551
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 6.175182481751826e-06,
+      "loss": 0.0242,
+      "step": 1552
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 6.171532846715329e-06,
+      "loss": 0.0322,
+      "step": 1553
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 6.1678832116788326e-06,
+      "loss": 0.0282,
+      "step": 1554
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 6.1642335766423365e-06,
+      "loss": 0.0387,
+      "step": 1555
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 6.16058394160584e-06,
+      "loss": 0.0463,
+      "step": 1556
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 6.156934306569343e-06,
+      "loss": 0.0422,
+      "step": 1557
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 6.153284671532847e-06,
+      "loss": 0.0207,
+      "step": 1558
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 6.149635036496351e-06,
+      "loss": 0.0741,
+      "step": 1559
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 6.145985401459854e-06,
+      "loss": 0.036,
+      "step": 1560
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 6.142335766423359e-06,
+      "loss": 0.0464,
+      "step": 1561
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 6.138686131386862e-06,
+      "loss": 0.05,
+      "step": 1562
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 6.135036496350365e-06,
+      "loss": 0.0544,
+      "step": 1563
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 6.13138686131387e-06,
+      "loss": 0.024,
+      "step": 1564
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 6.127737226277373e-06,
+      "loss": 0.0578,
+      "step": 1565
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 6.124087591240876e-06,
+      "loss": 0.053,
+      "step": 1566
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 6.1204379562043805e-06,
+      "loss": 0.0628,
+      "step": 1567
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 6.1167883211678835e-06,
+      "loss": 0.0369,
+      "step": 1568
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 6.1131386861313866e-06,
+      "loss": 0.0569,
+      "step": 1569
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 6.109489051094891e-06,
+      "loss": 0.0329,
+      "step": 1570
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 6.105839416058394e-06,
+      "loss": 0.0709,
+      "step": 1571
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 6.102189781021899e-06,
+      "loss": 0.0517,
+      "step": 1572
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 6.098540145985402e-06,
+      "loss": 0.0488,
+      "step": 1573
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 6.094890510948905e-06,
+      "loss": 0.0403,
+      "step": 1574
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 6.09124087591241e-06,
+      "loss": 0.1023,
+      "step": 1575
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 6.087591240875913e-06,
+      "loss": 0.0429,
+      "step": 1576
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 6.083941605839416e-06,
+      "loss": 0.0621,
+      "step": 1577
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 6.080291970802921e-06,
+      "loss": 0.0169,
+      "step": 1578
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 6.076642335766424e-06,
+      "loss": 0.0253,
+      "step": 1579
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 6.072992700729928e-06,
+      "loss": 0.047,
+      "step": 1580
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 6.0693430656934315e-06,
+      "loss": 0.0382,
+      "step": 1581
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 6.0656934306569345e-06,
+      "loss": 0.031,
+      "step": 1582
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 6.062043795620438e-06,
+      "loss": 0.0568,
+      "step": 1583
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 6.058394160583942e-06,
+      "loss": 0.0297,
+      "step": 1584
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 6.054744525547446e-06,
+      "loss": 0.028,
+      "step": 1585
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 6.051094890510949e-06,
+      "loss": 0.021,
+      "step": 1586
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 6.047445255474453e-06,
+      "loss": 0.0567,
+      "step": 1587
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 6.043795620437957e-06,
+      "loss": 0.0206,
+      "step": 1588
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 6.04014598540146e-06,
+      "loss": 0.0658,
+      "step": 1589
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 6.036496350364964e-06,
+      "loss": 0.0603,
+      "step": 1590
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 6.032846715328468e-06,
+      "loss": 0.0389,
+      "step": 1591
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 6.029197080291971e-06,
+      "loss": 0.0472,
+      "step": 1592
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 6.0255474452554755e-06,
+      "loss": 0.0458,
+      "step": 1593
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 6.0218978102189786e-06,
+      "loss": 0.07,
+      "step": 1594
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 6.018248175182482e-06,
+      "loss": 0.0324,
+      "step": 1595
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 6.014598540145986e-06,
+      "loss": 0.0556,
+      "step": 1596
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 6.010948905109489e-06,
+      "loss": 0.0458,
+      "step": 1597
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 6.007299270072994e-06,
+      "loss": 0.0532,
+      "step": 1598
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 6.003649635036497e-06,
+      "loss": 0.1199,
+      "step": 1599
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 6e-06,
+      "loss": 0.014,
+      "step": 1600
+    },
+    {
+      "epoch": 14.81,
+      "eval_loss": 1.2507344484329224,
+      "eval_runtime": 65.2468,
+      "eval_samples_per_second": 25.243,
+      "eval_steps_per_second": 1.579,
+      "eval_wer": 0.9443366356858339,
+      "step": 1600
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 5.996350364963505e-06,
+      "loss": 0.0414,
+      "step": 1601
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 5.992700729927008e-06,
+      "loss": 0.0533,
+      "step": 1602
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 5.989051094890511e-06,
+      "loss": 0.0822,
+      "step": 1603
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 5.985401459854016e-06,
+      "loss": 0.0356,
+      "step": 1604
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 5.981751824817519e-06,
+      "loss": 0.0589,
+      "step": 1605
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 5.978102189781022e-06,
+      "loss": 0.0426,
+      "step": 1606
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 5.9744525547445265e-06,
+      "loss": 0.0217,
+      "step": 1607
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 5.9708029197080295e-06,
+      "loss": 0.054,
+      "step": 1608
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 5.967153284671533e-06,
+      "loss": 0.0298,
+      "step": 1609
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 5.963503649635037e-06,
+      "loss": 0.0388,
+      "step": 1610
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 5.95985401459854e-06,
+      "loss": 0.0458,
+      "step": 1611
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 5.956204379562044e-06,
+      "loss": 0.0795,
+      "step": 1612
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 5.952554744525548e-06,
+      "loss": 0.0203,
+      "step": 1613
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 5.948905109489051e-06,
+      "loss": 0.0545,
+      "step": 1614
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 5.945255474452555e-06,
+      "loss": 0.0799,
+      "step": 1615
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 5.941605839416059e-06,
+      "loss": 0.0415,
+      "step": 1616
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 5.937956204379563e-06,
+      "loss": 0.0414,
+      "step": 1617
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 5.934306569343066e-06,
+      "loss": 0.0689,
+      "step": 1618
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 5.93065693430657e-06,
+      "loss": 0.0444,
+      "step": 1619
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 5.9270072992700736e-06,
+      "loss": 0.0317,
+      "step": 1620
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 5.923357664233577e-06,
+      "loss": 0.0579,
+      "step": 1621
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 5.919708029197081e-06,
+      "loss": 0.0346,
+      "step": 1622
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 5.916058394160584e-06,
+      "loss": 0.0546,
+      "step": 1623
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 5.912408759124088e-06,
+      "loss": 0.0474,
+      "step": 1624
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 5.908759124087592e-06,
+      "loss": 0.0333,
+      "step": 1625
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 5.905109489051095e-06,
+      "loss": 0.0644,
+      "step": 1626
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 5.9014598540146e-06,
+      "loss": 0.0324,
+      "step": 1627
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 5.897810218978103e-06,
+      "loss": 0.0143,
+      "step": 1628
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 5.894160583941606e-06,
+      "loss": 0.0488,
+      "step": 1629
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 5.890510948905111e-06,
+      "loss": 0.061,
+      "step": 1630
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 5.886861313868614e-06,
+      "loss": 0.0293,
+      "step": 1631
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 5.883211678832117e-06,
+      "loss": 0.0452,
+      "step": 1632
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 5.8795620437956215e-06,
+      "loss": 0.0206,
+      "step": 1633
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 5.8759124087591245e-06,
+      "loss": 0.0261,
+      "step": 1634
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 5.8722627737226276e-06,
+      "loss": 0.0424,
+      "step": 1635
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 5.868613138686132e-06,
+      "loss": 0.0308,
+      "step": 1636
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 5.864963503649635e-06,
+      "loss": 0.0124,
+      "step": 1637
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 5.861313868613138e-06,
+      "loss": 0.0392,
+      "step": 1638
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 5.857664233576643e-06,
+      "loss": 0.1221,
+      "step": 1639
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 5.854014598540146e-06,
+      "loss": 0.0533,
+      "step": 1640
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 5.85036496350365e-06,
+      "loss": 0.0272,
+      "step": 1641
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 5.846715328467154e-06,
+      "loss": 0.0401,
+      "step": 1642
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 5.843065693430657e-06,
+      "loss": 0.0211,
+      "step": 1643
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 5.839416058394161e-06,
+      "loss": 0.0453,
+      "step": 1644
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 5.835766423357665e-06,
+      "loss": 0.0733,
+      "step": 1645
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 5.832116788321169e-06,
+      "loss": 0.0389,
+      "step": 1646
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 5.828467153284672e-06,
+      "loss": 0.0302,
+      "step": 1647
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 5.8248175182481755e-06,
+      "loss": 0.0412,
+      "step": 1648
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 5.821167883211679e-06,
+      "loss": 0.0266,
+      "step": 1649
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 5.817518248175182e-06,
+      "loss": 0.0196,
+      "step": 1650
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 5.813868613138687e-06,
+      "loss": 0.0377,
+      "step": 1651
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 5.81021897810219e-06,
+      "loss": 0.0284,
+      "step": 1652
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 5.806569343065694e-06,
+      "loss": 0.0281,
+      "step": 1653
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 5.802919708029198e-06,
+      "loss": 0.0348,
+      "step": 1654
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 5.799270072992701e-06,
+      "loss": 0.0368,
+      "step": 1655
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 5.795620437956206e-06,
+      "loss": 0.0734,
+      "step": 1656
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 5.791970802919709e-06,
+      "loss": 0.022,
+      "step": 1657
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 5.788321167883212e-06,
+      "loss": 0.0413,
+      "step": 1658
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 5.7846715328467165e-06,
+      "loss": 0.0369,
+      "step": 1659
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 5.7810218978102195e-06,
+      "loss": 0.0334,
+      "step": 1660
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 5.777372262773723e-06,
+      "loss": 0.0709,
+      "step": 1661
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 5.773722627737227e-06,
+      "loss": 0.0742,
+      "step": 1662
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 5.77007299270073e-06,
+      "loss": 0.0749,
+      "step": 1663
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 5.766423357664233e-06,
+      "loss": 0.0558,
+      "step": 1664
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 5.762773722627738e-06,
+      "loss": 0.0244,
+      "step": 1665
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 5.759124087591241e-06,
+      "loss": 0.0684,
+      "step": 1666
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 5.755474452554744e-06,
+      "loss": 0.0437,
+      "step": 1667
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 5.751824817518249e-06,
+      "loss": 0.0167,
+      "step": 1668
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 5.748175182481752e-06,
+      "loss": 0.026,
+      "step": 1669
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 5.744525547445256e-06,
+      "loss": 0.0458,
+      "step": 1670
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 5.74087591240876e-06,
+      "loss": 0.0271,
+      "step": 1671
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 5.737226277372263e-06,
+      "loss": 0.0667,
+      "step": 1672
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 5.733576642335767e-06,
+      "loss": 0.0874,
+      "step": 1673
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 5.7299270072992705e-06,
+      "loss": 0.0508,
+      "step": 1674
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 5.726277372262774e-06,
+      "loss": 0.0341,
+      "step": 1675
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 5.7226277372262774e-06,
+      "loss": 0.0526,
+      "step": 1676
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 5.718978102189781e-06,
+      "loss": 0.0448,
+      "step": 1677
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 5.715328467153285e-06,
+      "loss": 0.0342,
+      "step": 1678
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 5.711678832116789e-06,
+      "loss": 0.077,
+      "step": 1679
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 5.708029197080293e-06,
+      "loss": 0.0397,
+      "step": 1680
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 5.704379562043796e-06,
+      "loss": 0.0334,
+      "step": 1681
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 5.7007299270073e-06,
+      "loss": 0.0519,
+      "step": 1682
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 5.697080291970804e-06,
+      "loss": 0.045,
+      "step": 1683
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 5.693430656934307e-06,
+      "loss": 0.0146,
+      "step": 1684
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 5.6897810218978115e-06,
+      "loss": 0.053,
+      "step": 1685
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 5.6861313868613146e-06,
+      "loss": 0.0533,
+      "step": 1686
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 5.682481751824818e-06,
+      "loss": 0.0326,
+      "step": 1687
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 5.678832116788322e-06,
+      "loss": 0.0303,
+      "step": 1688
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 5.675182481751825e-06,
+      "loss": 0.0355,
+      "step": 1689
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 5.671532846715328e-06,
+      "loss": 0.0288,
+      "step": 1690
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 5.667883211678833e-06,
+      "loss": 0.037,
+      "step": 1691
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 5.664233576642336e-06,
+      "loss": 0.0932,
+      "step": 1692
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 5.660583941605839e-06,
+      "loss": 0.0288,
+      "step": 1693
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 5.656934306569344e-06,
+      "loss": 0.0476,
+      "step": 1694
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 5.653284671532847e-06,
+      "loss": 0.0788,
+      "step": 1695
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 5.64963503649635e-06,
+      "loss": 0.02,
+      "step": 1696
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 5.645985401459855e-06,
+      "loss": 0.0412,
+      "step": 1697
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 5.642335766423358e-06,
+      "loss": 0.0618,
+      "step": 1698
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 5.638686131386862e-06,
+      "loss": 0.0355,
+      "step": 1699
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 5.6350364963503655e-06,
+      "loss": 0.0232,
+      "step": 1700
+    },
+    {
+      "epoch": 15.74,
+      "eval_loss": 1.3318383693695068,
+      "eval_runtime": 65.6812,
+      "eval_samples_per_second": 25.076,
+      "eval_steps_per_second": 1.568,
+      "eval_wer": 0.9464814625676642,
+      "step": 1700
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 5.6313868613138686e-06,
+      "loss": 0.0798,
+      "step": 1701
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 5.6277372262773724e-06,
+      "loss": 0.0533,
+      "step": 1702
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 5.624087591240876e-06,
+      "loss": 0.0306,
+      "step": 1703
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 5.62043795620438e-06,
+      "loss": 0.0368,
+      "step": 1704
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 5.616788321167884e-06,
+      "loss": 0.0533,
+      "step": 1705
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 5.613138686131387e-06,
+      "loss": 0.0255,
+      "step": 1706
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 5.609489051094891e-06,
+      "loss": 0.0376,
+      "step": 1707
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 5.605839416058395e-06,
+      "loss": 0.0608,
+      "step": 1708
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 5.602189781021899e-06,
+      "loss": 0.0093,
+      "step": 1709
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 5.598540145985402e-06,
+      "loss": 0.0671,
+      "step": 1710
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 5.594890510948906e-06,
+      "loss": 0.0765,
+      "step": 1711
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 5.5912408759124096e-06,
+      "loss": 0.0254,
+      "step": 1712
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 5.587591240875913e-06,
+      "loss": 0.0922,
+      "step": 1713
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 5.5839416058394165e-06,
+      "loss": 0.0474,
+      "step": 1714
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 5.58029197080292e-06,
+      "loss": 0.0839,
+      "step": 1715
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 5.576642335766423e-06,
+      "loss": 0.0237,
+      "step": 1716
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 5.572992700729928e-06,
+      "loss": 0.1307,
+      "step": 1717
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 5.569343065693431e-06,
+      "loss": 0.0262,
+      "step": 1718
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 5.565693430656934e-06,
+      "loss": 0.0362,
+      "step": 1719
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 5.562043795620439e-06,
+      "loss": 0.0352,
+      "step": 1720
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 5.558394160583942e-06,
+      "loss": 0.0277,
+      "step": 1721
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 5.554744525547445e-06,
+      "loss": 0.0529,
+      "step": 1722
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 5.55109489051095e-06,
+      "loss": 0.0575,
+      "step": 1723
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 5.547445255474453e-06,
+      "loss": 0.0617,
+      "step": 1724
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 5.543795620437956e-06,
+      "loss": 0.0528,
+      "step": 1725
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 5.5401459854014605e-06,
+      "loss": 0.0484,
+      "step": 1726
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 5.5364963503649636e-06,
+      "loss": 0.0362,
+      "step": 1727
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 5.5328467153284675e-06,
+      "loss": 0.0281,
+      "step": 1728
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 5.529197080291971e-06,
+      "loss": 0.0598,
+      "step": 1729
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 5.525547445255474e-06,
+      "loss": 0.0278,
+      "step": 1730
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 5.521897810218979e-06,
+      "loss": 0.0705,
+      "step": 1731
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 5.518248175182482e-06,
+      "loss": 0.0391,
+      "step": 1732
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 5.514598540145986e-06,
+      "loss": 0.0066,
+      "step": 1733
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 5.51094890510949e-06,
+      "loss": 0.0901,
+      "step": 1734
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 5.507299270072993e-06,
+      "loss": 0.0434,
+      "step": 1735
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 5.503649635036497e-06,
+      "loss": 0.0623,
+      "step": 1736
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.0541,
+      "step": 1737
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 5.496350364963504e-06,
+      "loss": 0.0409,
+      "step": 1738
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 5.492700729927008e-06,
+      "loss": 0.0297,
+      "step": 1739
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 5.4890510948905115e-06,
+      "loss": 0.0621,
+      "step": 1740
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 5.485401459854015e-06,
+      "loss": 0.0307,
+      "step": 1741
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 5.481751824817518e-06,
+      "loss": 0.014,
+      "step": 1742
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 5.478102189781022e-06,
+      "loss": 0.0206,
+      "step": 1743
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 5.474452554744526e-06,
+      "loss": 0.0542,
+      "step": 1744
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 5.470802919708029e-06,
+      "loss": 0.0165,
+      "step": 1745
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 5.467153284671534e-06,
+      "loss": 0.0264,
+      "step": 1746
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 5.463503649635037e-06,
+      "loss": 0.0696,
+      "step": 1747
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 5.45985401459854e-06,
+      "loss": 0.0181,
+      "step": 1748
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 5.456204379562045e-06,
+      "loss": 0.0353,
+      "step": 1749
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 5.452554744525548e-06,
+      "loss": 0.0534,
+      "step": 1750
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.448905109489051e-06,
+      "loss": 0.0294,
+      "step": 1751
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.4452554744525556e-06,
+      "loss": 0.0259,
+      "step": 1752
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 5.441605839416059e-06,
+      "loss": 0.0444,
+      "step": 1753
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.437956204379562e-06,
+      "loss": 0.0374,
+      "step": 1754
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.434306569343066e-06,
+      "loss": 0.0346,
+      "step": 1755
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.430656934306569e-06,
+      "loss": 0.0241,
+      "step": 1756
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 5.427007299270074e-06,
+      "loss": 0.0552,
+      "step": 1757
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.423357664233577e-06,
+      "loss": 0.0537,
+      "step": 1758
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.41970802919708e-06,
+      "loss": 0.0264,
+      "step": 1759
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.416058394160585e-06,
+      "loss": 0.0709,
+      "step": 1760
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.412408759124088e-06,
+      "loss": 0.0353,
+      "step": 1761
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.408759124087591e-06,
+      "loss": 0.0712,
+      "step": 1762
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 5.405109489051096e-06,
+      "loss": 0.049,
+      "step": 1763
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.401459854014599e-06,
+      "loss": 0.0214,
+      "step": 1764
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.397810218978103e-06,
+      "loss": 0.05,
+      "step": 1765
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.3941605839416065e-06,
+      "loss": 0.054,
+      "step": 1766
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.3905109489051096e-06,
+      "loss": 0.0265,
+      "step": 1767
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 5.3868613138686134e-06,
+      "loss": 0.0453,
+      "step": 1768
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.383211678832117e-06,
+      "loss": 0.0641,
+      "step": 1769
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.379562043795621e-06,
+      "loss": 0.0246,
+      "step": 1770
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.375912408759124e-06,
+      "loss": 0.0416,
+      "step": 1771
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.372262773722628e-06,
+      "loss": 0.0616,
+      "step": 1772
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.368613138686132e-06,
+      "loss": 0.0606,
+      "step": 1773
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.364963503649635e-06,
+      "loss": 0.0467,
+      "step": 1774
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.36131386861314e-06,
+      "loss": 0.072,
+      "step": 1775
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.357664233576643e-06,
+      "loss": 0.051,
+      "step": 1776
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.354014598540146e-06,
+      "loss": 0.0184,
+      "step": 1777
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.3503649635036506e-06,
+      "loss": 0.0493,
+      "step": 1778
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.346715328467154e-06,
+      "loss": 0.0216,
+      "step": 1779
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.343065693430657e-06,
+      "loss": 0.0156,
+      "step": 1780
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.339416058394161e-06,
+      "loss": 0.0588,
+      "step": 1781
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.335766423357664e-06,
+      "loss": 0.0474,
+      "step": 1782
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.332116788321169e-06,
+      "loss": 0.0208,
+      "step": 1783
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.328467153284672e-06,
+      "loss": 0.0659,
+      "step": 1784
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.324817518248175e-06,
+      "loss": 0.0446,
+      "step": 1785
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.32116788321168e-06,
+      "loss": 0.0143,
+      "step": 1786
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.317518248175183e-06,
+      "loss": 0.038,
+      "step": 1787
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.313868613138686e-06,
+      "loss": 0.0267,
+      "step": 1788
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.310218978102191e-06,
+      "loss": 0.0176,
+      "step": 1789
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.306569343065694e-06,
+      "loss": 0.044,
+      "step": 1790
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.302919708029197e-06,
+      "loss": 0.0371,
+      "step": 1791
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.2992700729927015e-06,
+      "loss": 0.0425,
+      "step": 1792
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.2956204379562046e-06,
+      "loss": 0.0142,
+      "step": 1793
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.2919708029197084e-06,
+      "loss": 0.0698,
+      "step": 1794
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.288321167883212e-06,
+      "loss": 0.0505,
+      "step": 1795
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.284671532846715e-06,
+      "loss": 0.0368,
+      "step": 1796
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.281021897810219e-06,
+      "loss": 0.0346,
+      "step": 1797
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.277372262773723e-06,
+      "loss": 0.0284,
+      "step": 1798
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.273722627737227e-06,
+      "loss": 0.0263,
+      "step": 1799
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.27007299270073e-06,
+      "loss": 0.0554,
+      "step": 1800
+    },
+    {
+      "epoch": 16.66,
+      "eval_loss": 1.2784050703048706,
+      "eval_runtime": 65.192,
+      "eval_samples_per_second": 25.264,
+      "eval_steps_per_second": 1.58,
+      "eval_wer": 0.9461750587274027,
+      "step": 1800
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.266423357664234e-06,
+      "loss": 0.0296,
+      "step": 1801
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.262773722627738e-06,
+      "loss": 0.0484,
+      "step": 1802
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.259124087591241e-06,
+      "loss": 0.0313,
+      "step": 1803
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.255474452554746e-06,
+      "loss": 0.0291,
+      "step": 1804
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.251824817518249e-06,
+      "loss": 0.0251,
+      "step": 1805
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.248175182481752e-06,
+      "loss": 0.089,
+      "step": 1806
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.244525547445256e-06,
+      "loss": 0.0383,
+      "step": 1807
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.240875912408759e-06,
+      "loss": 0.0578,
+      "step": 1808
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 5.237226277372264e-06,
+      "loss": 0.0435,
+      "step": 1809
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.233576642335767e-06,
+      "loss": 0.0505,
+      "step": 1810
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.22992700729927e-06,
+      "loss": 0.0248,
+      "step": 1811
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 5.226277372262775e-06,
+      "loss": 0.0672,
+      "step": 1812
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 5.222627737226278e-06,
+      "loss": 0.0328,
+      "step": 1813
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 5.218978102189781e-06,
+      "loss": 0.0294,
+      "step": 1814
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 5.215328467153286e-06,
+      "loss": 0.0307,
+      "step": 1815
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 5.211678832116789e-06,
+      "loss": 0.0443,
+      "step": 1816
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 5.208029197080292e-06,
+      "loss": 0.0176,
+      "step": 1817
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 5.2043795620437965e-06,
+      "loss": 0.0511,
+      "step": 1818
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 5.2007299270073e-06,
+      "loss": 0.0674,
+      "step": 1819
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 5.197080291970803e-06,
+      "loss": 0.0279,
+      "step": 1820
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 5.193430656934307e-06,
+      "loss": 0.0801,
+      "step": 1821
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 5.18978102189781e-06,
+      "loss": 0.0475,
+      "step": 1822
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 5.186131386861314e-06,
+      "loss": 0.0404,
+      "step": 1823
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 5.182481751824818e-06,
+      "loss": 0.0474,
+      "step": 1824
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 5.178832116788321e-06,
+      "loss": 0.0644,
+      "step": 1825
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 5.175182481751825e-06,
+      "loss": 0.0312,
+      "step": 1826
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 5.171532846715329e-06,
+      "loss": 0.0227,
+      "step": 1827
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 5.167883211678833e-06,
+      "loss": 0.0999,
+      "step": 1828
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 5.164233576642336e-06,
+      "loss": 0.0301,
+      "step": 1829
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 5.16058394160584e-06,
+      "loss": 0.0513,
+      "step": 1830
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 5.156934306569344e-06,
+      "loss": 0.0718,
+      "step": 1831
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 5.153284671532847e-06,
+      "loss": 0.018,
+      "step": 1832
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 5.149635036496351e-06,
+      "loss": 0.0238,
+      "step": 1833
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 5.1459854014598544e-06,
+      "loss": 0.0628,
+      "step": 1834
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 5.142335766423358e-06,
+      "loss": 0.0306,
+      "step": 1835
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 5.138686131386862e-06,
+      "loss": 0.0202,
+      "step": 1836
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 5.135036496350365e-06,
+      "loss": 0.0909,
+      "step": 1837
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 5.13138686131387e-06,
+      "loss": 0.0171,
+      "step": 1838
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 5.127737226277373e-06,
+      "loss": 0.0992,
+      "step": 1839
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 5.124087591240876e-06,
+      "loss": 0.0474,
+      "step": 1840
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 5.120437956204381e-06,
+      "loss": 0.0435,
+      "step": 1841
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 5.116788321167884e-06,
+      "loss": 0.0252,
+      "step": 1842
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 5.113138686131387e-06,
+      "loss": 0.0427,
+      "step": 1843
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 5.1094890510948916e-06,
+      "loss": 0.0468,
+      "step": 1844
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 5.105839416058395e-06,
+      "loss": 0.0445,
+      "step": 1845
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 5.102189781021898e-06,
+      "loss": 0.0675,
+      "step": 1846
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 5.098540145985402e-06,
+      "loss": 0.0443,
+      "step": 1847
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 5.094890510948905e-06,
+      "loss": 0.0431,
+      "step": 1848
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 5.0912408759124084e-06,
+      "loss": 0.0215,
+      "step": 1849
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 5.087591240875913e-06,
+      "loss": 0.0517,
+      "step": 1850
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 5.083941605839416e-06,
+      "loss": 0.0492,
+      "step": 1851
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 5.08029197080292e-06,
+      "loss": 0.0477,
+      "step": 1852
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 5.076642335766424e-06,
+      "loss": 0.0228,
+      "step": 1853
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 5.072992700729927e-06,
+      "loss": 0.0319,
+      "step": 1854
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 5.069343065693431e-06,
+      "loss": 0.061,
+      "step": 1855
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 5.065693430656935e-06,
+      "loss": 0.0177,
+      "step": 1856
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 5.062043795620439e-06,
+      "loss": 0.0575,
+      "step": 1857
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 5.058394160583942e-06,
+      "loss": 0.063,
+      "step": 1858
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 5.0547445255474456e-06,
+      "loss": 0.0122,
+      "step": 1859
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 5.0510948905109494e-06,
+      "loss": 0.0505,
+      "step": 1860
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 5.047445255474453e-06,
+      "loss": 0.0614,
+      "step": 1861
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 5.043795620437957e-06,
+      "loss": 0.0176,
+      "step": 1862
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 5.04014598540146e-06,
+      "loss": 0.0208,
+      "step": 1863
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 5.036496350364964e-06,
+      "loss": 0.0541,
+      "step": 1864
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 5.032846715328468e-06,
+      "loss": 0.0294,
+      "step": 1865
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 5.029197080291971e-06,
+      "loss": 0.0459,
+      "step": 1866
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 5.025547445255475e-06,
+      "loss": 0.0761,
+      "step": 1867
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 5.021897810218979e-06,
+      "loss": 0.0336,
+      "step": 1868
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 5.018248175182482e-06,
+      "loss": 0.0449,
+      "step": 1869
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 5.0145985401459866e-06,
+      "loss": 0.0875,
+      "step": 1870
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 5.01094890510949e-06,
+      "loss": 0.0504,
+      "step": 1871
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 5.007299270072993e-06,
+      "loss": 0.0879,
+      "step": 1872
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 5.003649635036497e-06,
+      "loss": 0.0733,
+      "step": 1873
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 5e-06,
+      "loss": 0.0448,
+      "step": 1874
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.996350364963504e-06,
+      "loss": 0.0475,
+      "step": 1875
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.992700729927007e-06,
+      "loss": 0.0486,
+      "step": 1876
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.989051094890511e-06,
+      "loss": 0.0461,
+      "step": 1877
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 4.985401459854015e-06,
+      "loss": 0.0494,
+      "step": 1878
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.981751824817518e-06,
+      "loss": 0.0475,
+      "step": 1879
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.978102189781022e-06,
+      "loss": 0.0757,
+      "step": 1880
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.974452554744526e-06,
+      "loss": 0.067,
+      "step": 1881
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 4.97080291970803e-06,
+      "loss": 0.0265,
+      "step": 1882
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 4.967153284671533e-06,
+      "loss": 0.0856,
+      "step": 1883
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 4.963503649635037e-06,
+      "loss": 0.0226,
+      "step": 1884
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.9598540145985406e-06,
+      "loss": 0.0729,
+      "step": 1885
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 4.956204379562044e-06,
+      "loss": 0.0451,
+      "step": 1886
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 4.9525547445255475e-06,
+      "loss": 0.0148,
+      "step": 1887
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 4.948905109489051e-06,
+      "loss": 0.0382,
+      "step": 1888
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 4.945255474452555e-06,
+      "loss": 0.0551,
+      "step": 1889
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 4.941605839416059e-06,
+      "loss": 0.0155,
+      "step": 1890
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 4.937956204379562e-06,
+      "loss": 0.0485,
+      "step": 1891
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 4.934306569343066e-06,
+      "loss": 0.0307,
+      "step": 1892
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 4.93065693430657e-06,
+      "loss": 0.0276,
+      "step": 1893
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 4.927007299270074e-06,
+      "loss": 0.0273,
+      "step": 1894
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 4.923357664233578e-06,
+      "loss": 0.0481,
+      "step": 1895
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 4.919708029197081e-06,
+      "loss": 0.0355,
+      "step": 1896
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 4.916058394160585e-06,
+      "loss": 0.0301,
+      "step": 1897
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 4.9124087591240885e-06,
+      "loss": 0.0339,
+      "step": 1898
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 4.9087591240875915e-06,
+      "loss": 0.062,
+      "step": 1899
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 4.905109489051095e-06,
+      "loss": 0.0316,
+      "step": 1900
+    },
+    {
+      "epoch": 17.59,
+      "eval_loss": 1.2502586841583252,
+      "eval_runtime": 64.9686,
+      "eval_samples_per_second": 25.351,
+      "eval_steps_per_second": 1.585,
+      "eval_wer": 0.948115616382392,
+      "step": 1900
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 4.901459854014599e-06,
+      "loss": 0.0475,
+      "step": 1901
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 4.897810218978102e-06,
+      "loss": 0.03,
+      "step": 1902
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 4.894160583941606e-06,
+      "loss": 0.0179,
+      "step": 1903
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 4.89051094890511e-06,
+      "loss": 0.0433,
+      "step": 1904
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 4.886861313868613e-06,
+      "loss": 0.0483,
+      "step": 1905
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 4.883211678832117e-06,
+      "loss": 0.0153,
+      "step": 1906
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 4.879562043795621e-06,
+      "loss": 0.0427,
+      "step": 1907
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 4.875912408759125e-06,
+      "loss": 0.0358,
+      "step": 1908
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 4.872262773722628e-06,
+      "loss": 0.0082,
+      "step": 1909
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 4.868613138686132e-06,
+      "loss": 0.0394,
+      "step": 1910
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 4.864963503649636e-06,
+      "loss": 0.1022,
+      "step": 1911
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 4.861313868613139e-06,
+      "loss": 0.011,
+      "step": 1912
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 4.8576642335766425e-06,
+      "loss": 0.0359,
+      "step": 1913
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 4.854014598540146e-06,
+      "loss": 0.0552,
+      "step": 1914
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 4.8503649635036494e-06,
+      "loss": 0.0119,
+      "step": 1915
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 4.846715328467153e-06,
+      "loss": 0.039,
+      "step": 1916
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 4.843065693430657e-06,
+      "loss": 0.0541,
+      "step": 1917
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 4.839416058394161e-06,
+      "loss": 0.042,
+      "step": 1918
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 4.835766423357665e-06,
+      "loss": 0.0333,
+      "step": 1919
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 4.832116788321168e-06,
+      "loss": 0.0297,
+      "step": 1920
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 4.828467153284672e-06,
+      "loss": 0.0357,
+      "step": 1921
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 4.824817518248176e-06,
+      "loss": 0.0395,
+      "step": 1922
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 4.82116788321168e-06,
+      "loss": 0.0235,
+      "step": 1923
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 4.8175182481751835e-06,
+      "loss": 0.0216,
+      "step": 1924
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 4.8138686131386866e-06,
+      "loss": 0.0419,
+      "step": 1925
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 4.8102189781021904e-06,
+      "loss": 0.0288,
+      "step": 1926
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 4.806569343065694e-06,
+      "loss": 0.0667,
+      "step": 1927
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 4.802919708029197e-06,
+      "loss": 0.013,
+      "step": 1928
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 4.799270072992701e-06,
+      "loss": 0.0432,
+      "step": 1929
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 4.795620437956205e-06,
+      "loss": 0.0619,
+      "step": 1930
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 4.791970802919708e-06,
+      "loss": 0.0188,
+      "step": 1931
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 4.788321167883212e-06,
+      "loss": 0.0524,
+      "step": 1932
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 4.784671532846716e-06,
+      "loss": 0.0353,
+      "step": 1933
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 4.78102189781022e-06,
+      "loss": 0.0333,
+      "step": 1934
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 4.777372262773723e-06,
+      "loss": 0.0174,
+      "step": 1935
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 4.773722627737227e-06,
+      "loss": 0.0674,
+      "step": 1936
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 4.770072992700731e-06,
+      "loss": 0.0237,
+      "step": 1937
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 4.766423357664234e-06,
+      "loss": 0.0232,
+      "step": 1938
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 4.7627737226277375e-06,
+      "loss": 0.0489,
+      "step": 1939
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 4.759124087591241e-06,
+      "loss": 0.0149,
+      "step": 1940
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 4.7554744525547444e-06,
+      "loss": 0.0374,
+      "step": 1941
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 4.751824817518248e-06,
+      "loss": 0.0223,
+      "step": 1942
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 4.748175182481752e-06,
+      "loss": 0.0228,
+      "step": 1943
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 4.744525547445255e-06,
+      "loss": 0.0321,
+      "step": 1944
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.740875912408759e-06,
+      "loss": 0.0499,
+      "step": 1945
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.737226277372263e-06,
+      "loss": 0.0448,
+      "step": 1946
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.733576642335767e-06,
+      "loss": 0.0569,
+      "step": 1947
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.729927007299271e-06,
+      "loss": 0.0257,
+      "step": 1948
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.726277372262774e-06,
+      "loss": 0.0215,
+      "step": 1949
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.722627737226278e-06,
+      "loss": 0.0479,
+      "step": 1950
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.7189781021897816e-06,
+      "loss": 0.0192,
+      "step": 1951
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.7153284671532855e-06,
+      "loss": 0.0282,
+      "step": 1952
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.7116788321167885e-06,
+      "loss": 0.0598,
+      "step": 1953
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.708029197080292e-06,
+      "loss": 0.0113,
+      "step": 1954
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.704379562043796e-06,
+      "loss": 0.0233,
+      "step": 1955
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.7007299270073e-06,
+      "loss": 0.0326,
+      "step": 1956
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.697080291970803e-06,
+      "loss": 0.046,
+      "step": 1957
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.693430656934307e-06,
+      "loss": 0.0273,
+      "step": 1958
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.689781021897811e-06,
+      "loss": 0.0294,
+      "step": 1959
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.686131386861315e-06,
+      "loss": 0.0323,
+      "step": 1960
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.682481751824818e-06,
+      "loss": 0.0224,
+      "step": 1961
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.678832116788322e-06,
+      "loss": 0.0261,
+      "step": 1962
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.675182481751826e-06,
+      "loss": 0.0489,
+      "step": 1963
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.671532846715329e-06,
+      "loss": 0.0223,
+      "step": 1964
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.6678832116788325e-06,
+      "loss": 0.0397,
+      "step": 1965
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.664233576642336e-06,
+      "loss": 0.0233,
+      "step": 1966
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.6605839416058395e-06,
+      "loss": 0.0244,
+      "step": 1967
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.656934306569343e-06,
+      "loss": 0.0375,
+      "step": 1968
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.653284671532847e-06,
+      "loss": 0.0321,
+      "step": 1969
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.64963503649635e-06,
+      "loss": 0.0437,
+      "step": 1970
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.645985401459854e-06,
+      "loss": 0.0324,
+      "step": 1971
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.642335766423358e-06,
+      "loss": 0.0497,
+      "step": 1972
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.638686131386862e-06,
+      "loss": 0.0169,
+      "step": 1973
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.635036496350365e-06,
+      "loss": 0.0748,
+      "step": 1974
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.631386861313869e-06,
+      "loss": 0.045,
+      "step": 1975
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.627737226277373e-06,
+      "loss": 0.0563,
+      "step": 1976
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.624087591240876e-06,
+      "loss": 0.0437,
+      "step": 1977
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.62043795620438e-06,
+      "loss": 0.0597,
+      "step": 1978
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.6167883211678835e-06,
+      "loss": 0.0196,
+      "step": 1979
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.613138686131387e-06,
+      "loss": 0.0344,
+      "step": 1980
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.609489051094891e-06,
+      "loss": 0.0326,
+      "step": 1981
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.605839416058394e-06,
+      "loss": 0.0479,
+      "step": 1982
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.602189781021898e-06,
+      "loss": 0.0228,
+      "step": 1983
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.598540145985402e-06,
+      "loss": 0.0518,
+      "step": 1984
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.594890510948906e-06,
+      "loss": 0.0235,
+      "step": 1985
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.59124087591241e-06,
+      "loss": 0.0206,
+      "step": 1986
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.587591240875913e-06,
+      "loss": 0.0408,
+      "step": 1987
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.583941605839417e-06,
+      "loss": 0.0216,
+      "step": 1988
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 4.580291970802921e-06,
+      "loss": 0.0266,
+      "step": 1989
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 4.576642335766424e-06,
+      "loss": 0.0286,
+      "step": 1990
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 4.5729927007299275e-06,
+      "loss": 0.0449,
+      "step": 1991
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 4.5693430656934314e-06,
+      "loss": 0.042,
+      "step": 1992
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 4.5656934306569345e-06,
+      "loss": 0.0325,
+      "step": 1993
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 4.562043795620438e-06,
+      "loss": 0.0275,
+      "step": 1994
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 4.558394160583942e-06,
+      "loss": 0.0199,
+      "step": 1995
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 4.554744525547445e-06,
+      "loss": 0.0563,
+      "step": 1996
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 4.551094890510949e-06,
+      "loss": 0.0629,
+      "step": 1997
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 4.547445255474453e-06,
+      "loss": 0.0207,
+      "step": 1998
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 4.543795620437956e-06,
+      "loss": 0.0135,
+      "step": 1999
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 4.54014598540146e-06,
+      "loss": 0.0524,
+      "step": 2000
+    },
+    {
+      "epoch": 18.51,
+      "eval_loss": 1.3919520378112793,
+      "eval_runtime": 64.859,
+      "eval_samples_per_second": 25.394,
+      "eval_steps_per_second": 1.588,
+      "eval_wer": 0.9603717699928506,
+      "step": 2000
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 4.536496350364964e-06,
+      "loss": 0.0083,
+      "step": 2001
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 4.532846715328468e-06,
+      "loss": 0.0401,
+      "step": 2002
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 4.529197080291971e-06,
+      "loss": 0.0355,
+      "step": 2003
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 4.525547445255475e-06,
+      "loss": 0.0471,
+      "step": 2004
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 4.5218978102189785e-06,
+      "loss": 0.0126,
+      "step": 2005
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 4.5182481751824815e-06,
+      "loss": 0.0359,
+      "step": 2006
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 4.5145985401459854e-06,
+      "loss": 0.0112,
+      "step": 2007
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 4.510948905109489e-06,
+      "loss": 0.0407,
+      "step": 2008
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 4.507299270072993e-06,
+      "loss": 0.0364,
+      "step": 2009
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 4.503649635036497e-06,
+      "loss": 0.0512,
+      "step": 2010
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 4.5e-06,
+      "loss": 0.0256,
+      "step": 2011
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 4.496350364963504e-06,
+      "loss": 0.0246,
+      "step": 2012
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 4.492700729927008e-06,
+      "loss": 0.029,
+      "step": 2013
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 4.489051094890512e-06,
+      "loss": 0.0315,
+      "step": 2014
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 4.485401459854015e-06,
+      "loss": 0.0676,
+      "step": 2015
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 4.481751824817519e-06,
+      "loss": 0.0407,
+      "step": 2016
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 4.4781021897810226e-06,
+      "loss": 0.0118,
+      "step": 2017
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 4.4744525547445264e-06,
+      "loss": 0.0408,
+      "step": 2018
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 4.4708029197080295e-06,
+      "loss": 0.0657,
+      "step": 2019
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 4.467153284671533e-06,
+      "loss": 0.0148,
+      "step": 2020
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 4.463503649635037e-06,
+      "loss": 0.0249,
+      "step": 2021
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 4.45985401459854e-06,
+      "loss": 0.1062,
+      "step": 2022
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 4.456204379562044e-06,
+      "loss": 0.0187,
+      "step": 2023
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 4.452554744525548e-06,
+      "loss": 0.0058,
+      "step": 2024
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 4.448905109489051e-06,
+      "loss": 0.0479,
+      "step": 2025
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 4.445255474452555e-06,
+      "loss": 0.0247,
+      "step": 2026
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 4.441605839416059e-06,
+      "loss": 0.026,
+      "step": 2027
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 4.437956204379563e-06,
+      "loss": 0.0457,
+      "step": 2028
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 4.434306569343066e-06,
+      "loss": 0.0665,
+      "step": 2029
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 4.43065693430657e-06,
+      "loss": 0.0216,
+      "step": 2030
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 4.4270072992700735e-06,
+      "loss": 0.0536,
+      "step": 2031
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 4.4233576642335766e-06,
+      "loss": 0.036,
+      "step": 2032
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 4.4197080291970804e-06,
+      "loss": 0.0444,
+      "step": 2033
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 4.416058394160584e-06,
+      "loss": 0.0418,
+      "step": 2034
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 4.412408759124087e-06,
+      "loss": 0.0432,
+      "step": 2035
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 4.408759124087591e-06,
+      "loss": 0.0078,
+      "step": 2036
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 4.405109489051095e-06,
+      "loss": 0.0155,
+      "step": 2037
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 4.401459854014599e-06,
+      "loss": 0.0353,
+      "step": 2038
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 4.397810218978102e-06,
+      "loss": 0.0217,
+      "step": 2039
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 4.394160583941606e-06,
+      "loss": 0.0363,
+      "step": 2040
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 4.39051094890511e-06,
+      "loss": 0.0627,
+      "step": 2041
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 4.386861313868614e-06,
+      "loss": 0.0211,
+      "step": 2042
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 4.3832116788321176e-06,
+      "loss": 0.0262,
+      "step": 2043
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 4.379562043795621e-06,
+      "loss": 0.0401,
+      "step": 2044
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 4.3759124087591245e-06,
+      "loss": 0.0281,
+      "step": 2045
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 4.372262773722628e-06,
+      "loss": 0.0522,
+      "step": 2046
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 4.368613138686132e-06,
+      "loss": 0.0212,
+      "step": 2047
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 4.364963503649635e-06,
+      "loss": 0.0194,
+      "step": 2048
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 4.361313868613139e-06,
+      "loss": 0.0545,
+      "step": 2049
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 4.357664233576643e-06,
+      "loss": 0.0297,
+      "step": 2050
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 4.354014598540146e-06,
+      "loss": 0.0122,
+      "step": 2051
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 4.35036496350365e-06,
+      "loss": 0.0254,
+      "step": 2052
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 4.346715328467154e-06,
+      "loss": 0.04,
+      "step": 2053
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 4.343065693430658e-06,
+      "loss": 0.0285,
+      "step": 2054
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 4.339416058394161e-06,
+      "loss": 0.0383,
+      "step": 2055
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 4.335766423357665e-06,
+      "loss": 0.0115,
+      "step": 2056
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 4.3321167883211685e-06,
+      "loss": 0.0076,
+      "step": 2057
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 4.328467153284672e-06,
+      "loss": 0.0658,
+      "step": 2058
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 4.3248175182481755e-06,
+      "loss": 0.0454,
+      "step": 2059
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 4.321167883211679e-06,
+      "loss": 0.0154,
+      "step": 2060
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 4.317518248175182e-06,
+      "loss": 0.0407,
+      "step": 2061
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 4.313868613138686e-06,
+      "loss": 0.0837,
+      "step": 2062
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 4.31021897810219e-06,
+      "loss": 0.0167,
+      "step": 2063
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 4.306569343065693e-06,
+      "loss": 0.0257,
+      "step": 2064
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 4.302919708029197e-06,
+      "loss": 0.0363,
+      "step": 2065
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 4.299270072992701e-06,
+      "loss": 0.0328,
+      "step": 2066
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 4.295620437956205e-06,
+      "loss": 0.0392,
+      "step": 2067
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 4.291970802919708e-06,
+      "loss": 0.0458,
+      "step": 2068
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 4.288321167883212e-06,
+      "loss": 0.0111,
+      "step": 2069
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 4.284671532846716e-06,
+      "loss": 0.0174,
+      "step": 2070
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 4.2810218978102195e-06,
+      "loss": 0.0394,
+      "step": 2071
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 4.277372262773723e-06,
+      "loss": 0.0501,
+      "step": 2072
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 4.2737226277372264e-06,
+      "loss": 0.0231,
+      "step": 2073
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 4.27007299270073e-06,
+      "loss": 0.0463,
+      "step": 2074
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 4.266423357664234e-06,
+      "loss": 0.0312,
+      "step": 2075
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 4.262773722627738e-06,
+      "loss": 0.0339,
+      "step": 2076
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 4.259124087591241e-06,
+      "loss": 0.0313,
+      "step": 2077
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 4.255474452554745e-06,
+      "loss": 0.0329,
+      "step": 2078
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 4.251824817518249e-06,
+      "loss": 0.0243,
+      "step": 2079
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 4.248175182481753e-06,
+      "loss": 0.0534,
+      "step": 2080
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 4.244525547445256e-06,
+      "loss": 0.0261,
+      "step": 2081
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 4.24087591240876e-06,
+      "loss": 0.0301,
+      "step": 2082
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 4.2372262773722636e-06,
+      "loss": 0.0502,
+      "step": 2083
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 4.233576642335767e-06,
+      "loss": 0.0327,
+      "step": 2084
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 4.2299270072992705e-06,
+      "loss": 0.0137,
+      "step": 2085
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 4.226277372262774e-06,
+      "loss": 0.0366,
+      "step": 2086
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 4.222627737226277e-06,
+      "loss": 0.0507,
+      "step": 2087
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 4.218978102189781e-06,
+      "loss": 0.0189,
+      "step": 2088
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 4.215328467153285e-06,
+      "loss": 0.0236,
+      "step": 2089
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 4.211678832116788e-06,
+      "loss": 0.0962,
+      "step": 2090
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 4.208029197080292e-06,
+      "loss": 0.0201,
+      "step": 2091
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 4.204379562043796e-06,
+      "loss": 0.0259,
+      "step": 2092
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 4.2007299270073e-06,
+      "loss": 0.0432,
+      "step": 2093
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 4.197080291970803e-06,
+      "loss": 0.0304,
+      "step": 2094
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 4.193430656934307e-06,
+      "loss": 0.0137,
+      "step": 2095
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 4.189781021897811e-06,
+      "loss": 0.0285,
+      "step": 2096
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 4.186131386861314e-06,
+      "loss": 0.0216,
+      "step": 2097
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 4.1824817518248176e-06,
+      "loss": 0.0468,
+      "step": 2098
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 4.1788321167883214e-06,
+      "loss": 0.0631,
+      "step": 2099
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 4.175182481751825e-06,
+      "loss": 0.0142,
+      "step": 2100
+    },
+    {
+      "epoch": 19.44,
+      "eval_loss": 1.422370433807373,
+      "eval_runtime": 64.842,
+      "eval_samples_per_second": 25.4,
+      "eval_steps_per_second": 1.588,
+      "eval_wer": 0.9697681544275355,
+      "step": 2100
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 4.171532846715328e-06,
+      "loss": 0.0519,
+      "step": 2101
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 4.167883211678832e-06,
+      "loss": 0.0135,
+      "step": 2102
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 4.164233576642336e-06,
+      "loss": 0.0232,
+      "step": 2103
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 4.16058394160584e-06,
+      "loss": 0.0444,
+      "step": 2104
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 4.156934306569344e-06,
+      "loss": 0.0462,
+      "step": 2105
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 4.153284671532847e-06,
+      "loss": 0.0313,
+      "step": 2106
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 4.149635036496351e-06,
+      "loss": 0.0497,
+      "step": 2107
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 4.145985401459855e-06,
+      "loss": 0.0509,
+      "step": 2108
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 4.1423357664233586e-06,
+      "loss": 0.0125,
+      "step": 2109
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 4.138686131386862e-06,
+      "loss": 0.0609,
+      "step": 2110
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 4.1350364963503655e-06,
+      "loss": 0.0693,
+      "step": 2111
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 4.131386861313869e-06,
+      "loss": 0.1316,
+      "step": 2112
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 4.127737226277372e-06,
+      "loss": 0.0297,
+      "step": 2113
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 4.124087591240876e-06,
+      "loss": 0.0436,
+      "step": 2114
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 4.12043795620438e-06,
+      "loss": 0.0283,
+      "step": 2115
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 4.116788321167883e-06,
+      "loss": 0.0362,
+      "step": 2116
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 4.113138686131387e-06,
+      "loss": 0.0769,
+      "step": 2117
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 4.109489051094891e-06,
+      "loss": 0.0558,
+      "step": 2118
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 4.105839416058395e-06,
+      "loss": 0.0194,
+      "step": 2119
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 4.102189781021898e-06,
+      "loss": 0.0304,
+      "step": 2120
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 4.098540145985402e-06,
+      "loss": 0.0532,
+      "step": 2121
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 4.094890510948906e-06,
+      "loss": 0.0148,
+      "step": 2122
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 4.091240875912409e-06,
+      "loss": 0.0166,
+      "step": 2123
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 4.0875912408759126e-06,
+      "loss": 0.043,
+      "step": 2124
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 4.0839416058394165e-06,
+      "loss": 0.0127,
+      "step": 2125
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 4.0802919708029195e-06,
+      "loss": 0.0576,
+      "step": 2126
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.076642335766423e-06,
+      "loss": 0.0214,
+      "step": 2127
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.072992700729927e-06,
+      "loss": 0.0345,
+      "step": 2128
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.069343065693431e-06,
+      "loss": 0.0303,
+      "step": 2129
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.065693430656934e-06,
+      "loss": 0.0489,
+      "step": 2130
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.062043795620438e-06,
+      "loss": 0.0186,
+      "step": 2131
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.058394160583942e-06,
+      "loss": 0.0225,
+      "step": 2132
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.054744525547446e-06,
+      "loss": 0.0461,
+      "step": 2133
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 4.05109489051095e-06,
+      "loss": 0.0462,
+      "step": 2134
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 4.047445255474453e-06,
+      "loss": 0.0198,
+      "step": 2135
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 4.043795620437957e-06,
+      "loss": 0.0475,
+      "step": 2136
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 4.0401459854014605e-06,
+      "loss": 0.0155,
+      "step": 2137
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 4.036496350364964e-06,
+      "loss": 0.0553,
+      "step": 2138
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.032846715328467e-06,
+      "loss": 0.047,
+      "step": 2139
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.029197080291971e-06,
+      "loss": 0.0424,
+      "step": 2140
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.025547445255475e-06,
+      "loss": 0.016,
+      "step": 2141
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.021897810218978e-06,
+      "loss": 0.048,
+      "step": 2142
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.018248175182482e-06,
+      "loss": 0.0286,
+      "step": 2143
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 4.014598540145986e-06,
+      "loss": 0.0207,
+      "step": 2144
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 4.01094890510949e-06,
+      "loss": 0.0424,
+      "step": 2145
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 4.007299270072993e-06,
+      "loss": 0.0451,
+      "step": 2146
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 4.003649635036497e-06,
+      "loss": 0.0115,
+      "step": 2147
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.0218,
+      "step": 2148
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 3.996350364963504e-06,
+      "loss": 0.0791,
+      "step": 2149
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 3.992700729927008e-06,
+      "loss": 0.0223,
+      "step": 2150
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 3.9890510948905115e-06,
+      "loss": 0.0323,
+      "step": 2151
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 3.9854014598540145e-06,
+      "loss": 0.0691,
+      "step": 2152
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 3.981751824817518e-06,
+      "loss": 0.012,
+      "step": 2153
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 3.978102189781022e-06,
+      "loss": 0.019,
+      "step": 2154
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 3.974452554744525e-06,
+      "loss": 0.0127,
+      "step": 2155
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 3.970802919708029e-06,
+      "loss": 0.0224,
+      "step": 2156
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 3.967153284671533e-06,
+      "loss": 0.0347,
+      "step": 2157
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 3.963503649635037e-06,
+      "loss": 0.0243,
+      "step": 2158
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 3.95985401459854e-06,
+      "loss": 0.0155,
+      "step": 2159
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.956204379562044e-06,
+      "loss": 0.0437,
+      "step": 2160
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.952554744525548e-06,
+      "loss": 0.0378,
+      "step": 2161
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.948905109489052e-06,
+      "loss": 0.0068,
+      "step": 2162
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.945255474452555e-06,
+      "loss": 0.0299,
+      "step": 2163
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.9416058394160585e-06,
+      "loss": 0.0094,
+      "step": 2164
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.9379562043795624e-06,
+      "loss": 0.0405,
+      "step": 2165
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.934306569343066e-06,
+      "loss": 0.0429,
+      "step": 2166
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.93065693430657e-06,
+      "loss": 0.0466,
+      "step": 2167
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.927007299270073e-06,
+      "loss": 0.0295,
+      "step": 2168
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.923357664233577e-06,
+      "loss": 0.0717,
+      "step": 2169
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.919708029197081e-06,
+      "loss": 0.035,
+      "step": 2170
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.916058394160585e-06,
+      "loss": 0.0442,
+      "step": 2171
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.912408759124088e-06,
+      "loss": 0.0357,
+      "step": 2172
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.908759124087592e-06,
+      "loss": 0.0399,
+      "step": 2173
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.905109489051096e-06,
+      "loss": 0.0122,
+      "step": 2174
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.901459854014599e-06,
+      "loss": 0.0517,
+      "step": 2175
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.897810218978103e-06,
+      "loss": 0.0443,
+      "step": 2176
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.8941605839416065e-06,
+      "loss": 0.0314,
+      "step": 2177
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.8905109489051095e-06,
+      "loss": 0.0318,
+      "step": 2178
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.886861313868613e-06,
+      "loss": 0.039,
+      "step": 2179
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.883211678832117e-06,
+      "loss": 0.0266,
+      "step": 2180
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.87956204379562e-06,
+      "loss": 0.0104,
+      "step": 2181
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.875912408759124e-06,
+      "loss": 0.0237,
+      "step": 2182
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.872262773722628e-06,
+      "loss": 0.0184,
+      "step": 2183
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.868613138686132e-06,
+      "loss": 0.0646,
+      "step": 2184
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.864963503649635e-06,
+      "loss": 0.0199,
+      "step": 2185
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.861313868613139e-06,
+      "loss": 0.0154,
+      "step": 2186
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.857664233576643e-06,
+      "loss": 0.0392,
+      "step": 2187
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.854014598540146e-06,
+      "loss": 0.0518,
+      "step": 2188
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.85036496350365e-06,
+      "loss": 0.0438,
+      "step": 2189
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.8467153284671536e-06,
+      "loss": 0.019,
+      "step": 2190
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.8430656934306574e-06,
+      "loss": 0.0191,
+      "step": 2191
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.8394160583941605e-06,
+      "loss": 0.0127,
+      "step": 2192
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.835766423357664e-06,
+      "loss": 0.0217,
+      "step": 2193
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.832116788321168e-06,
+      "loss": 0.03,
+      "step": 2194
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.828467153284672e-06,
+      "loss": 0.088,
+      "step": 2195
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.824817518248176e-06,
+      "loss": 0.02,
+      "step": 2196
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.821167883211679e-06,
+      "loss": 0.062,
+      "step": 2197
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.817518248175183e-06,
+      "loss": 0.0426,
+      "step": 2198
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.8138686131386864e-06,
+      "loss": 0.0337,
+      "step": 2199
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.8102189781021903e-06,
+      "loss": 0.0288,
+      "step": 2200
+    },
+    {
+      "epoch": 20.37,
+      "eval_loss": 1.3474684953689575,
+      "eval_runtime": 64.717,
+      "eval_samples_per_second": 25.449,
+      "eval_steps_per_second": 1.592,
+      "eval_wer": 0.9635379430088857,
+      "step": 2200
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.8065693430656937e-06,
+      "loss": 0.0667,
+      "step": 2201
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.8029197080291976e-06,
+      "loss": 0.0235,
+      "step": 2202
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.799270072992701e-06,
+      "loss": 0.0185,
+      "step": 2203
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.7956204379562045e-06,
+      "loss": 0.0179,
+      "step": 2204
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.7919708029197084e-06,
+      "loss": 0.0079,
+      "step": 2205
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.7883211678832123e-06,
+      "loss": 0.0378,
+      "step": 2206
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.7846715328467153e-06,
+      "loss": 0.0333,
+      "step": 2207
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.7810218978102192e-06,
+      "loss": 0.022,
+      "step": 2208
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.777372262773723e-06,
+      "loss": 0.0333,
+      "step": 2209
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.773722627737227e-06,
+      "loss": 0.0397,
+      "step": 2210
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.77007299270073e-06,
+      "loss": 0.029,
+      "step": 2211
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.766423357664234e-06,
+      "loss": 0.0463,
+      "step": 2212
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.7627737226277378e-06,
+      "loss": 0.0908,
+      "step": 2213
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.7591240875912412e-06,
+      "loss": 0.0285,
+      "step": 2214
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.7554744525547447e-06,
+      "loss": 0.0245,
+      "step": 2215
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.7518248175182486e-06,
+      "loss": 0.0204,
+      "step": 2216
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.748175182481752e-06,
+      "loss": 0.0133,
+      "step": 2217
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.744525547445256e-06,
+      "loss": 0.0124,
+      "step": 2218
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.7408759124087594e-06,
+      "loss": 0.0334,
+      "step": 2219
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.737226277372263e-06,
+      "loss": 0.0214,
+      "step": 2220
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.7335766423357667e-06,
+      "loss": 0.0383,
+      "step": 2221
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.7299270072992706e-06,
+      "loss": 0.0329,
+      "step": 2222
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.7262773722627736e-06,
+      "loss": 0.023,
+      "step": 2223
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.7226277372262775e-06,
+      "loss": 0.0308,
+      "step": 2224
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.7189781021897814e-06,
+      "loss": 0.0504,
+      "step": 2225
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.7153284671532853e-06,
+      "loss": 0.0535,
+      "step": 2226
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.7116788321167883e-06,
+      "loss": 0.047,
+      "step": 2227
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.708029197080292e-06,
+      "loss": 0.0314,
+      "step": 2228
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.704379562043796e-06,
+      "loss": 0.0675,
+      "step": 2229
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.7007299270072995e-06,
+      "loss": 0.0089,
+      "step": 2230
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.697080291970803e-06,
+      "loss": 0.0512,
+      "step": 2231
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.693430656934307e-06,
+      "loss": 0.0434,
+      "step": 2232
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.6897810218978103e-06,
+      "loss": 0.0525,
+      "step": 2233
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.6861313868613142e-06,
+      "loss": 0.0087,
+      "step": 2234
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.682481751824818e-06,
+      "loss": 0.0458,
+      "step": 2235
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.678832116788321e-06,
+      "loss": 0.0191,
+      "step": 2236
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.675182481751825e-06,
+      "loss": 0.0033,
+      "step": 2237
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.671532846715329e-06,
+      "loss": 0.0586,
+      "step": 2238
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.667883211678833e-06,
+      "loss": 0.0139,
+      "step": 2239
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.664233576642336e-06,
+      "loss": 0.0319,
+      "step": 2240
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.6605839416058397e-06,
+      "loss": 0.0352,
+      "step": 2241
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.6569343065693436e-06,
+      "loss": 0.0309,
+      "step": 2242
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.6532846715328466e-06,
+      "loss": 0.0735,
+      "step": 2243
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.6496350364963505e-06,
+      "loss": 0.0228,
+      "step": 2244
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.6459854014598544e-06,
+      "loss": 0.045,
+      "step": 2245
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.642335766423358e-06,
+      "loss": 0.0473,
+      "step": 2246
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.6386861313868617e-06,
+      "loss": 0.0326,
+      "step": 2247
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.635036496350365e-06,
+      "loss": 0.031,
+      "step": 2248
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.6313868613138687e-06,
+      "loss": 0.0396,
+      "step": 2249
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.6277372262773725e-06,
+      "loss": 0.0151,
+      "step": 2250
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.6240875912408764e-06,
+      "loss": 0.0278,
+      "step": 2251
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.6204379562043803e-06,
+      "loss": 0.0205,
+      "step": 2252
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.6167883211678833e-06,
+      "loss": 0.0705,
+      "step": 2253
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.6131386861313872e-06,
+      "loss": 0.0661,
+      "step": 2254
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.609489051094891e-06,
+      "loss": 0.0313,
+      "step": 2255
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.605839416058394e-06,
+      "loss": 0.0282,
+      "step": 2256
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.602189781021898e-06,
+      "loss": 0.0557,
+      "step": 2257
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.598540145985402e-06,
+      "loss": 0.0339,
+      "step": 2258
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.5948905109489054e-06,
+      "loss": 0.0236,
+      "step": 2259
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.591240875912409e-06,
+      "loss": 0.0288,
+      "step": 2260
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.5875912408759127e-06,
+      "loss": 0.0115,
+      "step": 2261
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.583941605839416e-06,
+      "loss": 0.0277,
+      "step": 2262
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.58029197080292e-06,
+      "loss": 0.0328,
+      "step": 2263
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.576642335766424e-06,
+      "loss": 0.0131,
+      "step": 2264
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.5729927007299274e-06,
+      "loss": 0.0171,
+      "step": 2265
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.569343065693431e-06,
+      "loss": 0.0921,
+      "step": 2266
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.5656934306569347e-06,
+      "loss": 0.0534,
+      "step": 2267
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.5620437956204386e-06,
+      "loss": 0.0333,
+      "step": 2268
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.5583941605839416e-06,
+      "loss": 0.0367,
+      "step": 2269
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.5547445255474455e-06,
+      "loss": 0.0157,
+      "step": 2270
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.5510948905109494e-06,
+      "loss": 0.0548,
+      "step": 2271
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.5474452554744524e-06,
+      "loss": 0.0432,
+      "step": 2272
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.5437956204379563e-06,
+      "loss": 0.0243,
+      "step": 2273
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.54014598540146e-06,
+      "loss": 0.0221,
+      "step": 2274
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.5364963503649637e-06,
+      "loss": 0.0163,
+      "step": 2275
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.5328467153284675e-06,
+      "loss": 0.0142,
+      "step": 2276
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.529197080291971e-06,
+      "loss": 0.041,
+      "step": 2277
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.525547445255475e-06,
+      "loss": 0.024,
+      "step": 2278
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.5218978102189783e-06,
+      "loss": 0.0367,
+      "step": 2279
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.5182481751824822e-06,
+      "loss": 0.0579,
+      "step": 2280
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.5145985401459857e-06,
+      "loss": 0.0184,
+      "step": 2281
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.510948905109489e-06,
+      "loss": 0.0144,
+      "step": 2282
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.507299270072993e-06,
+      "loss": 0.028,
+      "step": 2283
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.503649635036497e-06,
+      "loss": 0.012,
+      "step": 2284
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.5e-06,
+      "loss": 0.0332,
+      "step": 2285
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.496350364963504e-06,
+      "loss": 0.0158,
+      "step": 2286
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.4927007299270077e-06,
+      "loss": 0.01,
+      "step": 2287
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.489051094890511e-06,
+      "loss": 0.014,
+      "step": 2288
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.4854014598540146e-06,
+      "loss": 0.016,
+      "step": 2289
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.4817518248175185e-06,
+      "loss": 0.0236,
+      "step": 2290
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.4781021897810224e-06,
+      "loss": 0.0183,
+      "step": 2291
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.474452554744526e-06,
+      "loss": 0.0112,
+      "step": 2292
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.4708029197080293e-06,
+      "loss": 0.0428,
+      "step": 2293
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.467153284671533e-06,
+      "loss": 0.0559,
+      "step": 2294
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.4635036496350367e-06,
+      "loss": 0.0137,
+      "step": 2295
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.4598540145985405e-06,
+      "loss": 0.0298,
+      "step": 2296
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.4562043795620444e-06,
+      "loss": 0.0121,
+      "step": 2297
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.4525547445255475e-06,
+      "loss": 0.0093,
+      "step": 2298
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.4489051094890513e-06,
+      "loss": 0.0262,
+      "step": 2299
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.4452554744525552e-06,
+      "loss": 0.0106,
+      "step": 2300
+    },
+    {
+      "epoch": 21.29,
+      "eval_loss": 1.2232261896133423,
+      "eval_runtime": 64.8459,
+      "eval_samples_per_second": 25.399,
+      "eval_steps_per_second": 1.588,
+      "eval_wer": 0.926360943723828,
+      "step": 2300
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.4416058394160583e-06,
+      "loss": 0.0405,
+      "step": 2301
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.437956204379562e-06,
+      "loss": 0.0106,
+      "step": 2302
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.434306569343066e-06,
+      "loss": 0.0485,
+      "step": 2303
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.43065693430657e-06,
+      "loss": 0.0098,
+      "step": 2304
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.427007299270073e-06,
+      "loss": 0.0484,
+      "step": 2305
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.423357664233577e-06,
+      "loss": 0.0477,
+      "step": 2306
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.4197080291970807e-06,
+      "loss": 0.0187,
+      "step": 2307
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.416058394160584e-06,
+      "loss": 0.0703,
+      "step": 2308
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.412408759124088e-06,
+      "loss": 0.0512,
+      "step": 2309
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.4087591240875915e-06,
+      "loss": 0.0198,
+      "step": 2310
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.405109489051095e-06,
+      "loss": 0.0347,
+      "step": 2311
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.401459854014599e-06,
+      "loss": 0.0999,
+      "step": 2312
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.3978102189781027e-06,
+      "loss": 0.0194,
+      "step": 2313
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.3941605839416058e-06,
+      "loss": 0.0168,
+      "step": 2314
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.3905109489051096e-06,
+      "loss": 0.0228,
+      "step": 2315
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.3868613138686135e-06,
+      "loss": 0.0338,
+      "step": 2316
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.3832116788321174e-06,
+      "loss": 0.0704,
+      "step": 2317
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.3795620437956204e-06,
+      "loss": 0.0625,
+      "step": 2318
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.3759124087591243e-06,
+      "loss": 0.008,
+      "step": 2319
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.372262773722628e-06,
+      "loss": 0.0342,
+      "step": 2320
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.3686131386861317e-06,
+      "loss": 0.0328,
+      "step": 2321
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.364963503649635e-06,
+      "loss": 0.011,
+      "step": 2322
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.361313868613139e-06,
+      "loss": 0.0174,
+      "step": 2323
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.3576642335766425e-06,
+      "loss": 0.0651,
+      "step": 2324
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.3540145985401464e-06,
+      "loss": 0.0225,
+      "step": 2325
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.3503649635036502e-06,
+      "loss": 0.057,
+      "step": 2326
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.3467153284671533e-06,
+      "loss": 0.0311,
+      "step": 2327
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.343065693430657e-06,
+      "loss": 0.0395,
+      "step": 2328
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.339416058394161e-06,
+      "loss": 0.0131,
+      "step": 2329
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.335766423357665e-06,
+      "loss": 0.0239,
+      "step": 2330
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.332116788321168e-06,
+      "loss": 0.1472,
+      "step": 2331
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.328467153284672e-06,
+      "loss": 0.0255,
+      "step": 2332
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.3248175182481757e-06,
+      "loss": 0.0274,
+      "step": 2333
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.3211678832116788e-06,
+      "loss": 0.0181,
+      "step": 2334
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.3175182481751826e-06,
+      "loss": 0.0209,
+      "step": 2335
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.3138686131386865e-06,
+      "loss": 0.0276,
+      "step": 2336
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.31021897810219e-06,
+      "loss": 0.0297,
+      "step": 2337
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.306569343065694e-06,
+      "loss": 0.0368,
+      "step": 2338
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.3029197080291973e-06,
+      "loss": 0.0246,
+      "step": 2339
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.2992700729927008e-06,
+      "loss": 0.0496,
+      "step": 2340
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.2956204379562047e-06,
+      "loss": 0.0154,
+      "step": 2341
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.2919708029197085e-06,
+      "loss": 0.0185,
+      "step": 2342
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.288321167883212e-06,
+      "loss": 0.0465,
+      "step": 2343
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.2846715328467155e-06,
+      "loss": 0.0167,
+      "step": 2344
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.2810218978102193e-06,
+      "loss": 0.0136,
+      "step": 2345
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.2773722627737232e-06,
+      "loss": 0.053,
+      "step": 2346
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.2737226277372263e-06,
+      "loss": 0.0393,
+      "step": 2347
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.27007299270073e-06,
+      "loss": 0.0206,
+      "step": 2348
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.266423357664234e-06,
+      "loss": 0.0528,
+      "step": 2349
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.2627737226277375e-06,
+      "loss": 0.0262,
+      "step": 2350
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.259124087591241e-06,
+      "loss": 0.0347,
+      "step": 2351
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.255474452554745e-06,
+      "loss": 0.0226,
+      "step": 2352
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.2518248175182483e-06,
+      "loss": 0.0349,
+      "step": 2353
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.248175182481752e-06,
+      "loss": 0.0304,
+      "step": 2354
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.2445255474452556e-06,
+      "loss": 0.0251,
+      "step": 2355
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.2408759124087595e-06,
+      "loss": 0.0544,
+      "step": 2356
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.237226277372263e-06,
+      "loss": 0.0405,
+      "step": 2357
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.233576642335767e-06,
+      "loss": 0.026,
+      "step": 2358
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.2299270072992707e-06,
+      "loss": 0.0211,
+      "step": 2359
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.2262773722627738e-06,
+      "loss": 0.0298,
+      "step": 2360
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.2226277372262776e-06,
+      "loss": 0.0902,
+      "step": 2361
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.2189781021897815e-06,
+      "loss": 0.0487,
+      "step": 2362
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.2153284671532846e-06,
+      "loss": 0.0563,
+      "step": 2363
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.2116788321167884e-06,
+      "loss": 0.0233,
+      "step": 2364
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.2080291970802923e-06,
+      "loss": 0.0494,
+      "step": 2365
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.204379562043796e-06,
+      "loss": 0.0152,
+      "step": 2366
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.2007299270072992e-06,
+      "loss": 0.0118,
+      "step": 2367
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.197080291970803e-06,
+      "loss": 0.0066,
+      "step": 2368
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.193430656934307e-06,
+      "loss": 0.0283,
+      "step": 2369
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.1897810218978105e-06,
+      "loss": 0.0329,
+      "step": 2370
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.1861313868613144e-06,
+      "loss": 0.0603,
+      "step": 2371
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.182481751824818e-06,
+      "loss": 0.0416,
+      "step": 2372
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.1788321167883213e-06,
+      "loss": 0.0305,
+      "step": 2373
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.175182481751825e-06,
+      "loss": 0.0518,
+      "step": 2374
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.171532846715329e-06,
+      "loss": 0.0434,
+      "step": 2375
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.167883211678832e-06,
+      "loss": 0.0241,
+      "step": 2376
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.164233576642336e-06,
+      "loss": 0.1049,
+      "step": 2377
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.16058394160584e-06,
+      "loss": 0.0109,
+      "step": 2378
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.156934306569343e-06,
+      "loss": 0.0619,
+      "step": 2379
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.1532846715328468e-06,
+      "loss": 0.0157,
+      "step": 2380
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.1496350364963506e-06,
+      "loss": 0.0236,
+      "step": 2381
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.1459854014598545e-06,
+      "loss": 0.0573,
+      "step": 2382
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.142335766423358e-06,
+      "loss": 0.0353,
+      "step": 2383
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.1386861313868614e-06,
+      "loss": 0.0265,
+      "step": 2384
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.1350364963503653e-06,
+      "loss": 0.019,
+      "step": 2385
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.1313868613138688e-06,
+      "loss": 0.0229,
+      "step": 2386
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.1277372262773727e-06,
+      "loss": 0.0123,
+      "step": 2387
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.1240875912408765e-06,
+      "loss": 0.026,
+      "step": 2388
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.1204379562043796e-06,
+      "loss": 0.0216,
+      "step": 2389
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.1167883211678835e-06,
+      "loss": 0.0424,
+      "step": 2390
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.1131386861313873e-06,
+      "loss": 0.0142,
+      "step": 2391
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.1094890510948904e-06,
+      "loss": 0.036,
+      "step": 2392
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.1058394160583943e-06,
+      "loss": 0.0499,
+      "step": 2393
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.102189781021898e-06,
+      "loss": 0.0471,
+      "step": 2394
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.098540145985402e-06,
+      "loss": 0.0317,
+      "step": 2395
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.094890510948905e-06,
+      "loss": 0.0129,
+      "step": 2396
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.091240875912409e-06,
+      "loss": 0.049,
+      "step": 2397
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.087591240875913e-06,
+      "loss": 0.0492,
+      "step": 2398
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.0839416058394163e-06,
+      "loss": 0.0147,
+      "step": 2399
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.08029197080292e-06,
+      "loss": 0.0396,
+      "step": 2400
+    },
+    {
+      "epoch": 22.22,
+      "eval_loss": 1.33225417137146,
+      "eval_runtime": 64.6702,
+      "eval_samples_per_second": 25.468,
+      "eval_steps_per_second": 1.593,
+      "eval_wer": 0.9614952507404759,
+      "step": 2400
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.0766423357664236e-06,
+      "loss": 0.0513,
+      "step": 2401
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.072992700729927e-06,
+      "loss": 0.0367,
+      "step": 2402
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.069343065693431e-06,
+      "loss": 0.0136,
+      "step": 2403
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.065693430656935e-06,
+      "loss": 0.0599,
+      "step": 2404
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.062043795620438e-06,
+      "loss": 0.0143,
+      "step": 2405
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.0583941605839418e-06,
+      "loss": 0.0182,
+      "step": 2406
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.0547445255474457e-06,
+      "loss": 0.0529,
+      "step": 2407
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.0510948905109495e-06,
+      "loss": 0.0464,
+      "step": 2408
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.0474452554744526e-06,
+      "loss": 0.0237,
+      "step": 2409
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.0437956204379565e-06,
+      "loss": 0.0323,
+      "step": 2410
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.0401459854014603e-06,
+      "loss": 0.0643,
+      "step": 2411
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.036496350364964e-06,
+      "loss": 0.029,
+      "step": 2412
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.0328467153284673e-06,
+      "loss": 0.0737,
+      "step": 2413
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.029197080291971e-06,
+      "loss": 0.038,
+      "step": 2414
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.0255474452554746e-06,
+      "loss": 0.021,
+      "step": 2415
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.0218978102189785e-06,
+      "loss": 0.0437,
+      "step": 2416
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.018248175182482e-06,
+      "loss": 0.0124,
+      "step": 2417
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.0145985401459854e-06,
+      "loss": 0.0065,
+      "step": 2418
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.0109489051094893e-06,
+      "loss": 0.0291,
+      "step": 2419
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.007299270072993e-06,
+      "loss": 0.043,
+      "step": 2420
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.003649635036497e-06,
+      "loss": 0.0088,
+      "step": 2421
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3e-06,
+      "loss": 0.0318,
+      "step": 2422
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 2.996350364963504e-06,
+      "loss": 0.0444,
+      "step": 2423
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 2.992700729927008e-06,
+      "loss": 0.0152,
+      "step": 2424
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 2.989051094890511e-06,
+      "loss": 0.0149,
+      "step": 2425
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 2.9854014598540148e-06,
+      "loss": 0.0426,
+      "step": 2426
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 2.9817518248175186e-06,
+      "loss": 0.0186,
+      "step": 2427
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 2.978102189781022e-06,
+      "loss": 0.0263,
+      "step": 2428
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 2.9744525547445256e-06,
+      "loss": 0.0219,
+      "step": 2429
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 2.9708029197080294e-06,
+      "loss": 0.0072,
+      "step": 2430
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 2.967153284671533e-06,
+      "loss": 0.027,
+      "step": 2431
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 2.9635036496350368e-06,
+      "loss": 0.0334,
+      "step": 2432
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 2.9598540145985407e-06,
+      "loss": 0.0272,
+      "step": 2433
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 2.956204379562044e-06,
+      "loss": 0.0188,
+      "step": 2434
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 2.9525547445255476e-06,
+      "loss": 0.0084,
+      "step": 2435
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 2.9489051094890515e-06,
+      "loss": 0.017,
+      "step": 2436
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 2.9452554744525553e-06,
+      "loss": 0.0435,
+      "step": 2437
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 2.9416058394160584e-06,
+      "loss": 0.0361,
+      "step": 2438
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 2.9379562043795623e-06,
+      "loss": 0.0133,
+      "step": 2439
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 2.934306569343066e-06,
+      "loss": 0.0091,
+      "step": 2440
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 2.930656934306569e-06,
+      "loss": 0.046,
+      "step": 2441
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 2.927007299270073e-06,
+      "loss": 0.0552,
+      "step": 2442
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 2.923357664233577e-06,
+      "loss": 0.014,
+      "step": 2443
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 2.9197080291970804e-06,
+      "loss": 0.0247,
+      "step": 2444
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 2.9160583941605843e-06,
+      "loss": 0.0195,
+      "step": 2445
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 2.9124087591240877e-06,
+      "loss": 0.0273,
+      "step": 2446
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 2.908759124087591e-06,
+      "loss": 0.0198,
+      "step": 2447
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 2.905109489051095e-06,
+      "loss": 0.0192,
+      "step": 2448
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 2.901459854014599e-06,
+      "loss": 0.0111,
+      "step": 2449
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 2.897810218978103e-06,
+      "loss": 0.0313,
+      "step": 2450
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 2.894160583941606e-06,
+      "loss": 0.0362,
+      "step": 2451
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 2.8905109489051098e-06,
+      "loss": 0.0052,
+      "step": 2452
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 2.8868613138686137e-06,
+      "loss": 0.0143,
+      "step": 2453
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 2.8832116788321167e-06,
+      "loss": 0.0514,
+      "step": 2454
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 2.8795620437956206e-06,
+      "loss": 0.0317,
+      "step": 2455
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 2.8759124087591245e-06,
+      "loss": 0.0363,
+      "step": 2456
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 2.872262773722628e-06,
+      "loss": 0.066,
+      "step": 2457
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 2.8686131386861314e-06,
+      "loss": 0.0196,
+      "step": 2458
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 2.8649635036496353e-06,
+      "loss": 0.0012,
+      "step": 2459
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 2.8613138686131387e-06,
+      "loss": 0.0623,
+      "step": 2460
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 2.8576642335766426e-06,
+      "loss": 0.0226,
+      "step": 2461
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 2.8540145985401465e-06,
+      "loss": 0.0226,
+      "step": 2462
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 2.85036496350365e-06,
+      "loss": 0.0549,
+      "step": 2463
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 2.8467153284671534e-06,
+      "loss": 0.028,
+      "step": 2464
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 2.8430656934306573e-06,
+      "loss": 0.0329,
+      "step": 2465
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 2.839416058394161e-06,
+      "loss": 0.0456,
+      "step": 2466
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 2.835766423357664e-06,
+      "loss": 0.0384,
+      "step": 2467
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 2.832116788321168e-06,
+      "loss": 0.0253,
+      "step": 2468
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 2.828467153284672e-06,
+      "loss": 0.0252,
+      "step": 2469
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 2.824817518248175e-06,
+      "loss": 0.0366,
+      "step": 2470
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 2.821167883211679e-06,
+      "loss": 0.02,
+      "step": 2471
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 2.8175182481751828e-06,
+      "loss": 0.0303,
+      "step": 2472
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 2.8138686131386862e-06,
+      "loss": 0.0388,
+      "step": 2473
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 2.81021897810219e-06,
+      "loss": 0.0077,
+      "step": 2474
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 2.8065693430656936e-06,
+      "loss": 0.0594,
+      "step": 2475
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 2.8029197080291974e-06,
+      "loss": 0.0488,
+      "step": 2476
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 2.799270072992701e-06,
+      "loss": 0.0133,
+      "step": 2477
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 2.7956204379562048e-06,
+      "loss": 0.0172,
+      "step": 2478
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 2.7919708029197082e-06,
+      "loss": 0.0229,
+      "step": 2479
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 2.7883211678832117e-06,
+      "loss": 0.0217,
+      "step": 2480
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 2.7846715328467156e-06,
+      "loss": 0.0333,
+      "step": 2481
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 2.7810218978102195e-06,
+      "loss": 0.0336,
+      "step": 2482
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 2.7773722627737225e-06,
+      "loss": 0.0324,
+      "step": 2483
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 2.7737226277372264e-06,
+      "loss": 0.0411,
+      "step": 2484
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 2.7700729927007303e-06,
+      "loss": 0.0425,
+      "step": 2485
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 2.7664233576642337e-06,
+      "loss": 0.0196,
+      "step": 2486
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 2.762773722627737e-06,
+      "loss": 0.0491,
+      "step": 2487
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 2.759124087591241e-06,
+      "loss": 0.0266,
+      "step": 2488
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 2.755474452554745e-06,
+      "loss": 0.0198,
+      "step": 2489
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 2.7518248175182484e-06,
+      "loss": 0.0336,
+      "step": 2490
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 2.748175182481752e-06,
+      "loss": 0.0213,
+      "step": 2491
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 2.7445255474452558e-06,
+      "loss": 0.0083,
+      "step": 2492
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 2.740875912408759e-06,
+      "loss": 0.0334,
+      "step": 2493
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 2.737226277372263e-06,
+      "loss": 0.0236,
+      "step": 2494
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 2.733576642335767e-06,
+      "loss": 0.044,
+      "step": 2495
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 2.72992700729927e-06,
+      "loss": 0.0348,
+      "step": 2496
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 2.726277372262774e-06,
+      "loss": 0.0151,
+      "step": 2497
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 2.7226277372262778e-06,
+      "loss": 0.0397,
+      "step": 2498
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 2.718978102189781e-06,
+      "loss": 0.0356,
+      "step": 2499
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 2.7153284671532847e-06,
+      "loss": 0.0349,
+      "step": 2500
+    },
+    {
+      "epoch": 23.15,
+      "eval_loss": 1.2741411924362183,
+      "eval_runtime": 64.8459,
+      "eval_samples_per_second": 25.399,
+      "eval_steps_per_second": 1.588,
+      "eval_wer": 0.9587376161781228,
+      "step": 2500
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 2.7116788321167886e-06,
+      "loss": 0.014,
+      "step": 2501
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 2.7080291970802925e-06,
+      "loss": 0.036,
+      "step": 2502
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 2.7043795620437955e-06,
+      "loss": 0.0272,
+      "step": 2503
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 2.7007299270072994e-06,
+      "loss": 0.0229,
+      "step": 2504
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 2.6970802919708033e-06,
+      "loss": 0.0129,
+      "step": 2505
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 2.6934306569343067e-06,
+      "loss": 0.0456,
+      "step": 2506
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 2.6897810218978106e-06,
+      "loss": 0.035,
+      "step": 2507
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 2.686131386861314e-06,
+      "loss": 0.0246,
+      "step": 2508
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 2.6824817518248175e-06,
+      "loss": 0.0282,
+      "step": 2509
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 2.6788321167883214e-06,
+      "loss": 0.0321,
+      "step": 2510
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 2.6751824817518253e-06,
+      "loss": 0.0529,
+      "step": 2511
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 2.6715328467153283e-06,
+      "loss": 0.0277,
+      "step": 2512
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 2.667883211678832e-06,
+      "loss": 0.0324,
+      "step": 2513
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 2.664233576642336e-06,
+      "loss": 0.0189,
+      "step": 2514
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 2.66058394160584e-06,
+      "loss": 0.023,
+      "step": 2515
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 2.656934306569343e-06,
+      "loss": 0.0279,
+      "step": 2516
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 2.653284671532847e-06,
+      "loss": 0.039,
+      "step": 2517
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 2.6496350364963508e-06,
+      "loss": 0.0176,
+      "step": 2518
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 2.6459854014598542e-06,
+      "loss": 0.0503,
+      "step": 2519
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 2.6423357664233577e-06,
+      "loss": 0.0239,
+      "step": 2520
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 2.6386861313868616e-06,
+      "loss": 0.0051,
+      "step": 2521
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 2.635036496350365e-06,
+      "loss": 0.0319,
+      "step": 2522
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 2.631386861313869e-06,
+      "loss": 0.0247,
+      "step": 2523
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 2.627737226277373e-06,
+      "loss": 0.0415,
+      "step": 2524
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 2.624087591240876e-06,
+      "loss": 0.0444,
+      "step": 2525
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 2.6204379562043797e-06,
+      "loss": 0.0155,
+      "step": 2526
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 2.6167883211678836e-06,
+      "loss": 0.0276,
+      "step": 2527
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 2.6131386861313875e-06,
+      "loss": 0.0271,
+      "step": 2528
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 2.6094890510948905e-06,
+      "loss": 0.0236,
+      "step": 2529
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 2.6058394160583944e-06,
+      "loss": 0.0185,
+      "step": 2530
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 2.6021897810218983e-06,
+      "loss": 0.0472,
+      "step": 2531
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 2.5985401459854013e-06,
+      "loss": 0.0488,
+      "step": 2532
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 2.594890510948905e-06,
+      "loss": 0.0253,
+      "step": 2533
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 2.591240875912409e-06,
+      "loss": 0.0715,
+      "step": 2534
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 2.5875912408759125e-06,
+      "loss": 0.006,
+      "step": 2535
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 2.5839416058394164e-06,
+      "loss": 0.0146,
+      "step": 2536
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 2.58029197080292e-06,
+      "loss": 0.0248,
+      "step": 2537
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 2.5766423357664233e-06,
+      "loss": 0.0106,
+      "step": 2538
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 2.5729927007299272e-06,
+      "loss": 0.0345,
+      "step": 2539
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 2.569343065693431e-06,
+      "loss": 0.0286,
+      "step": 2540
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 2.565693430656935e-06,
+      "loss": 0.0165,
+      "step": 2541
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 2.562043795620438e-06,
+      "loss": 0.0257,
+      "step": 2542
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 2.558394160583942e-06,
+      "loss": 0.0143,
+      "step": 2543
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 2.5547445255474458e-06,
+      "loss": 0.0306,
+      "step": 2544
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 2.551094890510949e-06,
+      "loss": 0.0081,
+      "step": 2545
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 2.5474452554744527e-06,
+      "loss": 0.033,
+      "step": 2546
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 2.5437956204379566e-06,
+      "loss": 0.0098,
+      "step": 2547
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 2.54014598540146e-06,
+      "loss": 0.0169,
+      "step": 2548
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 2.5364963503649635e-06,
+      "loss": 0.0431,
+      "step": 2549
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 2.5328467153284674e-06,
+      "loss": 0.0552,
+      "step": 2550
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 2.529197080291971e-06,
+      "loss": 0.011,
+      "step": 2551
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 2.5255474452554747e-06,
+      "loss": 0.0449,
+      "step": 2552
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 2.5218978102189786e-06,
+      "loss": 0.0679,
+      "step": 2553
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 2.518248175182482e-06,
+      "loss": 0.0444,
+      "step": 2554
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 2.5145985401459855e-06,
+      "loss": 0.0477,
+      "step": 2555
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 2.5109489051094894e-06,
+      "loss": 0.027,
+      "step": 2556
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 2.5072992700729933e-06,
+      "loss": 0.0062,
+      "step": 2557
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 2.5036496350364963e-06,
+      "loss": 0.0288,
+      "step": 2558
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0368,
+      "step": 2559
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 2.4963503649635037e-06,
+      "loss": 0.0163,
+      "step": 2560
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 2.4927007299270075e-06,
+      "loss": 0.0311,
+      "step": 2561
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 2.489051094890511e-06,
+      "loss": 0.0369,
+      "step": 2562
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 2.485401459854015e-06,
+      "loss": 0.0217,
+      "step": 2563
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 2.4817518248175183e-06,
+      "loss": 0.0247,
+      "step": 2564
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 2.478102189781022e-06,
+      "loss": 0.0422,
+      "step": 2565
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 2.4744525547445257e-06,
+      "loss": 0.0284,
+      "step": 2566
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 2.4708029197080296e-06,
+      "loss": 0.004,
+      "step": 2567
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 2.467153284671533e-06,
+      "loss": 0.0414,
+      "step": 2568
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 2.463503649635037e-06,
+      "loss": 0.0327,
+      "step": 2569
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 2.4598540145985404e-06,
+      "loss": 0.014,
+      "step": 2570
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 2.4562043795620443e-06,
+      "loss": 0.0389,
+      "step": 2571
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 2.4525547445255477e-06,
+      "loss": 0.0267,
+      "step": 2572
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 2.448905109489051e-06,
+      "loss": 0.0174,
+      "step": 2573
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 2.445255474452555e-06,
+      "loss": 0.0332,
+      "step": 2574
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 2.4416058394160585e-06,
+      "loss": 0.0168,
+      "step": 2575
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 2.4379562043795624e-06,
+      "loss": 0.0023,
+      "step": 2576
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 2.434306569343066e-06,
+      "loss": 0.0406,
+      "step": 2577
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 2.4306569343065693e-06,
+      "loss": 0.0387,
+      "step": 2578
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 2.427007299270073e-06,
+      "loss": 0.0073,
+      "step": 2579
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 2.4233576642335767e-06,
+      "loss": 0.0122,
+      "step": 2580
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 2.4197080291970805e-06,
+      "loss": 0.0286,
+      "step": 2581
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 2.416058394160584e-06,
+      "loss": 0.0222,
+      "step": 2582
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 2.412408759124088e-06,
+      "loss": 0.0344,
+      "step": 2583
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 2.4087591240875918e-06,
+      "loss": 0.033,
+      "step": 2584
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 2.4051094890510952e-06,
+      "loss": 0.0347,
+      "step": 2585
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 2.4014598540145987e-06,
+      "loss": 0.05,
+      "step": 2586
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 2.3978102189781026e-06,
+      "loss": 0.0122,
+      "step": 2587
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 2.394160583941606e-06,
+      "loss": 0.0459,
+      "step": 2588
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 2.39051094890511e-06,
+      "loss": 0.026,
+      "step": 2589
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 2.3868613138686134e-06,
+      "loss": 0.0426,
+      "step": 2590
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 2.383211678832117e-06,
+      "loss": 0.0307,
+      "step": 2591
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 2.3795620437956207e-06,
+      "loss": 0.02,
+      "step": 2592
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 2.375912408759124e-06,
+      "loss": 0.0453,
+      "step": 2593
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 2.3722627737226276e-06,
+      "loss": 0.0096,
+      "step": 2594
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 2.3686131386861315e-06,
+      "loss": 0.0096,
+      "step": 2595
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 2.3649635036496354e-06,
+      "loss": 0.0563,
+      "step": 2596
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 2.361313868613139e-06,
+      "loss": 0.0261,
+      "step": 2597
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 2.3576642335766427e-06,
+      "loss": 0.031,
+      "step": 2598
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 2.354014598540146e-06,
+      "loss": 0.0121,
+      "step": 2599
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 2.35036496350365e-06,
+      "loss": 0.0121,
+      "step": 2600
+    },
+    {
+      "epoch": 24.07,
+      "eval_loss": 1.267063021659851,
+      "eval_runtime": 65.0146,
+      "eval_samples_per_second": 25.333,
+      "eval_steps_per_second": 1.584,
+      "eval_wer": 0.9586354815647022,
+      "step": 2600
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 2.3467153284671535e-06,
+      "loss": 0.045,
+      "step": 2601
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 2.3430656934306574e-06,
+      "loss": 0.0156,
+      "step": 2602
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 2.339416058394161e-06,
+      "loss": 0.009,
+      "step": 2603
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 2.3357664233576643e-06,
+      "loss": 0.0089,
+      "step": 2604
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 2.332116788321168e-06,
+      "loss": 0.0211,
+      "step": 2605
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 2.3284671532846717e-06,
+      "loss": 0.0403,
+      "step": 2606
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 2.324817518248175e-06,
+      "loss": 0.0295,
+      "step": 2607
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 2.321167883211679e-06,
+      "loss": 0.0653,
+      "step": 2608
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 2.3175182481751825e-06,
+      "loss": 0.0143,
+      "step": 2609
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 2.3138686131386863e-06,
+      "loss": 0.0321,
+      "step": 2610
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 2.31021897810219e-06,
+      "loss": 0.0416,
+      "step": 2611
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 2.3065693430656937e-06,
+      "loss": 0.0264,
+      "step": 2612
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 2.302919708029197e-06,
+      "loss": 0.0103,
+      "step": 2613
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 2.299270072992701e-06,
+      "loss": 0.0682,
+      "step": 2614
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 2.295620437956205e-06,
+      "loss": 0.0198,
+      "step": 2615
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 2.2919708029197084e-06,
+      "loss": 0.0575,
+      "step": 2616
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 2.288321167883212e-06,
+      "loss": 0.0647,
+      "step": 2617
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 2.2846715328467157e-06,
+      "loss": 0.0073,
+      "step": 2618
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 2.281021897810219e-06,
+      "loss": 0.0255,
+      "step": 2619
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 2.2773722627737226e-06,
+      "loss": 0.0264,
+      "step": 2620
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 2.2737226277372265e-06,
+      "loss": 0.0314,
+      "step": 2621
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 2.27007299270073e-06,
+      "loss": 0.0173,
+      "step": 2622
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 2.266423357664234e-06,
+      "loss": 0.0407,
+      "step": 2623
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 2.2627737226277373e-06,
+      "loss": 0.0184,
+      "step": 2624
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 2.2591240875912408e-06,
+      "loss": 0.0112,
+      "step": 2625
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 2.2554744525547447e-06,
+      "loss": 0.0288,
+      "step": 2626
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 2.2518248175182485e-06,
+      "loss": 0.0122,
+      "step": 2627
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 2.248175182481752e-06,
+      "loss": 0.0172,
+      "step": 2628
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 2.244525547445256e-06,
+      "loss": 0.0125,
+      "step": 2629
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 2.2408759124087593e-06,
+      "loss": 0.0079,
+      "step": 2630
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 2.2372262773722632e-06,
+      "loss": 0.0187,
+      "step": 2631
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 2.2335766423357667e-06,
+      "loss": 0.0228,
+      "step": 2632
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 2.22992700729927e-06,
+      "loss": 0.0173,
+      "step": 2633
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 2.226277372262774e-06,
+      "loss": 0.0275,
+      "step": 2634
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 2.2226277372262775e-06,
+      "loss": 0.0163,
+      "step": 2635
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 2.2189781021897814e-06,
+      "loss": 0.035,
+      "step": 2636
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 2.215328467153285e-06,
+      "loss": 0.019,
+      "step": 2637
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 2.2116788321167883e-06,
+      "loss": 0.0552,
+      "step": 2638
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 2.208029197080292e-06,
+      "loss": 0.026,
+      "step": 2639
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 2.2043795620437956e-06,
+      "loss": 0.0451,
+      "step": 2640
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 2.2007299270072995e-06,
+      "loss": 0.0071,
+      "step": 2641
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 2.197080291970803e-06,
+      "loss": 0.0471,
+      "step": 2642
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 2.193430656934307e-06,
+      "loss": 0.0119,
+      "step": 2643
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 2.1897810218978103e-06,
+      "loss": 0.014,
+      "step": 2644
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 2.186131386861314e-06,
+      "loss": 0.0207,
+      "step": 2645
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 2.1824817518248176e-06,
+      "loss": 0.0137,
+      "step": 2646
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 2.1788321167883215e-06,
+      "loss": 0.0326,
+      "step": 2647
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 2.175182481751825e-06,
+      "loss": 0.0274,
+      "step": 2648
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 2.171532846715329e-06,
+      "loss": 0.0065,
+      "step": 2649
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 2.1678832116788323e-06,
+      "loss": 0.0244,
+      "step": 2650
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 2.164233576642336e-06,
+      "loss": 0.0167,
+      "step": 2651
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 2.1605839416058397e-06,
+      "loss": 0.0262,
+      "step": 2652
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 2.156934306569343e-06,
+      "loss": 0.0194,
+      "step": 2653
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 2.1532846715328466e-06,
+      "loss": 0.041,
+      "step": 2654
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 2.1496350364963505e-06,
+      "loss": 0.0691,
+      "step": 2655
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 2.145985401459854e-06,
+      "loss": 0.0096,
+      "step": 2656
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 2.142335766423358e-06,
+      "loss": 0.0193,
+      "step": 2657
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 2.1386861313868617e-06,
+      "loss": 0.0095,
+      "step": 2658
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 2.135036496350365e-06,
+      "loss": 0.0108,
+      "step": 2659
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 2.131386861313869e-06,
+      "loss": 0.0178,
+      "step": 2660
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 2.1277372262773725e-06,
+      "loss": 0.0378,
+      "step": 2661
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 2.1240875912408764e-06,
+      "loss": 0.027,
+      "step": 2662
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 2.12043795620438e-06,
+      "loss": 0.0331,
+      "step": 2663
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 2.1167883211678833e-06,
+      "loss": 0.0403,
+      "step": 2664
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 2.113138686131387e-06,
+      "loss": 0.0205,
+      "step": 2665
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 2.1094890510948906e-06,
+      "loss": 0.0204,
+      "step": 2666
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 2.105839416058394e-06,
+      "loss": 0.0472,
+      "step": 2667
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 2.102189781021898e-06,
+      "loss": 0.0281,
+      "step": 2668
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 2.0985401459854014e-06,
+      "loss": 0.0131,
+      "step": 2669
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 2.0948905109489053e-06,
+      "loss": 0.0211,
+      "step": 2670
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 2.0912408759124088e-06,
+      "loss": 0.0253,
+      "step": 2671
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 2.0875912408759127e-06,
+      "loss": 0.0212,
+      "step": 2672
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 2.083941605839416e-06,
+      "loss": 0.0356,
+      "step": 2673
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 2.08029197080292e-06,
+      "loss": 0.0414,
+      "step": 2674
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 2.0766423357664235e-06,
+      "loss": 0.025,
+      "step": 2675
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 2.0729927007299273e-06,
+      "loss": 0.028,
+      "step": 2676
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 2.069343065693431e-06,
+      "loss": 0.0274,
+      "step": 2677
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 2.0656934306569347e-06,
+      "loss": 0.0073,
+      "step": 2678
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 2.062043795620438e-06,
+      "loss": 0.0935,
+      "step": 2679
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 2.0583941605839416e-06,
+      "loss": 0.0627,
+      "step": 2680
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 2.0547445255474455e-06,
+      "loss": 0.0099,
+      "step": 2681
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 2.051094890510949e-06,
+      "loss": 0.0344,
+      "step": 2682
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 2.047445255474453e-06,
+      "loss": 0.0169,
+      "step": 2683
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 2.0437956204379563e-06,
+      "loss": 0.0434,
+      "step": 2684
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 2.0401459854014597e-06,
+      "loss": 0.0214,
+      "step": 2685
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 2.0364963503649636e-06,
+      "loss": 0.0508,
+      "step": 2686
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 2.032846715328467e-06,
+      "loss": 0.0184,
+      "step": 2687
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 2.029197080291971e-06,
+      "loss": 0.0413,
+      "step": 2688
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 2.025547445255475e-06,
+      "loss": 0.0163,
+      "step": 2689
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 2.0218978102189783e-06,
+      "loss": 0.0349,
+      "step": 2690
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 2.018248175182482e-06,
+      "loss": 0.0177,
+      "step": 2691
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 2.0145985401459857e-06,
+      "loss": 0.0385,
+      "step": 2692
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 2.010948905109489e-06,
+      "loss": 0.0085,
+      "step": 2693
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 2.007299270072993e-06,
+      "loss": 0.024,
+      "step": 2694
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 2.0036496350364965e-06,
+      "loss": 0.0569,
+      "step": 2695
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.0331,
+      "step": 2696
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.996350364963504e-06,
+      "loss": 0.0186,
+      "step": 2697
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.9927007299270073e-06,
+      "loss": 0.0181,
+      "step": 2698
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 1.989051094890511e-06,
+      "loss": 0.0259,
+      "step": 2699
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 1.9854014598540146e-06,
+      "loss": 0.0224,
+      "step": 2700
+    },
+    {
+      "epoch": 24.99,
+      "eval_loss": 1.3000985383987427,
+      "eval_runtime": 64.6905,
+      "eval_samples_per_second": 25.46,
+      "eval_steps_per_second": 1.592,
+      "eval_wer": 0.961086712286794,
+      "step": 2700
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 1.9817518248175185e-06,
+      "loss": 0.1215,
+      "step": 2701
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 1.978102189781022e-06,
+      "loss": 0.0587,
+      "step": 2702
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.974452554744526e-06,
+      "loss": 0.0166,
+      "step": 2703
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 1.9708029197080293e-06,
+      "loss": 0.0167,
+      "step": 2704
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 1.967153284671533e-06,
+      "loss": 0.0291,
+      "step": 2705
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.9635036496350366e-06,
+      "loss": 0.0143,
+      "step": 2706
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.9598540145985405e-06,
+      "loss": 0.0183,
+      "step": 2707
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.956204379562044e-06,
+      "loss": 0.0258,
+      "step": 2708
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 1.952554744525548e-06,
+      "loss": 0.0154,
+      "step": 2709
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.9489051094890513e-06,
+      "loss": 0.0335,
+      "step": 2710
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.9452554744525548e-06,
+      "loss": 0.0096,
+      "step": 2711
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 1.9416058394160586e-06,
+      "loss": 0.0342,
+      "step": 2712
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 1.937956204379562e-06,
+      "loss": 0.0064,
+      "step": 2713
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 1.934306569343066e-06,
+      "loss": 0.0258,
+      "step": 2714
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 1.9306569343065694e-06,
+      "loss": 0.056,
+      "step": 2715
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.927007299270073e-06,
+      "loss": 0.0092,
+      "step": 2716
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.9233576642335768e-06,
+      "loss": 0.0243,
+      "step": 2717
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.9197080291970802e-06,
+      "loss": 0.0258,
+      "step": 2718
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.916058394160584e-06,
+      "loss": 0.0321,
+      "step": 2719
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 1.912408759124088e-06,
+      "loss": 0.0153,
+      "step": 2720
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 1.9087591240875915e-06,
+      "loss": 0.0316,
+      "step": 2721
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.9051094890510951e-06,
+      "loss": 0.0152,
+      "step": 2722
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.9014598540145988e-06,
+      "loss": 0.0574,
+      "step": 2723
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 1.8978102189781023e-06,
+      "loss": 0.0542,
+      "step": 2724
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.8941605839416061e-06,
+      "loss": 0.0366,
+      "step": 2725
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 1.8905109489051096e-06,
+      "loss": 0.0195,
+      "step": 2726
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 1.8868613138686135e-06,
+      "loss": 0.0311,
+      "step": 2727
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.883211678832117e-06,
+      "loss": 0.0329,
+      "step": 2728
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 1.8795620437956206e-06,
+      "loss": 0.025,
+      "step": 2729
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 1.8759124087591243e-06,
+      "loss": 0.0207,
+      "step": 2730
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.872262773722628e-06,
+      "loss": 0.0471,
+      "step": 2731
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.8686131386861314e-06,
+      "loss": 0.02,
+      "step": 2732
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 1.8649635036496353e-06,
+      "loss": 0.0441,
+      "step": 2733
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 1.8613138686131388e-06,
+      "loss": 0.0141,
+      "step": 2734
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.8576642335766426e-06,
+      "loss": 0.0218,
+      "step": 2735
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 1.854014598540146e-06,
+      "loss": 0.009,
+      "step": 2736
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 1.8503649635036498e-06,
+      "loss": 0.0428,
+      "step": 2737
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.8467153284671534e-06,
+      "loss": 0.0137,
+      "step": 2738
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 1.8430656934306571e-06,
+      "loss": 0.0199,
+      "step": 2739
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 1.8394160583941606e-06,
+      "loss": 0.0118,
+      "step": 2740
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.8357664233576645e-06,
+      "loss": 0.0235,
+      "step": 2741
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 1.832116788321168e-06,
+      "loss": 0.0174,
+      "step": 2742
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.8284671532846718e-06,
+      "loss": 0.0273,
+      "step": 2743
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.8248175182481753e-06,
+      "loss": 0.0181,
+      "step": 2744
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 1.821167883211679e-06,
+      "loss": 0.0148,
+      "step": 2745
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.8175182481751826e-06,
+      "loss": 0.0323,
+      "step": 2746
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 1.8138686131386863e-06,
+      "loss": 0.0311,
+      "step": 2747
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.8102189781021901e-06,
+      "loss": 0.0052,
+      "step": 2748
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 1.8065693430656936e-06,
+      "loss": 0.0483,
+      "step": 2749
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 1.802919708029197e-06,
+      "loss": 0.0538,
+      "step": 2750
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 1.799270072992701e-06,
+      "loss": 0.0064,
+      "step": 2751
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.7956204379562044e-06,
+      "loss": 0.031,
+      "step": 2752
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 1.791970802919708e-06,
+      "loss": 0.035,
+      "step": 2753
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.788321167883212e-06,
+      "loss": 0.0138,
+      "step": 2754
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.7846715328467154e-06,
+      "loss": 0.0348,
+      "step": 2755
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.7810218978102193e-06,
+      "loss": 0.0327,
+      "step": 2756
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 1.7773722627737228e-06,
+      "loss": 0.0067,
+      "step": 2757
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 1.7737226277372262e-06,
+      "loss": 0.0173,
+      "step": 2758
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 1.77007299270073e-06,
+      "loss": 0.0534,
+      "step": 2759
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.7664233576642338e-06,
+      "loss": 0.0364,
+      "step": 2760
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.7627737226277374e-06,
+      "loss": 0.0081,
+      "step": 2761
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 1.7591240875912411e-06,
+      "loss": 0.0321,
+      "step": 2762
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 1.7554744525547446e-06,
+      "loss": 0.0198,
+      "step": 2763
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 1.7518248175182485e-06,
+      "loss": 0.0366,
+      "step": 2764
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 1.748175182481752e-06,
+      "loss": 0.0269,
+      "step": 2765
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.7445255474452556e-06,
+      "loss": 0.0226,
+      "step": 2766
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.7408759124087593e-06,
+      "loss": 0.0216,
+      "step": 2767
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.737226277372263e-06,
+      "loss": 0.0316,
+      "step": 2768
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.7335766423357666e-06,
+      "loss": 0.0482,
+      "step": 2769
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 1.7299270072992703e-06,
+      "loss": 0.0208,
+      "step": 2770
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.7262773722627737e-06,
+      "loss": 0.0287,
+      "step": 2771
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 1.7226277372262776e-06,
+      "loss": 0.0552,
+      "step": 2772
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 1.718978102189781e-06,
+      "loss": 0.013,
+      "step": 2773
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.715328467153285e-06,
+      "loss": 0.0342,
+      "step": 2774
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 1.7116788321167884e-06,
+      "loss": 0.0191,
+      "step": 2775
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 1.708029197080292e-06,
+      "loss": 0.0285,
+      "step": 2776
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 1.7043795620437958e-06,
+      "loss": 0.0344,
+      "step": 2777
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 1.7007299270072994e-06,
+      "loss": 0.0263,
+      "step": 2778
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 1.6970802919708029e-06,
+      "loss": 0.0248,
+      "step": 2779
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.6934306569343068e-06,
+      "loss": 0.0071,
+      "step": 2780
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.6897810218978102e-06,
+      "loss": 0.0309,
+      "step": 2781
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 1.686131386861314e-06,
+      "loss": 0.0085,
+      "step": 2782
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 1.6824817518248176e-06,
+      "loss": 0.0047,
+      "step": 2783
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 1.6788321167883212e-06,
+      "loss": 0.0466,
+      "step": 2784
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 1.6751824817518251e-06,
+      "loss": 0.016,
+      "step": 2785
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.6715328467153286e-06,
+      "loss": 0.017,
+      "step": 2786
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.6678832116788325e-06,
+      "loss": 0.016,
+      "step": 2787
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.664233576642336e-06,
+      "loss": 0.0208,
+      "step": 2788
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 1.6605839416058394e-06,
+      "loss": 0.0231,
+      "step": 2789
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 1.6569343065693433e-06,
+      "loss": 0.0417,
+      "step": 2790
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 1.653284671532847e-06,
+      "loss": 0.0364,
+      "step": 2791
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.6496350364963504e-06,
+      "loss": 0.0159,
+      "step": 2792
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.6459854014598543e-06,
+      "loss": 0.0359,
+      "step": 2793
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.6423357664233577e-06,
+      "loss": 0.0648,
+      "step": 2794
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.6386861313868616e-06,
+      "loss": 0.0193,
+      "step": 2795
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 1.635036496350365e-06,
+      "loss": 0.0366,
+      "step": 2796
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 1.6313868613138687e-06,
+      "loss": 0.0344,
+      "step": 2797
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 1.6277372262773724e-06,
+      "loss": 0.0144,
+      "step": 2798
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.624087591240876e-06,
+      "loss": 0.0235,
+      "step": 2799
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 1.6204379562043798e-06,
+      "loss": 0.0449,
+      "step": 2800
+    },
+    {
+      "epoch": 25.92,
+      "eval_loss": 1.277716875076294,
+      "eval_runtime": 64.7175,
+      "eval_samples_per_second": 25.449,
+      "eval_steps_per_second": 1.592,
+      "eval_wer": 0.9572055969768154,
+      "step": 2800
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.6167883211678834e-06,
+      "loss": 0.0229,
+      "step": 2801
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 1.6131386861313869e-06,
+      "loss": 0.02,
+      "step": 2802
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 1.6094890510948908e-06,
+      "loss": 0.0662,
+      "step": 2803
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 1.6058394160583942e-06,
+      "loss": 0.0241,
+      "step": 2804
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.602189781021898e-06,
+      "loss": 0.0106,
+      "step": 2805
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.5985401459854016e-06,
+      "loss": 0.0736,
+      "step": 2806
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 1.5948905109489052e-06,
+      "loss": 0.0055,
+      "step": 2807
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 1.591240875912409e-06,
+      "loss": 0.0194,
+      "step": 2808
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 1.5875912408759126e-06,
+      "loss": 0.0316,
+      "step": 2809
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 1.583941605839416e-06,
+      "loss": 0.029,
+      "step": 2810
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.58029197080292e-06,
+      "loss": 0.0239,
+      "step": 2811
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 1.5766423357664234e-06,
+      "loss": 0.0078,
+      "step": 2812
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 1.5729927007299273e-06,
+      "loss": 0.0157,
+      "step": 2813
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.5693430656934307e-06,
+      "loss": 0.0265,
+      "step": 2814
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.5656934306569344e-06,
+      "loss": 0.0186,
+      "step": 2815
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 1.5620437956204383e-06,
+      "loss": 0.031,
+      "step": 2816
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 1.5583941605839417e-06,
+      "loss": 0.0547,
+      "step": 2817
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.5547445255474452e-06,
+      "loss": 0.0159,
+      "step": 2818
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 1.551094890510949e-06,
+      "loss": 0.0261,
+      "step": 2819
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 1.5474452554744525e-06,
+      "loss": 0.015,
+      "step": 2820
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 1.5437956204379564e-06,
+      "loss": 0.0111,
+      "step": 2821
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.54014598540146e-06,
+      "loss": 0.0192,
+      "step": 2822
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 1.5364963503649635e-06,
+      "loss": 0.043,
+      "step": 2823
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.5328467153284674e-06,
+      "loss": 0.0475,
+      "step": 2824
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 1.5291970802919709e-06,
+      "loss": 0.0217,
+      "step": 2825
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.5255474452554748e-06,
+      "loss": 0.0365,
+      "step": 2826
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.5218978102189782e-06,
+      "loss": 0.0367,
+      "step": 2827
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 1.518248175182482e-06,
+      "loss": 0.0149,
+      "step": 2828
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.5145985401459856e-06,
+      "loss": 0.0218,
+      "step": 2829
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 1.5109489051094892e-06,
+      "loss": 0.0263,
+      "step": 2830
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.5072992700729927e-06,
+      "loss": 0.0221,
+      "step": 2831
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 1.5036496350364966e-06,
+      "loss": 0.032,
+      "step": 2832
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 1.5e-06,
+      "loss": 0.0237,
+      "step": 2833
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 1.496350364963504e-06,
+      "loss": 0.0189,
+      "step": 2834
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 1.4927007299270074e-06,
+      "loss": 0.0077,
+      "step": 2835
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.489051094890511e-06,
+      "loss": 0.087,
+      "step": 2836
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 1.4854014598540147e-06,
+      "loss": 0.0178,
+      "step": 2837
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 1.4817518248175184e-06,
+      "loss": 0.021,
+      "step": 2838
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.478102189781022e-06,
+      "loss": 0.0135,
+      "step": 2839
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.4744525547445257e-06,
+      "loss": 0.0322,
+      "step": 2840
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 1.4708029197080292e-06,
+      "loss": 0.0111,
+      "step": 2841
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 1.467153284671533e-06,
+      "loss": 0.0414,
+      "step": 2842
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.4635036496350365e-06,
+      "loss": 0.0097,
+      "step": 2843
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 1.4598540145985402e-06,
+      "loss": 0.0251,
+      "step": 2844
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 1.4562043795620439e-06,
+      "loss": 0.0296,
+      "step": 2845
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 1.4525547445255475e-06,
+      "loss": 0.0415,
+      "step": 2846
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 1.4489051094890514e-06,
+      "loss": 0.0215,
+      "step": 2847
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 1.4452554744525549e-06,
+      "loss": 0.0268,
+      "step": 2848
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.4416058394160583e-06,
+      "loss": 0.0189,
+      "step": 2849
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.4379562043795622e-06,
+      "loss": 0.0304,
+      "step": 2850
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.4343065693430657e-06,
+      "loss": 0.0131,
+      "step": 2851
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.4306569343065694e-06,
+      "loss": 0.0182,
+      "step": 2852
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 1.4270072992700732e-06,
+      "loss": 0.0107,
+      "step": 2853
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 1.4233576642335767e-06,
+      "loss": 0.0094,
+      "step": 2854
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 1.4197080291970806e-06,
+      "loss": 0.0102,
+      "step": 2855
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.416058394160584e-06,
+      "loss": 0.0617,
+      "step": 2856
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.4124087591240875e-06,
+      "loss": 0.0248,
+      "step": 2857
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 1.4087591240875914e-06,
+      "loss": 0.0309,
+      "step": 2858
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 1.405109489051095e-06,
+      "loss": 0.0166,
+      "step": 2859
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 1.4014598540145987e-06,
+      "loss": 0.0441,
+      "step": 2860
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 1.3978102189781024e-06,
+      "loss": 0.0411,
+      "step": 2861
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.3941605839416059e-06,
+      "loss": 0.0133,
+      "step": 2862
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.3905109489051097e-06,
+      "loss": 0.0245,
+      "step": 2863
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.3868613138686132e-06,
+      "loss": 0.0635,
+      "step": 2864
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 1.3832116788321169e-06,
+      "loss": 0.0524,
+      "step": 2865
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 1.3795620437956205e-06,
+      "loss": 0.0298,
+      "step": 2866
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 1.3759124087591242e-06,
+      "loss": 0.0298,
+      "step": 2867
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 1.3722627737226279e-06,
+      "loss": 0.0043,
+      "step": 2868
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.3686131386861315e-06,
+      "loss": 0.0374,
+      "step": 2869
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 1.364963503649635e-06,
+      "loss": 0.0268,
+      "step": 2870
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.3613138686131389e-06,
+      "loss": 0.0328,
+      "step": 2871
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 1.3576642335766423e-06,
+      "loss": 0.0039,
+      "step": 2872
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 1.3540145985401462e-06,
+      "loss": 0.0049,
+      "step": 2873
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.3503649635036497e-06,
+      "loss": 0.0385,
+      "step": 2874
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.3467153284671534e-06,
+      "loss": 0.0129,
+      "step": 2875
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.343065693430657e-06,
+      "loss": 0.0188,
+      "step": 2876
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.3394160583941607e-06,
+      "loss": 0.033,
+      "step": 2877
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.3357664233576642e-06,
+      "loss": 0.0125,
+      "step": 2878
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 1.332116788321168e-06,
+      "loss": 0.023,
+      "step": 2879
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 1.3284671532846715e-06,
+      "loss": 0.0341,
+      "step": 2880
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 1.3248175182481754e-06,
+      "loss": 0.0265,
+      "step": 2881
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.3211678832116788e-06,
+      "loss": 0.0226,
+      "step": 2882
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 1.3175182481751825e-06,
+      "loss": 0.0434,
+      "step": 2883
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 1.3138686131386864e-06,
+      "loss": 0.0142,
+      "step": 2884
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.3102189781021899e-06,
+      "loss": 0.0191,
+      "step": 2885
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 1.3065693430656937e-06,
+      "loss": 0.0166,
+      "step": 2886
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 1.3029197080291972e-06,
+      "loss": 0.0233,
+      "step": 2887
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.2992700729927007e-06,
+      "loss": 0.037,
+      "step": 2888
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.2956204379562045e-06,
+      "loss": 0.0252,
+      "step": 2889
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 1.2919708029197082e-06,
+      "loss": 0.0142,
+      "step": 2890
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 1.2883211678832117e-06,
+      "loss": 0.0264,
+      "step": 2891
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.2846715328467155e-06,
+      "loss": 0.0261,
+      "step": 2892
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.281021897810219e-06,
+      "loss": 0.021,
+      "step": 2893
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.2773722627737229e-06,
+      "loss": 0.0125,
+      "step": 2894
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 1.2737226277372263e-06,
+      "loss": 0.0159,
+      "step": 2895
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 1.27007299270073e-06,
+      "loss": 0.0168,
+      "step": 2896
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 1.2664233576642337e-06,
+      "loss": 0.0115,
+      "step": 2897
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 1.2627737226277374e-06,
+      "loss": 0.0182,
+      "step": 2898
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.259124087591241e-06,
+      "loss": 0.01,
+      "step": 2899
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.2554744525547447e-06,
+      "loss": 0.0186,
+      "step": 2900
+    },
+    {
+      "epoch": 26.85,
+      "eval_loss": 1.2766320705413818,
+      "eval_runtime": 64.656,
+      "eval_samples_per_second": 25.473,
+      "eval_steps_per_second": 1.593,
+      "eval_wer": 0.960678173833112,
+      "step": 2900
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.2518248175182482e-06,
+      "loss": 0.0148,
+      "step": 2901
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.2481751824817518e-06,
+      "loss": 0.0232,
+      "step": 2902
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 1.2445255474452555e-06,
+      "loss": 0.0195,
+      "step": 2903
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 1.2408759124087592e-06,
+      "loss": 0.0304,
+      "step": 2904
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 1.2372262773722628e-06,
+      "loss": 0.0502,
+      "step": 2905
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.2335766423357665e-06,
+      "loss": 0.0224,
+      "step": 2906
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.2299270072992702e-06,
+      "loss": 0.0688,
+      "step": 2907
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 1.2262773722627739e-06,
+      "loss": 0.0358,
+      "step": 2908
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 1.2226277372262775e-06,
+      "loss": 0.0273,
+      "step": 2909
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.2189781021897812e-06,
+      "loss": 0.0426,
+      "step": 2910
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 1.2153284671532847e-06,
+      "loss": 0.0697,
+      "step": 2911
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.2116788321167883e-06,
+      "loss": 0.0126,
+      "step": 2912
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.208029197080292e-06,
+      "loss": 0.0161,
+      "step": 2913
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.2043795620437959e-06,
+      "loss": 0.0185,
+      "step": 2914
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 1.2007299270072993e-06,
+      "loss": 0.0149,
+      "step": 2915
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 1.197080291970803e-06,
+      "loss": 0.0298,
+      "step": 2916
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 1.1934306569343067e-06,
+      "loss": 0.0386,
+      "step": 2917
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 1.1897810218978104e-06,
+      "loss": 0.0399,
+      "step": 2918
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.1861313868613138e-06,
+      "loss": 0.0043,
+      "step": 2919
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 1.1824817518248177e-06,
+      "loss": 0.012,
+      "step": 2920
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 1.1788321167883214e-06,
+      "loss": 0.0382,
+      "step": 2921
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.175182481751825e-06,
+      "loss": 0.0303,
+      "step": 2922
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.1715328467153287e-06,
+      "loss": 0.0129,
+      "step": 2923
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 1.1678832116788322e-06,
+      "loss": 0.0297,
+      "step": 2924
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 1.1642335766423358e-06,
+      "loss": 0.0168,
+      "step": 2925
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.1605839416058395e-06,
+      "loss": 0.0182,
+      "step": 2926
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.1569343065693432e-06,
+      "loss": 0.0504,
+      "step": 2927
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 1.1532846715328468e-06,
+      "loss": 0.0487,
+      "step": 2928
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 1.1496350364963505e-06,
+      "loss": 0.0237,
+      "step": 2929
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 1.1459854014598542e-06,
+      "loss": 0.0256,
+      "step": 2930
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 1.1423357664233579e-06,
+      "loss": 0.0474,
+      "step": 2931
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.1386861313868613e-06,
+      "loss": 0.0384,
+      "step": 2932
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.135036496350365e-06,
+      "loss": 0.0124,
+      "step": 2933
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.1313868613138687e-06,
+      "loss": 0.0193,
+      "step": 2934
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.1277372262773723e-06,
+      "loss": 0.0417,
+      "step": 2935
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 1.124087591240876e-06,
+      "loss": 0.0173,
+      "step": 2936
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 1.1204379562043797e-06,
+      "loss": 0.0435,
+      "step": 2937
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 1.1167883211678833e-06,
+      "loss": 0.0403,
+      "step": 2938
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.113138686131387e-06,
+      "loss": 0.0178,
+      "step": 2939
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 1.1094890510948907e-06,
+      "loss": 0.0257,
+      "step": 2940
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.1058394160583941e-06,
+      "loss": 0.0199,
+      "step": 2941
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.1021897810218978e-06,
+      "loss": 0.0223,
+      "step": 2942
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 1.0985401459854015e-06,
+      "loss": 0.0209,
+      "step": 2943
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.0948905109489052e-06,
+      "loss": 0.0492,
+      "step": 2944
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 1.0912408759124088e-06,
+      "loss": 0.0275,
+      "step": 2945
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 1.0875912408759125e-06,
+      "loss": 0.015,
+      "step": 2946
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.0839416058394162e-06,
+      "loss": 0.0338,
+      "step": 2947
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.0802919708029198e-06,
+      "loss": 0.0265,
+      "step": 2948
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 1.0766423357664233e-06,
+      "loss": 0.0095,
+      "step": 2949
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 1.072992700729927e-06,
+      "loss": 0.0586,
+      "step": 2950
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.0693430656934308e-06,
+      "loss": 0.008,
+      "step": 2951
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 1.0656934306569345e-06,
+      "loss": 0.0363,
+      "step": 2952
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 1.0620437956204382e-06,
+      "loss": 0.0232,
+      "step": 2953
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.0583941605839416e-06,
+      "loss": 0.043,
+      "step": 2954
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 1.0547445255474453e-06,
+      "loss": 0.0241,
+      "step": 2955
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 1.051094890510949e-06,
+      "loss": 0.0319,
+      "step": 2956
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.0474452554744527e-06,
+      "loss": 0.0224,
+      "step": 2957
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 1.0437956204379563e-06,
+      "loss": 0.0147,
+      "step": 2958
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.04014598540146e-06,
+      "loss": 0.0456,
+      "step": 2959
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.0364963503649637e-06,
+      "loss": 0.0196,
+      "step": 2960
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 1.0328467153284673e-06,
+      "loss": 0.0044,
+      "step": 2961
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.0291970802919708e-06,
+      "loss": 0.0429,
+      "step": 2962
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 1.0255474452554745e-06,
+      "loss": 0.0045,
+      "step": 2963
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.0218978102189781e-06,
+      "loss": 0.0077,
+      "step": 2964
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 1.0182481751824818e-06,
+      "loss": 0.0472,
+      "step": 2965
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 1.0145985401459855e-06,
+      "loss": 0.0239,
+      "step": 2966
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 1.0109489051094892e-06,
+      "loss": 0.0073,
+      "step": 2967
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.0072992700729928e-06,
+      "loss": 0.0308,
+      "step": 2968
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 1.0036496350364965e-06,
+      "loss": 0.0503,
+      "step": 2969
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.0091,
+      "step": 2970
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 9.963503649635036e-07,
+      "loss": 0.0125,
+      "step": 2971
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 9.927007299270073e-07,
+      "loss": 0.0239,
+      "step": 2972
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 9.89051094890511e-07,
+      "loss": 0.0186,
+      "step": 2973
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 9.854014598540146e-07,
+      "loss": 0.0294,
+      "step": 2974
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 9.817518248175183e-07,
+      "loss": 0.0424,
+      "step": 2975
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 9.78102189781022e-07,
+      "loss": 0.0193,
+      "step": 2976
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 9.744525547445256e-07,
+      "loss": 0.0105,
+      "step": 2977
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 9.708029197080293e-07,
+      "loss": 0.0109,
+      "step": 2978
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 9.67153284671533e-07,
+      "loss": 0.02,
+      "step": 2979
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 9.635036496350364e-07,
+      "loss": 0.0161,
+      "step": 2980
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 9.598540145985401e-07,
+      "loss": 0.0111,
+      "step": 2981
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 9.56204379562044e-07,
+      "loss": 0.0197,
+      "step": 2982
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 9.525547445255476e-07,
+      "loss": 0.0136,
+      "step": 2983
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 9.489051094890511e-07,
+      "loss": 0.0535,
+      "step": 2984
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 9.452554744525548e-07,
+      "loss": 0.0503,
+      "step": 2985
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 9.416058394160585e-07,
+      "loss": 0.015,
+      "step": 2986
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 9.379562043795621e-07,
+      "loss": 0.0079,
+      "step": 2987
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 9.343065693430657e-07,
+      "loss": 0.0279,
+      "step": 2988
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 9.306569343065694e-07,
+      "loss": 0.0265,
+      "step": 2989
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 9.27007299270073e-07,
+      "loss": 0.02,
+      "step": 2990
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 9.233576642335767e-07,
+      "loss": 0.024,
+      "step": 2991
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 9.197080291970803e-07,
+      "loss": 0.0092,
+      "step": 2992
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 9.16058394160584e-07,
+      "loss": 0.0233,
+      "step": 2993
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 9.124087591240876e-07,
+      "loss": 0.0404,
+      "step": 2994
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 9.087591240875913e-07,
+      "loss": 0.0119,
+      "step": 2995
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 9.051094890510951e-07,
+      "loss": 0.0079,
+      "step": 2996
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 9.014598540145985e-07,
+      "loss": 0.0441,
+      "step": 2997
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 8.978102189781022e-07,
+      "loss": 0.0279,
+      "step": 2998
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 8.94160583941606e-07,
+      "loss": 0.0329,
+      "step": 2999
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 8.905109489051097e-07,
+      "loss": 0.0365,
+      "step": 3000
+    },
+    {
+      "epoch": 27.77,
+      "eval_loss": 1.2935019731521606,
+      "eval_runtime": 64.4179,
+      "eval_samples_per_second": 25.567,
+      "eval_steps_per_second": 1.599,
+      "eval_wer": 0.9597589623123276,
+      "step": 3000
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 8.868613138686131e-07,
+      "loss": 0.0066,
+      "step": 3001
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 8.832116788321169e-07,
+      "loss": 0.0129,
+      "step": 3002
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 8.795620437956206e-07,
+      "loss": 0.0318,
+      "step": 3003
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 8.759124087591242e-07,
+      "loss": 0.0612,
+      "step": 3004
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 8.722627737226278e-07,
+      "loss": 0.0203,
+      "step": 3005
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 8.686131386861315e-07,
+      "loss": 0.0149,
+      "step": 3006
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 8.649635036496351e-07,
+      "loss": 0.0207,
+      "step": 3007
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 8.613138686131388e-07,
+      "loss": 0.0315,
+      "step": 3008
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 8.576642335766425e-07,
+      "loss": 0.0331,
+      "step": 3009
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 8.54014598540146e-07,
+      "loss": 0.0396,
+      "step": 3010
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 8.503649635036497e-07,
+      "loss": 0.0136,
+      "step": 3011
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 8.467153284671534e-07,
+      "loss": 0.0449,
+      "step": 3012
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 8.43065693430657e-07,
+      "loss": 0.0188,
+      "step": 3013
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 8.394160583941606e-07,
+      "loss": 0.0141,
+      "step": 3014
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 8.357664233576643e-07,
+      "loss": 0.0305,
+      "step": 3015
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 8.32116788321168e-07,
+      "loss": 0.028,
+      "step": 3016
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 8.284671532846716e-07,
+      "loss": 0.0152,
+      "step": 3017
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 8.248175182481752e-07,
+      "loss": 0.0051,
+      "step": 3018
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 8.211678832116789e-07,
+      "loss": 0.0269,
+      "step": 3019
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 8.175182481751825e-07,
+      "loss": 0.0245,
+      "step": 3020
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 8.138686131386862e-07,
+      "loss": 0.0326,
+      "step": 3021
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 8.102189781021899e-07,
+      "loss": 0.0143,
+      "step": 3022
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 8.065693430656934e-07,
+      "loss": 0.0122,
+      "step": 3023
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 8.029197080291971e-07,
+      "loss": 0.024,
+      "step": 3024
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 7.992700729927008e-07,
+      "loss": 0.0139,
+      "step": 3025
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 7.956204379562045e-07,
+      "loss": 0.0198,
+      "step": 3026
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 7.91970802919708e-07,
+      "loss": 0.0289,
+      "step": 3027
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 7.883211678832117e-07,
+      "loss": 0.0196,
+      "step": 3028
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 7.846715328467154e-07,
+      "loss": 0.015,
+      "step": 3029
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 7.810218978102191e-07,
+      "loss": 0.0211,
+      "step": 3030
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 7.773722627737226e-07,
+      "loss": 0.0387,
+      "step": 3031
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 7.737226277372263e-07,
+      "loss": 0.022,
+      "step": 3032
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 7.7007299270073e-07,
+      "loss": 0.0333,
+      "step": 3033
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 7.664233576642337e-07,
+      "loss": 0.043,
+      "step": 3034
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 7.627737226277374e-07,
+      "loss": 0.0088,
+      "step": 3035
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 7.59124087591241e-07,
+      "loss": 0.0272,
+      "step": 3036
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 7.554744525547446e-07,
+      "loss": 0.0414,
+      "step": 3037
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 7.518248175182483e-07,
+      "loss": 0.0156,
+      "step": 3038
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 7.48175182481752e-07,
+      "loss": 0.0131,
+      "step": 3039
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 7.445255474452555e-07,
+      "loss": 0.0611,
+      "step": 3040
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 7.408759124087592e-07,
+      "loss": 0.033,
+      "step": 3041
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 7.372262773722629e-07,
+      "loss": 0.0328,
+      "step": 3042
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 7.335766423357665e-07,
+      "loss": 0.0424,
+      "step": 3043
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 7.299270072992701e-07,
+      "loss": 0.0297,
+      "step": 3044
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 7.262773722627738e-07,
+      "loss": 0.0259,
+      "step": 3045
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 7.226277372262774e-07,
+      "loss": 0.0184,
+      "step": 3046
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 7.189781021897811e-07,
+      "loss": 0.0344,
+      "step": 3047
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 7.153284671532847e-07,
+      "loss": 0.0266,
+      "step": 3048
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 7.116788321167883e-07,
+      "loss": 0.0304,
+      "step": 3049
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 7.08029197080292e-07,
+      "loss": 0.0054,
+      "step": 3050
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 7.043795620437957e-07,
+      "loss": 0.0204,
+      "step": 3051
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 7.007299270072994e-07,
+      "loss": 0.0271,
+      "step": 3052
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 6.970802919708029e-07,
+      "loss": 0.034,
+      "step": 3053
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 6.934306569343066e-07,
+      "loss": 0.0122,
+      "step": 3054
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 6.897810218978103e-07,
+      "loss": 0.0252,
+      "step": 3055
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 6.861313868613139e-07,
+      "loss": 0.0289,
+      "step": 3056
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 6.824817518248175e-07,
+      "loss": 0.0112,
+      "step": 3057
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 6.788321167883212e-07,
+      "loss": 0.0336,
+      "step": 3058
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 6.751824817518248e-07,
+      "loss": 0.0387,
+      "step": 3059
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 6.715328467153285e-07,
+      "loss": 0.0349,
+      "step": 3060
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 6.678832116788321e-07,
+      "loss": 0.0436,
+      "step": 3061
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 6.642335766423358e-07,
+      "loss": 0.0473,
+      "step": 3062
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 6.605839416058394e-07,
+      "loss": 0.035,
+      "step": 3063
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 6.569343065693432e-07,
+      "loss": 0.0531,
+      "step": 3064
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 6.532846715328469e-07,
+      "loss": 0.018,
+      "step": 3065
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 6.496350364963503e-07,
+      "loss": 0.0056,
+      "step": 3066
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 6.459854014598541e-07,
+      "loss": 0.024,
+      "step": 3067
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 6.423357664233578e-07,
+      "loss": 0.0217,
+      "step": 3068
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 6.386861313868614e-07,
+      "loss": 0.0193,
+      "step": 3069
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 6.35036496350365e-07,
+      "loss": 0.0168,
+      "step": 3070
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 6.313868613138687e-07,
+      "loss": 0.0092,
+      "step": 3071
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 6.277372262773724e-07,
+      "loss": 0.0103,
+      "step": 3072
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 6.240875912408759e-07,
+      "loss": 0.029,
+      "step": 3073
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 6.204379562043796e-07,
+      "loss": 0.0244,
+      "step": 3074
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 6.167883211678833e-07,
+      "loss": 0.0171,
+      "step": 3075
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 6.131386861313869e-07,
+      "loss": 0.0273,
+      "step": 3076
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 6.094890510948906e-07,
+      "loss": 0.0446,
+      "step": 3077
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 6.058394160583942e-07,
+      "loss": 0.0045,
+      "step": 3078
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 6.021897810218979e-07,
+      "loss": 0.0371,
+      "step": 3079
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 5.985401459854015e-07,
+      "loss": 0.0358,
+      "step": 3080
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 5.948905109489052e-07,
+      "loss": 0.0136,
+      "step": 3081
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 5.912408759124088e-07,
+      "loss": 0.0236,
+      "step": 3082
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 5.875912408759125e-07,
+      "loss": 0.0103,
+      "step": 3083
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 5.839416058394161e-07,
+      "loss": 0.0231,
+      "step": 3084
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 5.802919708029198e-07,
+      "loss": 0.0537,
+      "step": 3085
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 5.766423357664234e-07,
+      "loss": 0.0253,
+      "step": 3086
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 5.729927007299271e-07,
+      "loss": 0.0444,
+      "step": 3087
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 5.693430656934307e-07,
+      "loss": 0.0069,
+      "step": 3088
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 5.656934306569343e-07,
+      "loss": 0.0205,
+      "step": 3089
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 5.62043795620438e-07,
+      "loss": 0.0192,
+      "step": 3090
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 5.583941605839417e-07,
+      "loss": 0.0179,
+      "step": 3091
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 5.547445255474453e-07,
+      "loss": 0.0111,
+      "step": 3092
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 5.510948905109489e-07,
+      "loss": 0.0641,
+      "step": 3093
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 5.474452554744526e-07,
+      "loss": 0.0245,
+      "step": 3094
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 5.437956204379562e-07,
+      "loss": 0.0344,
+      "step": 3095
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 5.401459854014599e-07,
+      "loss": 0.029,
+      "step": 3096
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 5.364963503649635e-07,
+      "loss": 0.0305,
+      "step": 3097
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 5.328467153284673e-07,
+      "loss": 0.0157,
+      "step": 3098
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 5.291970802919708e-07,
+      "loss": 0.0339,
+      "step": 3099
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 5.255474452554745e-07,
+      "loss": 0.0105,
+      "step": 3100
+    },
+    {
+      "epoch": 28.7,
+      "eval_loss": 1.2760541439056396,
+      "eval_runtime": 65.306,
+      "eval_samples_per_second": 25.22,
+      "eval_steps_per_second": 1.577,
+      "eval_wer": 0.9588397507915433,
+      "step": 3100
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 5.218978102189782e-07,
+      "loss": 0.0252,
+      "step": 3101
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 5.182481751824818e-07,
+      "loss": 0.0111,
+      "step": 3102
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 5.145985401459854e-07,
+      "loss": 0.0141,
+      "step": 3103
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 5.109489051094891e-07,
+      "loss": 0.0146,
+      "step": 3104
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 5.072992700729927e-07,
+      "loss": 0.018,
+      "step": 3105
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 5.036496350364964e-07,
+      "loss": 0.0332,
+      "step": 3106
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 0.0182,
+      "step": 3107
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 4.963503649635036e-07,
+      "loss": 0.0126,
+      "step": 3108
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 4.927007299270073e-07,
+      "loss": 0.0274,
+      "step": 3109
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 4.89051094890511e-07,
+      "loss": 0.0144,
+      "step": 3110
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 4.854014598540147e-07,
+      "loss": 0.0391,
+      "step": 3111
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 4.817518248175182e-07,
+      "loss": 0.0335,
+      "step": 3112
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 4.78102189781022e-07,
+      "loss": 0.0556,
+      "step": 3113
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 4.7445255474452557e-07,
+      "loss": 0.0144,
+      "step": 3114
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 4.7080291970802924e-07,
+      "loss": 0.023,
+      "step": 3115
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 4.6715328467153285e-07,
+      "loss": 0.0149,
+      "step": 3116
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 4.635036496350365e-07,
+      "loss": 0.0092,
+      "step": 3117
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 4.5985401459854014e-07,
+      "loss": 0.0272,
+      "step": 3118
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 4.562043795620438e-07,
+      "loss": 0.0214,
+      "step": 3119
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 4.5255474452554754e-07,
+      "loss": 0.0053,
+      "step": 3120
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 4.489051094890511e-07,
+      "loss": 0.0382,
+      "step": 3121
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 4.452554744525548e-07,
+      "loss": 0.0197,
+      "step": 3122
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 4.4160583941605844e-07,
+      "loss": 0.0178,
+      "step": 3123
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 4.379562043795621e-07,
+      "loss": 0.0359,
+      "step": 3124
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 4.3430656934306573e-07,
+      "loss": 0.0255,
+      "step": 3125
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 4.306569343065694e-07,
+      "loss": 0.0034,
+      "step": 3126
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 4.27007299270073e-07,
+      "loss": 0.0252,
+      "step": 3127
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 4.233576642335767e-07,
+      "loss": 0.0313,
+      "step": 3128
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 4.197080291970803e-07,
+      "loss": 0.0191,
+      "step": 3129
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 4.16058394160584e-07,
+      "loss": 0.0192,
+      "step": 3130
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 4.124087591240876e-07,
+      "loss": 0.0167,
+      "step": 3131
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 4.0875912408759127e-07,
+      "loss": 0.0212,
+      "step": 3132
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 4.0510948905109494e-07,
+      "loss": 0.0532,
+      "step": 3133
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 4.0145985401459856e-07,
+      "loss": 0.0077,
+      "step": 3134
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.9781021897810223e-07,
+      "loss": 0.0168,
+      "step": 3135
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.9416058394160584e-07,
+      "loss": 0.0076,
+      "step": 3136
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.9051094890510957e-07,
+      "loss": 0.0036,
+      "step": 3137
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.8686131386861313e-07,
+      "loss": 0.0285,
+      "step": 3138
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.8321167883211686e-07,
+      "loss": 0.0097,
+      "step": 3139
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.795620437956205e-07,
+      "loss": 0.0305,
+      "step": 3140
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.7591240875912414e-07,
+      "loss": 0.0204,
+      "step": 3141
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.7226277372262776e-07,
+      "loss": 0.0425,
+      "step": 3142
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.6861313868613143e-07,
+      "loss": 0.0153,
+      "step": 3143
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.6496350364963505e-07,
+      "loss": 0.0303,
+      "step": 3144
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.613138686131387e-07,
+      "loss": 0.0096,
+      "step": 3145
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.5766423357664234e-07,
+      "loss": 0.0148,
+      "step": 3146
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.54014598540146e-07,
+      "loss": 0.0142,
+      "step": 3147
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.503649635036497e-07,
+      "loss": 0.0211,
+      "step": 3148
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.467153284671533e-07,
+      "loss": 0.03,
+      "step": 3149
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.4306569343065697e-07,
+      "loss": 0.0182,
+      "step": 3150
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.394160583941606e-07,
+      "loss": 0.0124,
+      "step": 3151
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.3576642335766426e-07,
+      "loss": 0.0135,
+      "step": 3152
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.321167883211679e-07,
+      "loss": 0.0081,
+      "step": 3153
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.284671532846716e-07,
+      "loss": 0.0077,
+      "step": 3154
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.2481751824817516e-07,
+      "loss": 0.0375,
+      "step": 3155
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.211678832116789e-07,
+      "loss": 0.033,
+      "step": 3156
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.175182481751825e-07,
+      "loss": 0.024,
+      "step": 3157
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.138686131386862e-07,
+      "loss": 0.0268,
+      "step": 3158
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.102189781021898e-07,
+      "loss": 0.0358,
+      "step": 3159
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.0656934306569346e-07,
+      "loss": 0.03,
+      "step": 3160
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.029197080291971e-07,
+      "loss": 0.0167,
+      "step": 3161
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 2.9927007299270075e-07,
+      "loss": 0.0082,
+      "step": 3162
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 2.956204379562044e-07,
+      "loss": 0.0155,
+      "step": 3163
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 2.9197080291970804e-07,
+      "loss": 0.0283,
+      "step": 3164
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 2.883211678832117e-07,
+      "loss": 0.0174,
+      "step": 3165
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 2.8467153284671533e-07,
+      "loss": 0.039,
+      "step": 3166
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 2.81021897810219e-07,
+      "loss": 0.013,
+      "step": 3167
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 2.7737226277372267e-07,
+      "loss": 0.0129,
+      "step": 3168
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 2.737226277372263e-07,
+      "loss": 0.0399,
+      "step": 3169
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 2.7007299270072996e-07,
+      "loss": 0.0048,
+      "step": 3170
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 2.6642335766423363e-07,
+      "loss": 0.0093,
+      "step": 3171
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 2.6277372262773725e-07,
+      "loss": 0.0223,
+      "step": 3172
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 2.591240875912409e-07,
+      "loss": 0.0103,
+      "step": 3173
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 2.5547445255474454e-07,
+      "loss": 0.0333,
+      "step": 3174
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 2.518248175182482e-07,
+      "loss": 0.0046,
+      "step": 3175
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 2.481751824817518e-07,
+      "loss": 0.0403,
+      "step": 3176
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 2.445255474452555e-07,
+      "loss": 0.0161,
+      "step": 3177
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 2.408759124087591e-07,
+      "loss": 0.0227,
+      "step": 3178
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 2.3722627737226278e-07,
+      "loss": 0.0261,
+      "step": 3179
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 2.3357664233576643e-07,
+      "loss": 0.0013,
+      "step": 3180
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 2.2992700729927007e-07,
+      "loss": 0.0093,
+      "step": 3181
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 2.2627737226277377e-07,
+      "loss": 0.0081,
+      "step": 3182
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 2.226277372262774e-07,
+      "loss": 0.0159,
+      "step": 3183
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 2.1897810218978106e-07,
+      "loss": 0.0099,
+      "step": 3184
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 2.153284671532847e-07,
+      "loss": 0.0181,
+      "step": 3185
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 2.1167883211678835e-07,
+      "loss": 0.0274,
+      "step": 3186
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 2.08029197080292e-07,
+      "loss": 0.0242,
+      "step": 3187
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 2.0437956204379563e-07,
+      "loss": 0.0125,
+      "step": 3188
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 2.0072992700729928e-07,
+      "loss": 0.0181,
+      "step": 3189
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 1.9708029197080292e-07,
+      "loss": 0.0144,
+      "step": 3190
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 1.9343065693430657e-07,
+      "loss": 0.0235,
+      "step": 3191
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.8978102189781024e-07,
+      "loss": 0.0169,
+      "step": 3192
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.8613138686131388e-07,
+      "loss": 0.0131,
+      "step": 3193
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 1.8248175182481753e-07,
+      "loss": 0.0179,
+      "step": 3194
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 1.7883211678832117e-07,
+      "loss": 0.0222,
+      "step": 3195
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 1.7518248175182484e-07,
+      "loss": 0.0156,
+      "step": 3196
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 1.7153284671532848e-07,
+      "loss": 0.0209,
+      "step": 3197
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.6788321167883213e-07,
+      "loss": 0.0136,
+      "step": 3198
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.642335766423358e-07,
+      "loss": 0.0123,
+      "step": 3199
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.6058394160583944e-07,
+      "loss": 0.021,
+      "step": 3200
+    },
+    {
+      "epoch": 29.63,
+      "eval_loss": 1.2686101198196411,
+      "eval_runtime": 64.5688,
+      "eval_samples_per_second": 25.508,
+      "eval_steps_per_second": 1.595,
+      "eval_wer": 0.9528138085997344,
+      "step": 3200
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.569343065693431e-07,
+      "loss": 0.0201,
+      "step": 3201
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 1.5328467153284673e-07,
+      "loss": 0.0046,
+      "step": 3202
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 1.4963503649635038e-07,
+      "loss": 0.0074,
+      "step": 3203
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 1.4598540145985402e-07,
+      "loss": 0.0243,
+      "step": 3204
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 1.4233576642335766e-07,
+      "loss": 0.0128,
+      "step": 3205
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.3868613138686134e-07,
+      "loss": 0.0236,
+      "step": 3206
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 1.3503649635036498e-07,
+      "loss": 0.0394,
+      "step": 3207
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 1.3138686131386862e-07,
+      "loss": 0.008,
+      "step": 3208
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 1.2773722627737227e-07,
+      "loss": 0.0347,
+      "step": 3209
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 1.240875912408759e-07,
+      "loss": 0.0081,
+      "step": 3210
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 1.2043795620437956e-07,
+      "loss": 0.0448,
+      "step": 3211
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.1678832116788321e-07,
+      "loss": 0.0182,
+      "step": 3212
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.1313868613138688e-07,
+      "loss": 0.0227,
+      "step": 3213
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 1.0948905109489053e-07,
+      "loss": 0.0187,
+      "step": 3214
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 1.0583941605839417e-07,
+      "loss": 0.0141,
+      "step": 3215
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 1.0218978102189782e-07,
+      "loss": 0.0124,
+      "step": 3216
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 9.854014598540146e-08,
+      "loss": 0.004,
+      "step": 3217
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 9.489051094890512e-08,
+      "loss": 0.0132,
+      "step": 3218
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 9.124087591240876e-08,
+      "loss": 0.0318,
+      "step": 3219
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 8.759124087591242e-08,
+      "loss": 0.0076,
+      "step": 3220
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 8.394160583941606e-08,
+      "loss": 0.0108,
+      "step": 3221
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 8.029197080291972e-08,
+      "loss": 0.0276,
+      "step": 3222
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 7.664233576642337e-08,
+      "loss": 0.028,
+      "step": 3223
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 7.299270072992701e-08,
+      "loss": 0.0096,
+      "step": 3224
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 6.934306569343067e-08,
+      "loss": 0.0216,
+      "step": 3225
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 6.569343065693431e-08,
+      "loss": 0.0389,
+      "step": 3226
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 6.204379562043796e-08,
+      "loss": 0.0376,
+      "step": 3227
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 5.839416058394161e-08,
+      "loss": 0.0402,
+      "step": 3228
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 5.4744525547445264e-08,
+      "loss": 0.0228,
+      "step": 3229
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 5.109489051094891e-08,
+      "loss": 0.0206,
+      "step": 3230
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 4.744525547445256e-08,
+      "loss": 0.048,
+      "step": 3231
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 4.379562043795621e-08,
+      "loss": 0.0107,
+      "step": 3232
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 4.014598540145986e-08,
+      "loss": 0.0244,
+      "step": 3233
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.6496350364963505e-08,
+      "loss": 0.0156,
+      "step": 3234
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.2846715328467156e-08,
+      "loss": 0.0511,
+      "step": 3235
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 2.9197080291970803e-08,
+      "loss": 0.022,
+      "step": 3236
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 2.5547445255474454e-08,
+      "loss": 0.0131,
+      "step": 3237
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 2.1897810218978105e-08,
+      "loss": 0.0069,
+      "step": 3238
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 1.8248175182481753e-08,
+      "loss": 0.02,
+      "step": 3239
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.4598540145985402e-08,
+      "loss": 0.037,
+      "step": 3240
+    },
+    {
+      "epoch": 29.99,
+      "step": 3240,
+      "total_flos": 8.734490735348705e+19,
+      "train_loss": 0.37093941706698386,
+      "train_runtime": 17761.7738,
+      "train_samples_per_second": 5.874,
+      "train_steps_per_second": 0.182
+    }
+  ],
+  "max_steps": 3240,
+  "num_train_epochs": 30,
+  "total_flos": 8.734490735348705e+19,
+  "trial_name": null,
+  "trial_params": null
+}