diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,13294 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 20.0,
+  "global_step": 2180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0,
+      "loss": 38.5792,
+      "step": 1
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0,
+      "loss": 37.0162,
+      "step": 2
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-06,
+      "loss": 39.6498,
+      "step": 3
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2e-06,
+      "loss": 37.9353,
+      "step": 4
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2e-06,
+      "loss": 37.3926,
+      "step": 5
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 3e-06,
+      "loss": 38.9023,
+      "step": 6
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4e-06,
+      "loss": 37.7523,
+      "step": 7
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 5e-06,
+      "loss": 37.6599,
+      "step": 8
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6e-06,
+      "loss": 36.2328,
+      "step": 9
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7e-06,
+      "loss": 32.7578,
+      "step": 10
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 8e-06,
+      "loss": 36.521,
+      "step": 11
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9e-06,
+      "loss": 36.5731,
+      "step": 12
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1e-05,
+      "loss": 38.2846,
+      "step": 13
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1e-05,
+      "loss": 33.8493,
+      "step": 14
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2e-05,
+      "loss": 36.5712,
+      "step": 15
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3e-05,
+      "loss": 34.8204,
+      "step": 16
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.4e-05,
+      "loss": 33.1536,
+      "step": 17
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.5e-05,
+      "loss": 34.1219,
+      "step": 18
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.6e-05,
+      "loss": 36.0866,
+      "step": 19
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 32.673,
+      "step": 20
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8e-05,
+      "loss": 37.8781,
+      "step": 21
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9e-05,
+      "loss": 31.2217,
+      "step": 22
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2e-05,
+      "loss": 37.2818,
+      "step": 23
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.1000000000000002e-05,
+      "loss": 34.9108,
+      "step": 24
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.2e-05,
+      "loss": 37.8733,
+      "step": 25
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.3e-05,
+      "loss": 38.0091,
+      "step": 26
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.4e-05,
+      "loss": 36.1274,
+      "step": 27
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.5e-05,
+      "loss": 34.8621,
+      "step": 28
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6e-05,
+      "loss": 31.4434,
+      "step": 29
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.7e-05,
+      "loss": 28.9744,
+      "step": 30
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.8e-05,
+      "loss": 29.4037,
+      "step": 31
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9000000000000004e-05,
+      "loss": 29.3016,
+      "step": 32
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3e-05,
+      "loss": 28.9117,
+      "step": 33
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.1e-05,
+      "loss": 28.1706,
+      "step": 34
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.2e-05,
+      "loss": 26.2272,
+      "step": 35
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3e-05,
+      "loss": 24.7177,
+      "step": 36
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 24.2448,
+      "step": 37
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 23.2494,
+      "step": 38
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6e-05,
+      "loss": 23.724,
+      "step": 39
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7e-05,
+      "loss": 22.8256,
+      "step": 40
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.8e-05,
+      "loss": 22.0102,
+      "step": 41
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.9e-05,
+      "loss": 20.4507,
+      "step": 42
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4e-05,
+      "loss": 19.7054,
+      "step": 43
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.1e-05,
+      "loss": 18.8839,
+      "step": 44
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 18.8487,
+      "step": 45
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.2999999999999995e-05,
+      "loss": 19.5023,
+      "step": 46
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.4e-05,
+      "loss": 20.0338,
+      "step": 47
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 21.0323,
+      "step": 48
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.6e-05,
+      "loss": 20.2858,
+      "step": 49
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.7000000000000004e-05,
+      "loss": 17.9252,
+      "step": 50
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.8e-05,
+      "loss": 20.4729,
+      "step": 51
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.9000000000000005e-05,
+      "loss": 18.9176,
+      "step": 52
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5e-05,
+      "loss": 20.5586,
+      "step": 53
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1e-05,
+      "loss": 19.3151,
+      "step": 54
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2e-05,
+      "loss": 17.6831,
+      "step": 55
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3e-05,
+      "loss": 18.0072,
+      "step": 56
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.4e-05,
+      "loss": 18.041,
+      "step": 57
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.5e-05,
+      "loss": 17.3918,
+      "step": 58
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 5.6e-05,
+      "loss": 16.5541,
+      "step": 59
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 5.7e-05,
+      "loss": 16.5265,
+      "step": 60
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.800000000000001e-05,
+      "loss": 17.0845,
+      "step": 61
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 5.9e-05,
+      "loss": 16.8278,
+      "step": 62
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 6e-05,
+      "loss": 16.1681,
+      "step": 63
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.1e-05,
+      "loss": 15.374,
+      "step": 64
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.2e-05,
+      "loss": 14.8997,
+      "step": 65
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.3e-05,
+      "loss": 14.3692,
+      "step": 66
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.4e-05,
+      "loss": 16.3112,
+      "step": 67
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 14.3636,
+      "step": 68
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.6e-05,
+      "loss": 14.9014,
+      "step": 69
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.7e-05,
+      "loss": 15.0173,
+      "step": 70
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 14.7624,
+      "step": 71
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.900000000000001e-05,
+      "loss": 13.5398,
+      "step": 72
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 13.993,
+      "step": 73
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.099999999999999e-05,
+      "loss": 14.214,
+      "step": 74
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.2e-05,
+      "loss": 14.3255,
+      "step": 75
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.2e-05,
+      "loss": 15.4427,
+      "step": 76
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3e-05,
+      "loss": 15.4043,
+      "step": 77
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.4e-05,
+      "loss": 15.5085,
+      "step": 78
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.5e-05,
+      "loss": 14.457,
+      "step": 79
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.6e-05,
+      "loss": 14.4357,
+      "step": 80
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 7.7e-05,
+      "loss": 13.9193,
+      "step": 81
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.8e-05,
+      "loss": 13.7483,
+      "step": 82
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.9e-05,
+      "loss": 13.7121,
+      "step": 83
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 8e-05,
+      "loss": 13.3313,
+      "step": 84
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.1e-05,
+      "loss": 13.4978,
+      "step": 85
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.2e-05,
+      "loss": 12.4055,
+      "step": 86
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 8.300000000000001e-05,
+      "loss": 12.8827,
+      "step": 87
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.400000000000001e-05,
+      "loss": 11.9554,
+      "step": 88
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.5e-05,
+      "loss": 12.515,
+      "step": 89
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.599999999999999e-05,
+      "loss": 12.1079,
+      "step": 90
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.7e-05,
+      "loss": 11.9346,
+      "step": 91
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.8e-05,
+      "loss": 12.0008,
+      "step": 92
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 8.9e-05,
+      "loss": 11.4768,
+      "step": 93
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 11.501,
+      "step": 94
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.1e-05,
+      "loss": 11.2109,
+      "step": 95
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.2e-05,
+      "loss": 11.383,
+      "step": 96
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.3e-05,
+      "loss": 11.4743,
+      "step": 97
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.400000000000001e-05,
+      "loss": 11.2624,
+      "step": 98
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.5e-05,
+      "loss": 10.2995,
+      "step": 99
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.6e-05,
+      "loss": 13.6687,
+      "step": 100
+    },
+    {
+      "epoch": 0.92,
+      "eval_loss": 12.456720352172852,
+      "eval_runtime": 35.8714,
+      "eval_samples_per_second": 45.914,
+      "eval_steps_per_second": 2.871,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.7e-05,
+      "loss": 11.5921,
+      "step": 101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.800000000000001e-05,
+      "loss": 11.1113,
+      "step": 102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.900000000000001e-05,
+      "loss": 10.4269,
+      "step": 103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001,
+      "loss": 10.6538,
+      "step": 104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.000101,
+      "loss": 10.1798,
+      "step": 105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000102,
+      "loss": 10.1534,
+      "step": 106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000103,
+      "loss": 9.3504,
+      "step": 107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.000104,
+      "loss": 9.7509,
+      "step": 108
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.000105,
+      "loss": 9.93,
+      "step": 109
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000106,
+      "loss": 9.5797,
+      "step": 110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000107,
+      "loss": 10.0161,
+      "step": 111
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000108,
+      "loss": 9.5943,
+      "step": 112
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.000109,
+      "loss": 9.5237,
+      "step": 113
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00011,
+      "loss": 8.9127,
+      "step": 114
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.000111,
+      "loss": 9.7015,
+      "step": 115
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.000112,
+      "loss": 8.7284,
+      "step": 116
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00011300000000000001,
+      "loss": 9.1047,
+      "step": 117
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000114,
+      "loss": 8.1133,
+      "step": 118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.000115,
+      "loss": 7.8652,
+      "step": 119
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00011600000000000001,
+      "loss": 8.3011,
+      "step": 120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00011700000000000001,
+      "loss": 8.0768,
+      "step": 121
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.000118,
+      "loss": 7.885,
+      "step": 122
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00011899999999999999,
+      "loss": 7.5906,
+      "step": 123
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012,
+      "loss": 7.3021,
+      "step": 124
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.000121,
+      "loss": 7.5019,
+      "step": 125
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000122,
+      "loss": 6.9366,
+      "step": 126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000123,
+      "loss": 6.7988,
+      "step": 127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000124,
+      "loss": 7.0515,
+      "step": 128
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.000125,
+      "loss": 6.3643,
+      "step": 129
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000126,
+      "loss": 6.438,
+      "step": 130
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.000127,
+      "loss": 6.3211,
+      "step": 131
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.000128,
+      "loss": 6.4368,
+      "step": 132
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00012900000000000002,
+      "loss": 6.2782,
+      "step": 133
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 7.5731,
+      "step": 134
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.000131,
+      "loss": 6.3343,
+      "step": 135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.000132,
+      "loss": 6.0038,
+      "step": 136
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.000133,
+      "loss": 6.2357,
+      "step": 137
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.000134,
+      "loss": 5.9097,
+      "step": 138
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.000135,
+      "loss": 5.8111,
+      "step": 139
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 5.6715,
+      "step": 140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00013700000000000002,
+      "loss": 5.5072,
+      "step": 141
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00013800000000000002,
+      "loss": 5.5216,
+      "step": 142
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00013900000000000002,
+      "loss": 5.4142,
+      "step": 143
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 5.0936,
+      "step": 144
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 5.1922,
+      "step": 145
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 5.096,
+      "step": 146
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014299999999999998,
+      "loss": 4.9215,
+      "step": 147
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.000144,
+      "loss": 5.0267,
+      "step": 148
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.000145,
+      "loss": 4.7844,
+      "step": 149
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.000146,
+      "loss": 4.7324,
+      "step": 150
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.000147,
+      "loss": 4.653,
+      "step": 151
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.000148,
+      "loss": 4.5143,
+      "step": 152
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000149,
+      "loss": 4.5446,
+      "step": 153
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00015,
+      "loss": 4.5102,
+      "step": 154
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.000151,
+      "loss": 4.4171,
+      "step": 155
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.000152,
+      "loss": 4.4132,
+      "step": 156
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.000153,
+      "loss": 4.4959,
+      "step": 157
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000154,
+      "loss": 4.2654,
+      "step": 158
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.000155,
+      "loss": 4.4156,
+      "step": 159
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.000156,
+      "loss": 4.209,
+      "step": 160
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.000157,
+      "loss": 4.0884,
+      "step": 161
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.000158,
+      "loss": 4.0367,
+      "step": 162
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00015900000000000002,
+      "loss": 3.9577,
+      "step": 163
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00016,
+      "loss": 3.9339,
+      "step": 164
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.000161,
+      "loss": 3.8781,
+      "step": 165
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000162,
+      "loss": 3.8486,
+      "step": 166
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.000163,
+      "loss": 3.8499,
+      "step": 167
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.000164,
+      "loss": 3.7434,
+      "step": 168
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.000165,
+      "loss": 3.7463,
+      "step": 169
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00016600000000000002,
+      "loss": 3.6791,
+      "step": 170
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00016700000000000002,
+      "loss": 3.8176,
+      "step": 171
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00016800000000000002,
+      "loss": 3.636,
+      "step": 172
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00016900000000000002,
+      "loss": 3.6684,
+      "step": 173
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00017,
+      "loss": 3.5822,
+      "step": 174
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.000171,
+      "loss": 3.5428,
+      "step": 175
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 3.5849,
+      "step": 176
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000173,
+      "loss": 3.4901,
+      "step": 177
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.000174,
+      "loss": 3.5467,
+      "step": 178
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000175,
+      "loss": 3.5479,
+      "step": 179
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.000176,
+      "loss": 3.5558,
+      "step": 180
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.000177,
+      "loss": 3.5238,
+      "step": 181
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.000178,
+      "loss": 3.5594,
+      "step": 182
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.000179,
+      "loss": 3.5816,
+      "step": 183
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 3.622,
+      "step": 184
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.000181,
+      "loss": 3.4921,
+      "step": 185
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.000182,
+      "loss": 3.4206,
+      "step": 186
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.000183,
+      "loss": 3.4675,
+      "step": 187
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.000184,
+      "loss": 3.3869,
+      "step": 188
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.000185,
+      "loss": 3.4207,
+      "step": 189
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.000186,
+      "loss": 3.382,
+      "step": 190
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.000187,
+      "loss": 3.4191,
+      "step": 191
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00018800000000000002,
+      "loss": 3.4345,
+      "step": 192
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.000189,
+      "loss": 3.4167,
+      "step": 193
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00019,
+      "loss": 3.4177,
+      "step": 194
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.000191,
+      "loss": 3.4371,
+      "step": 195
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.000192,
+      "loss": 3.3914,
+      "step": 196
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.000193,
+      "loss": 3.4371,
+      "step": 197
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.000194,
+      "loss": 3.3699,
+      "step": 198
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 3.4108,
+      "step": 199
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00019600000000000002,
+      "loss": 3.4219,
+      "step": 200
+    },
+    {
+      "epoch": 1.83,
+      "eval_loss": 3.4703733921051025,
+      "eval_runtime": 36.3034,
+      "eval_samples_per_second": 45.368,
+      "eval_steps_per_second": 2.837,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00019700000000000002,
+      "loss": 3.4513,
+      "step": 201
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00019800000000000002,
+      "loss": 3.3579,
+      "step": 202
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.000199,
+      "loss": 3.4298,
+      "step": 203
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002,
+      "loss": 3.5704,
+      "step": 204
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.000201,
+      "loss": 3.3696,
+      "step": 205
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.000202,
+      "loss": 3.3891,
+      "step": 206
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00020300000000000003,
+      "loss": 3.4279,
+      "step": 207
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.000204,
+      "loss": 3.4615,
+      "step": 208
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.000205,
+      "loss": 3.557,
+      "step": 209
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.000206,
+      "loss": 3.3266,
+      "step": 210
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.000207,
+      "loss": 3.4195,
+      "step": 211
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.000208,
+      "loss": 3.3603,
+      "step": 212
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00020899999999999998,
+      "loss": 3.3904,
+      "step": 213
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00021,
+      "loss": 3.3271,
+      "step": 214
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.000211,
+      "loss": 3.3431,
+      "step": 215
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.000212,
+      "loss": 3.3564,
+      "step": 216
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.000213,
+      "loss": 3.3976,
+      "step": 217
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.000214,
+      "loss": 3.3896,
+      "step": 218
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.000215,
+      "loss": 3.3424,
+      "step": 219
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.000216,
+      "loss": 3.2927,
+      "step": 220
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00021700000000000002,
+      "loss": 3.3482,
+      "step": 221
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.000218,
+      "loss": 3.3295,
+      "step": 222
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.000219,
+      "loss": 3.3461,
+      "step": 223
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00022,
+      "loss": 3.3533,
+      "step": 224
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.000221,
+      "loss": 3.2775,
+      "step": 225
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.000222,
+      "loss": 3.3536,
+      "step": 226
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.000223,
+      "loss": 3.3446,
+      "step": 227
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.000224,
+      "loss": 3.2963,
+      "step": 228
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 3.2853,
+      "step": 229
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00022600000000000002,
+      "loss": 3.3773,
+      "step": 230
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00022700000000000002,
+      "loss": 3.3429,
+      "step": 231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.000228,
+      "loss": 3.3102,
+      "step": 232
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.000229,
+      "loss": 3.3508,
+      "step": 233
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00023,
+      "loss": 3.3172,
+      "step": 234
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.000231,
+      "loss": 3.3283,
+      "step": 235
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 3.3122,
+      "step": 236
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00023300000000000003,
+      "loss": 3.3299,
+      "step": 237
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.00023400000000000002,
+      "loss": 3.3632,
+      "step": 238
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.000235,
+      "loss": 3.2774,
+      "step": 239
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.000236,
+      "loss": 3.3316,
+      "step": 240
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.000237,
+      "loss": 3.319,
+      "step": 241
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00023799999999999998,
+      "loss": 3.2672,
+      "step": 242
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00023899999999999998,
+      "loss": 3.4331,
+      "step": 243
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00024,
+      "loss": 3.3221,
+      "step": 244
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.000241,
+      "loss": 3.321,
+      "step": 245
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.000242,
+      "loss": 3.3073,
+      "step": 246
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.000243,
+      "loss": 3.298,
+      "step": 247
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000244,
+      "loss": 3.3197,
+      "step": 248
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000245,
+      "loss": 3.3396,
+      "step": 249
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.000246,
+      "loss": 3.2956,
+      "step": 250
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.000247,
+      "loss": 3.2701,
+      "step": 251
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.000248,
+      "loss": 3.318,
+      "step": 252
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.000249,
+      "loss": 3.3168,
+      "step": 253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00025,
+      "loss": 3.2829,
+      "step": 254
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00025100000000000003,
+      "loss": 3.2794,
+      "step": 255
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.000252,
+      "loss": 3.3066,
+      "step": 256
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.000253,
+      "loss": 3.328,
+      "step": 257
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.000254,
+      "loss": 3.2912,
+      "step": 258
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.000255,
+      "loss": 3.3403,
+      "step": 259
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.000256,
+      "loss": 3.2552,
+      "step": 260
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.000257,
+      "loss": 3.2576,
+      "step": 261
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00025800000000000004,
+      "loss": 3.2743,
+      "step": 262
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.000259,
+      "loss": 3.2991,
+      "step": 263
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 3.32,
+      "step": 264
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.000261,
+      "loss": 3.2453,
+      "step": 265
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.000262,
+      "loss": 3.2475,
+      "step": 266
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.000263,
+      "loss": 3.3388,
+      "step": 267
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.000264,
+      "loss": 3.3154,
+      "step": 268
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 3.2821,
+      "step": 269
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.000266,
+      "loss": 3.2592,
+      "step": 270
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00026700000000000004,
+      "loss": 3.2073,
+      "step": 271
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.000268,
+      "loss": 3.3326,
+      "step": 272
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00026900000000000003,
+      "loss": 3.2591,
+      "step": 273
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00027,
+      "loss": 3.3079,
+      "step": 274
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00027100000000000003,
+      "loss": 3.2713,
+      "step": 275
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 3.2584,
+      "step": 276
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.000273,
+      "loss": 3.2096,
+      "step": 277
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00027400000000000005,
+      "loss": 3.2723,
+      "step": 278
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.000275,
+      "loss": 3.2386,
+      "step": 279
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00027600000000000004,
+      "loss": 3.2419,
+      "step": 280
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.000277,
+      "loss": 3.1943,
+      "step": 281
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.00027800000000000004,
+      "loss": 3.2158,
+      "step": 282
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.000279,
+      "loss": 3.277,
+      "step": 283
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 3.1875,
+      "step": 284
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00028100000000000005,
+      "loss": 3.2061,
+      "step": 285
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 3.1763,
+      "step": 286
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.000283,
+      "loss": 3.21,
+      "step": 287
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00028399999999999996,
+      "loss": 3.239,
+      "step": 288
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.000285,
+      "loss": 3.2045,
+      "step": 289
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00028599999999999996,
+      "loss": 3.2626,
+      "step": 290
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.000287,
+      "loss": 3.2211,
+      "step": 291
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.000288,
+      "loss": 3.3022,
+      "step": 292
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.000289,
+      "loss": 3.2887,
+      "step": 293
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00029,
+      "loss": 3.2014,
+      "step": 294
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 3.2001,
+      "step": 295
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.000292,
+      "loss": 3.2143,
+      "step": 296
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00029299999999999997,
+      "loss": 3.1513,
+      "step": 297
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.000294,
+      "loss": 3.1761,
+      "step": 298
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.000295,
+      "loss": 3.1955,
+      "step": 299
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.000296,
+      "loss": 3.1846,
+      "step": 300
+    },
+    {
+      "epoch": 2.75,
+      "eval_loss": 3.2280642986297607,
+      "eval_runtime": 36.3511,
+      "eval_samples_per_second": 45.308,
+      "eval_steps_per_second": 2.833,
+      "eval_wer": 0.9934558117195005,
+      "step": 300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.000297,
+      "loss": 3.1597,
+      "step": 301
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.000298,
+      "loss": 3.1815,
+      "step": 302
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.000299,
+      "loss": 3.1767,
+      "step": 303
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0003,
+      "loss": 3.2083,
+      "step": 304
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.000301,
+      "loss": 3.1591,
+      "step": 305
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.000302,
+      "loss": 3.1413,
+      "step": 306
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.000303,
+      "loss": 3.2094,
+      "step": 307
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000304,
+      "loss": 3.1181,
+      "step": 308
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000305,
+      "loss": 3.1231,
+      "step": 309
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.000306,
+      "loss": 3.1626,
+      "step": 310
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.000307,
+      "loss": 3.1493,
+      "step": 311
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.000308,
+      "loss": 3.0951,
+      "step": 312
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00030900000000000003,
+      "loss": 3.1452,
+      "step": 313
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00031,
+      "loss": 3.159,
+      "step": 314
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.000311,
+      "loss": 3.1309,
+      "step": 315
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.000312,
+      "loss": 3.1552,
+      "step": 316
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.000313,
+      "loss": 3.1456,
+      "step": 317
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.000314,
+      "loss": 3.2217,
+      "step": 318
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000315,
+      "loss": 3.115,
+      "step": 319
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.000316,
+      "loss": 3.1785,
+      "step": 320
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.000317,
+      "loss": 3.1313,
+      "step": 321
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.00031800000000000003,
+      "loss": 3.1596,
+      "step": 322
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.000319,
+      "loss": 3.1644,
+      "step": 323
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00032,
+      "loss": 3.1194,
+      "step": 324
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.000321,
+      "loss": 3.1795,
+      "step": 325
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.000322,
+      "loss": 3.0813,
+      "step": 326
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.000323,
+      "loss": 3.2504,
+      "step": 327
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.000324,
+      "loss": 3.1434,
+      "step": 328
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 3.0752,
+      "step": 329
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.000326,
+      "loss": 3.1518,
+      "step": 330
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00032700000000000003,
+      "loss": 3.0821,
+      "step": 331
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.000328,
+      "loss": 3.1113,
+      "step": 332
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.00032900000000000003,
+      "loss": 3.1609,
+      "step": 333
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.00033,
+      "loss": 3.1282,
+      "step": 334
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.000331,
+      "loss": 3.0544,
+      "step": 335
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 3.0813,
+      "step": 336
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.000333,
+      "loss": 3.1289,
+      "step": 337
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00033400000000000004,
+      "loss": 3.0654,
+      "step": 338
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.000335,
+      "loss": 3.05,
+      "step": 339
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00033600000000000004,
+      "loss": 3.086,
+      "step": 340
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.000337,
+      "loss": 2.9991,
+      "step": 341
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00033800000000000003,
+      "loss": 3.0787,
+      "step": 342
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00033900000000000005,
+      "loss": 3.0205,
+      "step": 343
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00034,
+      "loss": 3.0272,
+      "step": 344
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00034100000000000005,
+      "loss": 2.9642,
+      "step": 345
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.000342,
+      "loss": 2.9674,
+      "step": 346
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00034300000000000004,
+      "loss": 3.0552,
+      "step": 347
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00034399999999999996,
+      "loss": 2.9619,
+      "step": 348
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.000345,
+      "loss": 3.0189,
+      "step": 349
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.000346,
+      "loss": 2.9918,
+      "step": 350
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.000347,
+      "loss": 2.8969,
+      "step": 351
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.000348,
+      "loss": 2.998,
+      "step": 352
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00034899999999999997,
+      "loss": 2.9714,
+      "step": 353
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00035,
+      "loss": 2.9747,
+      "step": 354
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00035099999999999997,
+      "loss": 2.904,
+      "step": 355
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.000352,
+      "loss": 2.9562,
+      "step": 356
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00035299999999999996,
+      "loss": 2.9535,
+      "step": 357
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.000354,
+      "loss": 2.9477,
+      "step": 358
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.000355,
+      "loss": 2.9082,
+      "step": 359
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.000356,
+      "loss": 2.8054,
+      "step": 360
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.000357,
+      "loss": 2.8714,
+      "step": 361
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.000358,
+      "loss": 2.8054,
+      "step": 362
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.000359,
+      "loss": 2.7559,
+      "step": 363
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 2.759,
+      "step": 364
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.000361,
+      "loss": 2.846,
+      "step": 365
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.000362,
+      "loss": 2.7483,
+      "step": 366
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.000363,
+      "loss": 2.8022,
+      "step": 367
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.000364,
+      "loss": 2.7079,
+      "step": 368
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.000365,
+      "loss": 2.5973,
+      "step": 369
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.000366,
+      "loss": 2.6236,
+      "step": 370
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000367,
+      "loss": 2.6806,
+      "step": 371
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.000368,
+      "loss": 2.7707,
+      "step": 372
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.000369,
+      "loss": 2.7137,
+      "step": 373
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00037,
+      "loss": 2.5566,
+      "step": 374
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.000371,
+      "loss": 2.5661,
+      "step": 375
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.000372,
+      "loss": 2.6638,
+      "step": 376
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.000373,
+      "loss": 2.6151,
+      "step": 377
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.000374,
+      "loss": 2.4922,
+      "step": 378
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.000375,
+      "loss": 2.5417,
+      "step": 379
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00037600000000000003,
+      "loss": 2.391,
+      "step": 380
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000377,
+      "loss": 2.4796,
+      "step": 381
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000378,
+      "loss": 2.4312,
+      "step": 382
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.000379,
+      "loss": 2.285,
+      "step": 383
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00038,
+      "loss": 2.2847,
+      "step": 384
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.000381,
+      "loss": 2.1494,
+      "step": 385
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.000382,
+      "loss": 2.1588,
+      "step": 386
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00038300000000000004,
+      "loss": 2.1269,
+      "step": 387
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.000384,
+      "loss": 1.9845,
+      "step": 388
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 2.1187,
+      "step": 389
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.000386,
+      "loss": 2.0799,
+      "step": 390
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00038700000000000003,
+      "loss": 2.1227,
+      "step": 391
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.000388,
+      "loss": 2.0675,
+      "step": 392
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.000389,
+      "loss": 2.0205,
+      "step": 393
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 2.0579,
+      "step": 394
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.000391,
+      "loss": 2.1067,
+      "step": 395
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 2.1526,
+      "step": 396
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.000393,
+      "loss": 1.7567,
+      "step": 397
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00039400000000000004,
+      "loss": 1.85,
+      "step": 398
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.000395,
+      "loss": 1.8782,
+      "step": 399
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 2.0076,
+      "step": 400
+    },
+    {
+      "epoch": 3.67,
+      "eval_loss": 1.7414734363555908,
+      "eval_runtime": 36.4256,
+      "eval_samples_per_second": 45.215,
+      "eval_steps_per_second": 2.828,
+      "eval_wer": 0.5221541786743515,
+      "step": 400
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00039700000000000005,
+      "loss": 2.0936,
+      "step": 401
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.000398,
+      "loss": 2.0625,
+      "step": 402
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00039900000000000005,
+      "loss": 1.8438,
+      "step": 403
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0004,
+      "loss": 1.8604,
+      "step": 404
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00040100000000000004,
+      "loss": 1.8046,
+      "step": 405
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.000402,
+      "loss": 1.7043,
+      "step": 406
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00040300000000000004,
+      "loss": 1.7386,
+      "step": 407
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.000404,
+      "loss": 1.7309,
+      "step": 408
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 1.5746,
+      "step": 409
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00040600000000000006,
+      "loss": 1.5256,
+      "step": 410
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00040699999999999997,
+      "loss": 1.5076,
+      "step": 411
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.000408,
+      "loss": 1.5335,
+      "step": 412
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00040899999999999997,
+      "loss": 1.5103,
+      "step": 413
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00041,
+      "loss": 1.5431,
+      "step": 414
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00041099999999999996,
+      "loss": 1.4869,
+      "step": 415
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.000412,
+      "loss": 1.6193,
+      "step": 416
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000413,
+      "loss": 1.5562,
+      "step": 417
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.000414,
+      "loss": 1.5564,
+      "step": 418
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.000415,
+      "loss": 1.472,
+      "step": 419
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.000416,
+      "loss": 1.7672,
+      "step": 420
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.000417,
+      "loss": 1.5295,
+      "step": 421
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00041799999999999997,
+      "loss": 1.6139,
+      "step": 422
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.000419,
+      "loss": 1.6035,
+      "step": 423
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00042,
+      "loss": 1.5985,
+      "step": 424
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.000421,
+      "loss": 1.4869,
+      "step": 425
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.000422,
+      "loss": 1.5062,
+      "step": 426
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.000423,
+      "loss": 1.6841,
+      "step": 427
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.000424,
+      "loss": 1.5998,
+      "step": 428
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.000425,
+      "loss": 1.452,
+      "step": 429
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.000426,
+      "loss": 1.494,
+      "step": 430
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.000427,
+      "loss": 1.4583,
+      "step": 431
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.000428,
+      "loss": 1.3967,
+      "step": 432
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000429,
+      "loss": 1.5856,
+      "step": 433
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00043,
+      "loss": 1.411,
+      "step": 434
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.000431,
+      "loss": 1.3648,
+      "step": 435
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.000432,
+      "loss": 1.431,
+      "step": 436
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.000433,
+      "loss": 1.3989,
+      "step": 437
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00043400000000000003,
+      "loss": 1.3966,
+      "step": 438
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.000435,
+      "loss": 1.3775,
+      "step": 439
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.000436,
+      "loss": 1.284,
+      "step": 440
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.000437,
+      "loss": 1.2616,
+      "step": 441
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.000438,
+      "loss": 1.2357,
+      "step": 442
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.000439,
+      "loss": 1.1726,
+      "step": 443
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.00044,
+      "loss": 1.1024,
+      "step": 444
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.000441,
+      "loss": 1.2494,
+      "step": 445
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.000442,
+      "loss": 1.1615,
+      "step": 446
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.00044300000000000003,
+      "loss": 1.1828,
+      "step": 447
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.000444,
+      "loss": 1.2224,
+      "step": 448
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 1.0664,
+      "step": 449
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.000446,
+      "loss": 1.0797,
+      "step": 450
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.000447,
+      "loss": 1.0265,
+      "step": 451
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.000448,
+      "loss": 1.2902,
+      "step": 452
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.000449,
+      "loss": 1.1238,
+      "step": 453
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 1.1059,
+      "step": 454
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.000451,
+      "loss": 1.3351,
+      "step": 455
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 1.3794,
+      "step": 456
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.000453,
+      "loss": 1.3166,
+      "step": 457
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00045400000000000003,
+      "loss": 1.1832,
+      "step": 458
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.000455,
+      "loss": 1.1953,
+      "step": 459
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.000456,
+      "loss": 1.4968,
+      "step": 460
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00045700000000000005,
+      "loss": 1.2917,
+      "step": 461
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.000458,
+      "loss": 1.1942,
+      "step": 462
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00045900000000000004,
+      "loss": 1.1631,
+      "step": 463
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00046,
+      "loss": 1.2899,
+      "step": 464
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00046100000000000004,
+      "loss": 1.1476,
+      "step": 465
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.000462,
+      "loss": 1.123,
+      "step": 466
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00046300000000000003,
+      "loss": 1.0419,
+      "step": 467
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.00046400000000000006,
+      "loss": 1.0275,
+      "step": 468
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.000465,
+      "loss": 0.9758,
+      "step": 469
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00046600000000000005,
+      "loss": 0.8816,
+      "step": 470
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.000467,
+      "loss": 1.1259,
+      "step": 471
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.00046800000000000005,
+      "loss": 1.2087,
+      "step": 472
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00046899999999999996,
+      "loss": 1.1363,
+      "step": 473
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.00047,
+      "loss": 1.006,
+      "step": 474
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.000471,
+      "loss": 1.1897,
+      "step": 475
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.000472,
+      "loss": 1.1039,
+      "step": 476
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.000473,
+      "loss": 1.0023,
+      "step": 477
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.000474,
+      "loss": 1.1296,
+      "step": 478
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.000475,
+      "loss": 1.1849,
+      "step": 479
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.00047599999999999997,
+      "loss": 1.0822,
+      "step": 480
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.000477,
+      "loss": 1.1014,
+      "step": 481
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.00047799999999999996,
+      "loss": 1.0653,
+      "step": 482
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.000479,
+      "loss": 1.0525,
+      "step": 483
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00048,
+      "loss": 1.0081,
+      "step": 484
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000481,
+      "loss": 1.0895,
+      "step": 485
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.000482,
+      "loss": 1.2735,
+      "step": 486
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.000483,
+      "loss": 1.1423,
+      "step": 487
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.000484,
+      "loss": 1.0427,
+      "step": 488
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 1.2217,
+      "step": 489
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.000486,
+      "loss": 1.0567,
+      "step": 490
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.000487,
+      "loss": 1.0211,
+      "step": 491
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.000488,
+      "loss": 1.0223,
+      "step": 492
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.000489,
+      "loss": 0.9925,
+      "step": 493
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00049,
+      "loss": 0.9798,
+      "step": 494
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.000491,
+      "loss": 1.0249,
+      "step": 495
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.000492,
+      "loss": 1.0168,
+      "step": 496
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.0004930000000000001,
+      "loss": 0.9643,
+      "step": 497
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.000494,
+      "loss": 1.0591,
+      "step": 498
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.000495,
+      "loss": 0.8722,
+      "step": 499
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.000496,
+      "loss": 1.0244,
+      "step": 500
+    },
+    {
+      "epoch": 4.59,
+      "eval_loss": 1.029004693031311,
+      "eval_runtime": 36.6474,
+      "eval_samples_per_second": 44.942,
+      "eval_steps_per_second": 2.811,
+      "eval_wer": 0.3322926673070765,
+      "step": 500
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.000497,
+      "loss": 1.0921,
+      "step": 501
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.000498,
+      "loss": 1.047,
+      "step": 502
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.000499,
+      "loss": 0.9539,
+      "step": 503
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0005,
+      "loss": 1.0536,
+      "step": 504
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.000499702380952381,
+      "loss": 1.1221,
+      "step": 505
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.0004994047619047619,
+      "loss": 1.0429,
+      "step": 506
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0004991071428571428,
+      "loss": 1.1396,
+      "step": 507
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.0004988095238095239,
+      "loss": 1.0315,
+      "step": 508
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.0004985119047619048,
+      "loss": 1.0947,
+      "step": 509
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0004982142857142858,
+      "loss": 1.1754,
+      "step": 510
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.0004979166666666667,
+      "loss": 1.3003,
+      "step": 511
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0004976190476190476,
+      "loss": 1.0628,
+      "step": 512
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0004973214285714286,
+      "loss": 1.1121,
+      "step": 513
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0004970238095238095,
+      "loss": 1.0423,
+      "step": 514
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0004967261904761904,
+      "loss": 1.047,
+      "step": 515
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.0004964285714285715,
+      "loss": 0.9824,
+      "step": 516
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0004961309523809524,
+      "loss": 1.0348,
+      "step": 517
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0004958333333333334,
+      "loss": 1.0037,
+      "step": 518
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.0004955357142857143,
+      "loss": 0.8242,
+      "step": 519
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0004952380952380952,
+      "loss": 0.9696,
+      "step": 520
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0004949404761904762,
+      "loss": 1.008,
+      "step": 521
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0004946428571428571,
+      "loss": 0.9254,
+      "step": 522
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0004943452380952382,
+      "loss": 0.8288,
+      "step": 523
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0004940476190476191,
+      "loss": 0.8741,
+      "step": 524
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00049375,
+      "loss": 1.0546,
+      "step": 525
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.000493452380952381,
+      "loss": 1.0411,
+      "step": 526
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0004931547619047619,
+      "loss": 1.0639,
+      "step": 527
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0004928571428571429,
+      "loss": 0.8977,
+      "step": 528
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0004925595238095238,
+      "loss": 0.9564,
+      "step": 529
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0004922619047619049,
+      "loss": 0.8622,
+      "step": 530
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.0004919642857142857,
+      "loss": 0.9812,
+      "step": 531
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0004916666666666666,
+      "loss": 0.9064,
+      "step": 532
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.0004913690476190476,
+      "loss": 1.0261,
+      "step": 533
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0004910714285714286,
+      "loss": 1.0859,
+      "step": 534
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0004907738095238095,
+      "loss": 1.0929,
+      "step": 535
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.0004904761904761905,
+      "loss": 1.2656,
+      "step": 536
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0004901785714285714,
+      "loss": 0.9044,
+      "step": 537
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0004898809523809524,
+      "loss": 0.992,
+      "step": 538
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0004895833333333333,
+      "loss": 0.905,
+      "step": 539
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0004892857142857142,
+      "loss": 0.9496,
+      "step": 540
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0004889880952380953,
+      "loss": 1.0831,
+      "step": 541
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.0004886904761904762,
+      "loss": 0.9685,
+      "step": 542
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.0004883928571428572,
+      "loss": 1.0129,
+      "step": 543
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.0004880952380952381,
+      "loss": 1.1261,
+      "step": 544
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00048779761904761904,
+      "loss": 1.3113,
+      "step": 545
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0004875,
+      "loss": 0.833,
+      "step": 546
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.00048720238095238097,
+      "loss": 0.7039,
+      "step": 547
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0004869047619047619,
+      "loss": 0.8449,
+      "step": 548
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00048660714285714285,
+      "loss": 0.7254,
+      "step": 549
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0004863095238095238,
+      "loss": 0.6353,
+      "step": 550
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0004860119047619048,
+      "loss": 0.8371,
+      "step": 551
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0004857142857142857,
+      "loss": 0.5982,
+      "step": 552
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00048541666666666666,
+      "loss": 0.7029,
+      "step": 553
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00048511904761904765,
+      "loss": 0.6792,
+      "step": 554
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0004848214285714286,
+      "loss": 0.7449,
+      "step": 555
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.00048452380952380953,
+      "loss": 0.6403,
+      "step": 556
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00048422619047619047,
+      "loss": 0.7698,
+      "step": 557
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.00048392857142857146,
+      "loss": 0.7152,
+      "step": 558
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.0004836309523809524,
+      "loss": 0.7124,
+      "step": 559
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00048333333333333334,
+      "loss": 0.7273,
+      "step": 560
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.0004830357142857143,
+      "loss": 0.5464,
+      "step": 561
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00048273809523809527,
+      "loss": 0.7211,
+      "step": 562
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.0004824404761904762,
+      "loss": 0.7491,
+      "step": 563
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00048214285714285715,
+      "loss": 0.7464,
+      "step": 564
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00048184523809523814,
+      "loss": 0.7698,
+      "step": 565
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0004815476190476191,
+      "loss": 0.7327,
+      "step": 566
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00048125,
+      "loss": 0.6873,
+      "step": 567
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.00048095238095238095,
+      "loss": 0.8524,
+      "step": 568
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00048065476190476195,
+      "loss": 1.0414,
+      "step": 569
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0004803571428571429,
+      "loss": 0.9219,
+      "step": 570
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0004800595238095238,
+      "loss": 0.6374,
+      "step": 571
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00047976190476190476,
+      "loss": 0.6421,
+      "step": 572
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.00047946428571428576,
+      "loss": 0.6721,
+      "step": 573
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0004791666666666667,
+      "loss": 0.677,
+      "step": 574
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00047886904761904763,
+      "loss": 0.6869,
+      "step": 575
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.0004785714285714286,
+      "loss": 0.6019,
+      "step": 576
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00047827380952380957,
+      "loss": 0.6728,
+      "step": 577
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0004779761904761905,
+      "loss": 0.6036,
+      "step": 578
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00047767857142857144,
+      "loss": 0.6075,
+      "step": 579
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00047738095238095244,
+      "loss": 0.7073,
+      "step": 580
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.0004770833333333334,
+      "loss": 0.5775,
+      "step": 581
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0004767857142857143,
+      "loss": 0.606,
+      "step": 582
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0004764880952380952,
+      "loss": 0.6927,
+      "step": 583
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.0004761904761904762,
+      "loss": 0.6912,
+      "step": 584
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00047589285714285713,
+      "loss": 0.7068,
+      "step": 585
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00047559523809523807,
+      "loss": 0.7384,
+      "step": 586
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00047529761904761906,
+      "loss": 0.9084,
+      "step": 587
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.000475,
+      "loss": 0.7458,
+      "step": 588
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00047470238095238094,
+      "loss": 0.6443,
+      "step": 589
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.0004744047619047619,
+      "loss": 0.7792,
+      "step": 590
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00047410714285714287,
+      "loss": 0.7438,
+      "step": 591
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0004738095238095238,
+      "loss": 0.7982,
+      "step": 592
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.00047351190476190475,
+      "loss": 0.7389,
+      "step": 593
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.0004732142857142857,
+      "loss": 0.7046,
+      "step": 594
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0004729166666666667,
+      "loss": 0.7215,
+      "step": 595
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0004726190476190476,
+      "loss": 0.7232,
+      "step": 596
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00047232142857142856,
+      "loss": 0.7941,
+      "step": 597
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.00047202380952380955,
+      "loss": 0.7464,
+      "step": 598
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0004717261904761905,
+      "loss": 0.7256,
+      "step": 599
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0004714285714285714,
+      "loss": 0.7095,
+      "step": 600
+    },
+    {
+      "epoch": 5.5,
+      "eval_loss": 0.8424153327941895,
+      "eval_runtime": 36.6389,
+      "eval_samples_per_second": 44.952,
+      "eval_steps_per_second": 2.811,
+      "eval_wer": 0.2859029779058597,
+      "step": 600
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00047113095238095236,
+      "loss": 0.6716,
+      "step": 601
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00047083333333333336,
+      "loss": 0.7452,
+      "step": 602
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0004705357142857143,
+      "loss": 0.6411,
+      "step": 603
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00047023809523809523,
+      "loss": 0.6482,
+      "step": 604
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0004699404761904762,
+      "loss": 0.7119,
+      "step": 605
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00046964285714285717,
+      "loss": 0.6345,
+      "step": 606
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0004693452380952381,
+      "loss": 0.6356,
+      "step": 607
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.00046904761904761904,
+      "loss": 0.6445,
+      "step": 608
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00046875,
+      "loss": 0.7391,
+      "step": 609
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.000468452380952381,
+      "loss": 0.7151,
+      "step": 610
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0004681547619047619,
+      "loss": 0.6977,
+      "step": 611
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00046785714285714285,
+      "loss": 0.7554,
+      "step": 612
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00046755952380952385,
+      "loss": 0.6745,
+      "step": 613
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0004672619047619048,
+      "loss": 0.6777,
+      "step": 614
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.0004669642857142857,
+      "loss": 0.7894,
+      "step": 615
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00046666666666666666,
+      "loss": 0.6759,
+      "step": 616
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00046636904761904765,
+      "loss": 0.626,
+      "step": 617
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0004660714285714286,
+      "loss": 0.7807,
+      "step": 618
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00046577380952380953,
+      "loss": 0.8262,
+      "step": 619
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00046547619047619047,
+      "loss": 0.7334,
+      "step": 620
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.00046517857142857146,
+      "loss": 0.7908,
+      "step": 621
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0004648809523809524,
+      "loss": 0.7076,
+      "step": 622
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00046458333333333334,
+      "loss": 0.6868,
+      "step": 623
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00046428571428571433,
+      "loss": 0.7658,
+      "step": 624
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00046398809523809527,
+      "loss": 0.6708,
+      "step": 625
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.0004636904761904762,
+      "loss": 0.6814,
+      "step": 626
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.00046339285714285715,
+      "loss": 0.6423,
+      "step": 627
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00046309523809523814,
+      "loss": 0.6029,
+      "step": 628
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0004627976190476191,
+      "loss": 0.7093,
+      "step": 629
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.0004625,
+      "loss": 0.6608,
+      "step": 630
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00046220238095238096,
+      "loss": 0.6257,
+      "step": 631
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00046190476190476195,
+      "loss": 0.5837,
+      "step": 632
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.0004616071428571429,
+      "loss": 0.5281,
+      "step": 633
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00046130952380952383,
+      "loss": 0.5757,
+      "step": 634
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0004610119047619048,
+      "loss": 0.652,
+      "step": 635
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0004607142857142857,
+      "loss": 0.8051,
+      "step": 636
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00046041666666666665,
+      "loss": 0.5603,
+      "step": 637
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.0004601190476190476,
+      "loss": 0.6722,
+      "step": 638
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0004598214285714286,
+      "loss": 0.6133,
+      "step": 639
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0004595238095238095,
+      "loss": 0.6847,
+      "step": 640
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00045922619047619045,
+      "loss": 0.6421,
+      "step": 641
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0004589285714285714,
+      "loss": 0.7918,
+      "step": 642
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0004586309523809524,
+      "loss": 0.6256,
+      "step": 643
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0004583333333333333,
+      "loss": 0.6713,
+      "step": 644
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00045803571428571426,
+      "loss": 0.9041,
+      "step": 645
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00045773809523809526,
+      "loss": 0.6144,
+      "step": 646
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0004574404761904762,
+      "loss": 0.659,
+      "step": 647
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00045714285714285713,
+      "loss": 0.6476,
+      "step": 648
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00045684523809523807,
+      "loss": 0.7583,
+      "step": 649
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00045654761904761907,
+      "loss": 0.6461,
+      "step": 650
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00045625,
+      "loss": 0.7822,
+      "step": 651
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00045595238095238094,
+      "loss": 0.7319,
+      "step": 652
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0004556547619047619,
+      "loss": 0.6277,
+      "step": 653
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0004553571428571429,
+      "loss": 0.7953,
+      "step": 654
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0004550595238095238,
+      "loss": 0.6052,
+      "step": 655
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00045476190476190475,
+      "loss": 0.5027,
+      "step": 656
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00045446428571428574,
+      "loss": 0.5211,
+      "step": 657
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0004541666666666667,
+      "loss": 0.5342,
+      "step": 658
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0004538690476190476,
+      "loss": 0.4475,
+      "step": 659
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00045357142857142856,
+      "loss": 0.5194,
+      "step": 660
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00045327380952380955,
+      "loss": 0.5385,
+      "step": 661
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0004529761904761905,
+      "loss": 0.4412,
+      "step": 662
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00045267857142857143,
+      "loss": 0.4688,
+      "step": 663
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00045238095238095237,
+      "loss": 0.5524,
+      "step": 664
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00045208333333333336,
+      "loss": 0.4355,
+      "step": 665
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.0004517857142857143,
+      "loss": 0.4869,
+      "step": 666
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00045148809523809524,
+      "loss": 0.4428,
+      "step": 667
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00045119047619047623,
+      "loss": 0.4751,
+      "step": 668
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00045089285714285717,
+      "loss": 0.4747,
+      "step": 669
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0004505952380952381,
+      "loss": 0.4015,
+      "step": 670
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00045029761904761905,
+      "loss": 0.678,
+      "step": 671
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 0.5198,
+      "step": 672
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.000449702380952381,
+      "loss": 0.4171,
+      "step": 673
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.0004494047619047619,
+      "loss": 0.5392,
+      "step": 674
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00044910714285714286,
+      "loss": 0.4987,
+      "step": 675
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.00044880952380952385,
+      "loss": 0.504,
+      "step": 676
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0004485119047619048,
+      "loss": 0.5521,
+      "step": 677
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00044821428571428573,
+      "loss": 0.5211,
+      "step": 678
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.0004479166666666667,
+      "loss": 0.6712,
+      "step": 679
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00044761904761904766,
+      "loss": 0.426,
+      "step": 680
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0004473214285714286,
+      "loss": 0.5331,
+      "step": 681
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00044702380952380954,
+      "loss": 0.5609,
+      "step": 682
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.00044672619047619053,
+      "loss": 0.5555,
+      "step": 683
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00044642857142857147,
+      "loss": 0.4505,
+      "step": 684
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0004461309523809524,
+      "loss": 0.4444,
+      "step": 685
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00044583333333333335,
+      "loss": 0.3861,
+      "step": 686
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00044553571428571434,
+      "loss": 0.4424,
+      "step": 687
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0004452380952380952,
+      "loss": 0.4101,
+      "step": 688
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00044494047619047616,
+      "loss": 0.4728,
+      "step": 689
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.0004446428571428571,
+      "loss": 0.4905,
+      "step": 690
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.0004443452380952381,
+      "loss": 0.5642,
+      "step": 691
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00044404761904761903,
+      "loss": 0.4848,
+      "step": 692
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00044374999999999997,
+      "loss": 0.5765,
+      "step": 693
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00044345238095238096,
+      "loss": 0.4728,
+      "step": 694
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.0004431547619047619,
+      "loss": 0.6184,
+      "step": 695
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00044285714285714284,
+      "loss": 0.5025,
+      "step": 696
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0004425595238095238,
+      "loss": 0.5541,
+      "step": 697
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.0004422619047619048,
+      "loss": 0.4937,
+      "step": 698
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0004419642857142857,
+      "loss": 0.6063,
+      "step": 699
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00044166666666666665,
+      "loss": 0.619,
+      "step": 700
+    },
+    {
+      "epoch": 6.42,
+      "eval_loss": 0.7389244437217712,
+      "eval_runtime": 36.5706,
+      "eval_samples_per_second": 45.036,
+      "eval_steps_per_second": 2.816,
+      "eval_wer": 0.22316282420749278,
+      "step": 700
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0004413690476190476,
+      "loss": 0.5363,
+      "step": 701
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.0004410714285714286,
+      "loss": 0.472,
+      "step": 702
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0004407738095238095,
+      "loss": 0.5292,
+      "step": 703
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.00044047619047619046,
+      "loss": 0.6937,
+      "step": 704
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00044017857142857145,
+      "loss": 0.466,
+      "step": 705
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0004398809523809524,
+      "loss": 0.58,
+      "step": 706
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00043958333333333333,
+      "loss": 0.3872,
+      "step": 707
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00043928571428571427,
+      "loss": 0.5127,
+      "step": 708
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00043898809523809526,
+      "loss": 0.4371,
+      "step": 709
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0004386904761904762,
+      "loss": 0.3844,
+      "step": 710
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00043839285714285714,
+      "loss": 0.6007,
+      "step": 711
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0004380952380952381,
+      "loss": 0.4901,
+      "step": 712
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00043779761904761907,
+      "loss": 0.4795,
+      "step": 713
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.0004375,
+      "loss": 0.6768,
+      "step": 714
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.00043720238095238095,
+      "loss": 0.4323,
+      "step": 715
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00043690476190476194,
+      "loss": 0.4503,
+      "step": 716
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0004366071428571429,
+      "loss": 0.4226,
+      "step": 717
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.0004363095238095238,
+      "loss": 0.4997,
+      "step": 718
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.00043601190476190476,
+      "loss": 0.4085,
+      "step": 719
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00043571428571428575,
+      "loss": 0.4442,
+      "step": 720
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0004354166666666667,
+      "loss": 0.5479,
+      "step": 721
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0004351190476190476,
+      "loss": 0.423,
+      "step": 722
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.00043482142857142857,
+      "loss": 0.5197,
+      "step": 723
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00043452380952380956,
+      "loss": 0.4856,
+      "step": 724
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0004342261904761905,
+      "loss": 0.3755,
+      "step": 725
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00043392857142857144,
+      "loss": 0.7071,
+      "step": 726
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00043363095238095243,
+      "loss": 0.6221,
+      "step": 727
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00043333333333333337,
+      "loss": 0.5698,
+      "step": 728
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0004330357142857143,
+      "loss": 0.5208,
+      "step": 729
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00043273809523809525,
+      "loss": 0.5065,
+      "step": 730
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00043244047619047624,
+      "loss": 0.4483,
+      "step": 731
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0004321428571428572,
+      "loss": 0.5604,
+      "step": 732
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0004318452380952381,
+      "loss": 0.5028,
+      "step": 733
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00043154761904761905,
+      "loss": 0.4628,
+      "step": 734
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00043125000000000005,
+      "loss": 0.4757,
+      "step": 735
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.000430952380952381,
+      "loss": 0.4778,
+      "step": 736
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.0004306547619047619,
+      "loss": 0.3465,
+      "step": 737
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0004303571428571429,
+      "loss": 0.3989,
+      "step": 738
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00043005952380952386,
+      "loss": 0.4782,
+      "step": 739
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0004297619047619048,
+      "loss": 0.4147,
+      "step": 740
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0004294642857142857,
+      "loss": 0.4747,
+      "step": 741
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00042916666666666667,
+      "loss": 0.4796,
+      "step": 742
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.0004288690476190476,
+      "loss": 0.4597,
+      "step": 743
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.00042857142857142855,
+      "loss": 0.5301,
+      "step": 744
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0004282738095238095,
+      "loss": 0.5437,
+      "step": 745
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0004279761904761905,
+      "loss": 0.4825,
+      "step": 746
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.0004276785714285714,
+      "loss": 0.4869,
+      "step": 747
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00042738095238095236,
+      "loss": 0.5642,
+      "step": 748
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0004270833333333333,
+      "loss": 0.4817,
+      "step": 749
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0004267857142857143,
+      "loss": 0.5704,
+      "step": 750
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.00042648809523809523,
+      "loss": 0.5132,
+      "step": 751
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00042619047619047617,
+      "loss": 0.5341,
+      "step": 752
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00042589285714285716,
+      "loss": 0.6579,
+      "step": 753
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0004255952380952381,
+      "loss": 0.6194,
+      "step": 754
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00042529761904761904,
+      "loss": 0.499,
+      "step": 755
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.000425,
+      "loss": 0.4829,
+      "step": 756
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00042470238095238097,
+      "loss": 0.482,
+      "step": 757
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.0004244047619047619,
+      "loss": 0.512,
+      "step": 758
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00042410714285714285,
+      "loss": 0.4165,
+      "step": 759
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.0004238095238095238,
+      "loss": 0.6565,
+      "step": 760
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0004235119047619048,
+      "loss": 0.5331,
+      "step": 761
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0004232142857142857,
+      "loss": 0.4662,
+      "step": 762
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00042291666666666666,
+      "loss": 0.6726,
+      "step": 763
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00042261904761904765,
+      "loss": 0.4184,
+      "step": 764
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.0004223214285714286,
+      "loss": 0.3985,
+      "step": 765
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0004220238095238095,
+      "loss": 0.4009,
+      "step": 766
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00042172619047619046,
+      "loss": 0.4845,
+      "step": 767
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00042142857142857146,
+      "loss": 0.2801,
+      "step": 768
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0004211309523809524,
+      "loss": 0.3393,
+      "step": 769
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.00042083333333333333,
+      "loss": 0.306,
+      "step": 770
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0004205357142857143,
+      "loss": 0.4196,
+      "step": 771
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.00042023809523809527,
+      "loss": 0.3151,
+      "step": 772
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0004199404761904762,
+      "loss": 0.367,
+      "step": 773
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00041964285714285714,
+      "loss": 0.3447,
+      "step": 774
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00041934523809523814,
+      "loss": 0.4018,
+      "step": 775
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0004190476190476191,
+      "loss": 0.3205,
+      "step": 776
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00041875,
+      "loss": 0.3915,
+      "step": 777
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00041845238095238095,
+      "loss": 0.3067,
+      "step": 778
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.00041815476190476195,
+      "loss": 0.3742,
+      "step": 779
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.0004178571428571429,
+      "loss": 0.4019,
+      "step": 780
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0004175595238095238,
+      "loss": 0.3765,
+      "step": 781
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00041726190476190476,
+      "loss": 0.3259,
+      "step": 782
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00041696428571428575,
+      "loss": 0.4031,
+      "step": 783
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.0004166666666666667,
+      "loss": 0.3407,
+      "step": 784
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.00041636904761904763,
+      "loss": 0.4152,
+      "step": 785
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.0004160714285714286,
+      "loss": 0.4165,
+      "step": 786
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00041577380952380956,
+      "loss": 0.4546,
+      "step": 787
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0004154761904761905,
+      "loss": 0.4783,
+      "step": 788
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00041517857142857144,
+      "loss": 0.3194,
+      "step": 789
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.00041488095238095243,
+      "loss": 0.3305,
+      "step": 790
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00041458333333333337,
+      "loss": 0.3199,
+      "step": 791
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0004142857142857143,
+      "loss": 0.3623,
+      "step": 792
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0004139880952380952,
+      "loss": 0.3482,
+      "step": 793
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0004136904761904762,
+      "loss": 0.3427,
+      "step": 794
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.00041339285714285713,
+      "loss": 0.3405,
+      "step": 795
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00041309523809523807,
+      "loss": 0.3817,
+      "step": 796
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00041279761904761906,
+      "loss": 0.275,
+      "step": 797
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.0004125,
+      "loss": 0.2878,
+      "step": 798
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00041220238095238094,
+      "loss": 0.3708,
+      "step": 799
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.0004119047619047619,
+      "loss": 0.3541,
+      "step": 800
+    },
+    {
+      "epoch": 7.34,
+      "eval_loss": 0.7048654556274414,
+      "eval_runtime": 36.8141,
+      "eval_samples_per_second": 44.738,
+      "eval_steps_per_second": 2.798,
+      "eval_wer": 0.20431075888568684,
+      "step": 800
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00041160714285714287,
+      "loss": 0.3143,
+      "step": 801
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0004113095238095238,
+      "loss": 0.3602,
+      "step": 802
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00041101190476190475,
+      "loss": 0.3614,
+      "step": 803
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0004107142857142857,
+      "loss": 0.4119,
+      "step": 804
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0004104166666666667,
+      "loss": 0.3522,
+      "step": 805
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0004101190476190476,
+      "loss": 0.4333,
+      "step": 806
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00040982142857142855,
+      "loss": 0.4502,
+      "step": 807
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00040952380952380955,
+      "loss": 0.4419,
+      "step": 808
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.0004092261904761905,
+      "loss": 0.4096,
+      "step": 809
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0004089285714285714,
+      "loss": 0.361,
+      "step": 810
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00040863095238095236,
+      "loss": 0.3831,
+      "step": 811
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00040833333333333336,
+      "loss": 0.343,
+      "step": 812
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0004080357142857143,
+      "loss": 0.6617,
+      "step": 813
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00040773809523809523,
+      "loss": 0.4464,
+      "step": 814
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00040744047619047617,
+      "loss": 0.3023,
+      "step": 815
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00040714285714285717,
+      "loss": 0.3661,
+      "step": 816
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0004068452380952381,
+      "loss": 0.3297,
+      "step": 817
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00040654761904761904,
+      "loss": 0.3128,
+      "step": 818
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00040625000000000004,
+      "loss": 0.3508,
+      "step": 819
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.000405952380952381,
+      "loss": 0.4379,
+      "step": 820
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.0004056547619047619,
+      "loss": 0.3658,
+      "step": 821
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00040535714285714285,
+      "loss": 0.3452,
+      "step": 822
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00040505952380952384,
+      "loss": 0.2632,
+      "step": 823
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0004047619047619048,
+      "loss": 0.329,
+      "step": 824
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.0004044642857142857,
+      "loss": 0.3464,
+      "step": 825
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.00040416666666666666,
+      "loss": 0.331,
+      "step": 826
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00040386904761904765,
+      "loss": 0.4117,
+      "step": 827
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.0004035714285714286,
+      "loss": 0.4321,
+      "step": 828
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00040327380952380953,
+      "loss": 0.4561,
+      "step": 829
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00040297619047619047,
+      "loss": 0.5066,
+      "step": 830
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.00040267857142857146,
+      "loss": 0.422,
+      "step": 831
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0004023809523809524,
+      "loss": 0.4191,
+      "step": 832
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00040208333333333334,
+      "loss": 0.3214,
+      "step": 833
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00040178571428571433,
+      "loss": 0.4118,
+      "step": 834
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00040148809523809527,
+      "loss": 0.4488,
+      "step": 835
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.0004011904761904762,
+      "loss": 0.3547,
+      "step": 836
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00040089285714285715,
+      "loss": 0.4759,
+      "step": 837
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00040059523809523814,
+      "loss": 0.5,
+      "step": 838
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.0004002976190476191,
+      "loss": 0.3379,
+      "step": 839
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0004,
+      "loss": 0.3661,
+      "step": 840
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00039970238095238096,
+      "loss": 0.3711,
+      "step": 841
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00039940476190476195,
+      "loss": 0.3131,
+      "step": 842
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.0003991071428571429,
+      "loss": 0.3414,
+      "step": 843
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00039880952380952383,
+      "loss": 0.3539,
+      "step": 844
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.0003985119047619048,
+      "loss": 0.3457,
+      "step": 845
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0003982142857142857,
+      "loss": 0.299,
+      "step": 846
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00039791666666666664,
+      "loss": 0.3792,
+      "step": 847
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0003976190476190476,
+      "loss": 0.4334,
+      "step": 848
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0003973214285714286,
+      "loss": 0.3011,
+      "step": 849
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.0003970238095238095,
+      "loss": 0.4186,
+      "step": 850
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00039672619047619045,
+      "loss": 0.4048,
+      "step": 851
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0003964285714285714,
+      "loss": 0.4028,
+      "step": 852
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.0003961309523809524,
+      "loss": 0.3563,
+      "step": 853
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.0003958333333333333,
+      "loss": 0.4591,
+      "step": 854
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00039553571428571426,
+      "loss": 0.3766,
+      "step": 855
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.00039523809523809526,
+      "loss": 0.4764,
+      "step": 856
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0003949404761904762,
+      "loss": 0.3666,
+      "step": 857
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00039464285714285713,
+      "loss": 0.4205,
+      "step": 858
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00039434523809523807,
+      "loss": 0.4185,
+      "step": 859
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00039404761904761906,
+      "loss": 0.4093,
+      "step": 860
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00039375,
+      "loss": 0.3376,
+      "step": 861
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.00039345238095238094,
+      "loss": 0.4816,
+      "step": 862
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0003931547619047619,
+      "loss": 0.5414,
+      "step": 863
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0003928571428571429,
+      "loss": 0.4455,
+      "step": 864
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0003925595238095238,
+      "loss": 0.4133,
+      "step": 865
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.00039226190476190475,
+      "loss": 0.3418,
+      "step": 866
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00039196428571428574,
+      "loss": 0.3913,
+      "step": 867
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.0003916666666666667,
+      "loss": 0.3829,
+      "step": 868
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0003913690476190476,
+      "loss": 0.2972,
+      "step": 869
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00039107142857142856,
+      "loss": 0.4337,
+      "step": 870
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.00039077380952380955,
+      "loss": 0.4927,
+      "step": 871
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0003904761904761905,
+      "loss": 0.5129,
+      "step": 872
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.00039017857142857143,
+      "loss": 0.2227,
+      "step": 873
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00038988095238095237,
+      "loss": 0.3338,
+      "step": 874
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00038958333333333336,
+      "loss": 0.2917,
+      "step": 875
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0003892857142857143,
+      "loss": 0.2678,
+      "step": 876
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00038898809523809524,
+      "loss": 0.2627,
+      "step": 877
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00038869047619047623,
+      "loss": 0.3013,
+      "step": 878
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00038839285714285717,
+      "loss": 0.2443,
+      "step": 879
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0003880952380952381,
+      "loss": 0.3228,
+      "step": 880
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00038779761904761905,
+      "loss": 0.2655,
+      "step": 881
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00038750000000000004,
+      "loss": 0.2557,
+      "step": 882
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.000387202380952381,
+      "loss": 0.3123,
+      "step": 883
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.0003869047619047619,
+      "loss": 0.2731,
+      "step": 884
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00038660714285714286,
+      "loss": 0.2922,
+      "step": 885
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00038630952380952385,
+      "loss": 0.3141,
+      "step": 886
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0003860119047619048,
+      "loss": 0.2942,
+      "step": 887
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0003857142857142857,
+      "loss": 0.2612,
+      "step": 888
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.0003854166666666667,
+      "loss": 0.3409,
+      "step": 889
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.00038511904761904766,
+      "loss": 0.3169,
+      "step": 890
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.0003848214285714286,
+      "loss": 0.2686,
+      "step": 891
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00038452380952380954,
+      "loss": 0.3858,
+      "step": 892
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00038422619047619053,
+      "loss": 0.4038,
+      "step": 893
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00038392857142857147,
+      "loss": 0.3448,
+      "step": 894
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.0003836309523809524,
+      "loss": 0.2985,
+      "step": 895
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00038333333333333334,
+      "loss": 0.3257,
+      "step": 896
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00038303571428571434,
+      "loss": 0.3967,
+      "step": 897
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.0003827380952380952,
+      "loss": 0.3701,
+      "step": 898
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00038244047619047616,
+      "loss": 0.2376,
+      "step": 899
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0003821428571428571,
+      "loss": 0.2946,
+      "step": 900
+    },
+    {
+      "epoch": 8.26,
+      "eval_loss": 0.7064677476882935,
+      "eval_runtime": 37.1375,
+      "eval_samples_per_second": 44.349,
+      "eval_steps_per_second": 2.773,
+      "eval_wer": 0.21533781620236953,
+      "step": 900
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.0003818452380952381,
+      "loss": 0.2596,
+      "step": 901
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00038154761904761903,
+      "loss": 0.2161,
+      "step": 902
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00038124999999999997,
+      "loss": 0.2869,
+      "step": 903
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00038095238095238096,
+      "loss": 0.3107,
+      "step": 904
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0003806547619047619,
+      "loss": 0.2823,
+      "step": 905
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.00038035714285714284,
+      "loss": 0.2716,
+      "step": 906
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.0003800595238095238,
+      "loss": 0.2612,
+      "step": 907
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00037976190476190477,
+      "loss": 0.2548,
+      "step": 908
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0003794642857142857,
+      "loss": 0.2585,
+      "step": 909
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00037916666666666665,
+      "loss": 0.2938,
+      "step": 910
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.0003788690476190476,
+      "loss": 0.2688,
+      "step": 911
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.0003785714285714286,
+      "loss": 0.2881,
+      "step": 912
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.0003782738095238095,
+      "loss": 0.2676,
+      "step": 913
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00037797619047619046,
+      "loss": 0.2446,
+      "step": 914
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.00037767857142857145,
+      "loss": 0.2669,
+      "step": 915
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.0003773809523809524,
+      "loss": 0.2959,
+      "step": 916
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00037708333333333333,
+      "loss": 0.2877,
+      "step": 917
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00037678571428571427,
+      "loss": 0.4083,
+      "step": 918
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00037648809523809526,
+      "loss": 0.2427,
+      "step": 919
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.0003761904761904762,
+      "loss": 0.3435,
+      "step": 920
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00037589285714285714,
+      "loss": 0.3214,
+      "step": 921
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0003755952380952381,
+      "loss": 0.3621,
+      "step": 922
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00037529761904761907,
+      "loss": 0.275,
+      "step": 923
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.000375,
+      "loss": 0.2609,
+      "step": 924
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00037470238095238095,
+      "loss": 0.3094,
+      "step": 925
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.00037440476190476194,
+      "loss": 0.2402,
+      "step": 926
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0003741071428571429,
+      "loss": 0.2685,
+      "step": 927
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0003738095238095238,
+      "loss": 0.284,
+      "step": 928
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00037351190476190476,
+      "loss": 0.2738,
+      "step": 929
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00037321428571428575,
+      "loss": 0.2941,
+      "step": 930
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.0003729166666666667,
+      "loss": 0.3129,
+      "step": 931
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0003726190476190476,
+      "loss": 0.2564,
+      "step": 932
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00037232142857142856,
+      "loss": 0.2302,
+      "step": 933
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00037202380952380956,
+      "loss": 0.2942,
+      "step": 934
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0003717261904761905,
+      "loss": 0.2854,
+      "step": 935
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00037142857142857143,
+      "loss": 0.3366,
+      "step": 936
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00037113095238095243,
+      "loss": 0.2416,
+      "step": 937
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00037083333333333337,
+      "loss": 0.3815,
+      "step": 938
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0003705357142857143,
+      "loss": 0.2273,
+      "step": 939
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00037023809523809524,
+      "loss": 0.258,
+      "step": 940
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00036994047619047624,
+      "loss": 0.3317,
+      "step": 941
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0003696428571428572,
+      "loss": 0.2857,
+      "step": 942
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.0003693452380952381,
+      "loss": 0.2989,
+      "step": 943
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00036904761904761905,
+      "loss": 0.3646,
+      "step": 944
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00036875000000000005,
+      "loss": 0.4941,
+      "step": 945
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.000368452380952381,
+      "loss": 0.369,
+      "step": 946
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0003681547619047619,
+      "loss": 0.4309,
+      "step": 947
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.0003678571428571429,
+      "loss": 0.2453,
+      "step": 948
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00036755952380952385,
+      "loss": 0.3439,
+      "step": 949
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0003672619047619048,
+      "loss": 0.223,
+      "step": 950
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0003669642857142857,
+      "loss": 0.3599,
+      "step": 951
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00036666666666666667,
+      "loss": 0.2941,
+      "step": 952
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0003663690476190476,
+      "loss": 0.2427,
+      "step": 953
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00036607142857142855,
+      "loss": 0.32,
+      "step": 954
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.0003657738095238095,
+      "loss": 0.2487,
+      "step": 955
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0003654761904761905,
+      "loss": 0.2606,
+      "step": 956
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.0003651785714285714,
+      "loss": 0.2732,
+      "step": 957
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.00036488095238095236,
+      "loss": 0.2453,
+      "step": 958
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0003645833333333333,
+      "loss": 0.2618,
+      "step": 959
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0003642857142857143,
+      "loss": 0.3253,
+      "step": 960
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00036398809523809523,
+      "loss": 0.3533,
+      "step": 961
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00036369047619047617,
+      "loss": 0.2846,
+      "step": 962
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00036339285714285716,
+      "loss": 0.2442,
+      "step": 963
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0003630952380952381,
+      "loss": 0.2696,
+      "step": 964
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00036279761904761904,
+      "loss": 0.3175,
+      "step": 965
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0003625,
+      "loss": 0.3171,
+      "step": 966
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.00036220238095238097,
+      "loss": 0.3092,
+      "step": 967
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0003619047619047619,
+      "loss": 0.3577,
+      "step": 968
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00036160714285714285,
+      "loss": 0.372,
+      "step": 969
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0003613095238095238,
+      "loss": 0.3897,
+      "step": 970
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.0003610119047619048,
+      "loss": 0.4045,
+      "step": 971
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0003607142857142857,
+      "loss": 0.5022,
+      "step": 972
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00036041666666666665,
+      "loss": 0.257,
+      "step": 973
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00036011904761904765,
+      "loss": 0.2919,
+      "step": 974
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0003598214285714286,
+      "loss": 0.3137,
+      "step": 975
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0003595238095238095,
+      "loss": 0.2893,
+      "step": 976
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.00035922619047619046,
+      "loss": 0.2547,
+      "step": 977
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00035892857142857146,
+      "loss": 0.3911,
+      "step": 978
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0003586309523809524,
+      "loss": 0.3134,
+      "step": 979
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00035833333333333333,
+      "loss": 0.311,
+      "step": 980
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00035803571428571427,
+      "loss": 0.4336,
+      "step": 981
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00035773809523809527,
+      "loss": 0.1838,
+      "step": 982
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.0003574404761904762,
+      "loss": 0.2559,
+      "step": 983
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00035714285714285714,
+      "loss": 0.2247,
+      "step": 984
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00035684523809523814,
+      "loss": 0.2119,
+      "step": 985
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0003565476190476191,
+      "loss": 0.2202,
+      "step": 986
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00035625,
+      "loss": 0.2031,
+      "step": 987
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00035595238095238095,
+      "loss": 0.19,
+      "step": 988
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00035565476190476194,
+      "loss": 0.2541,
+      "step": 989
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.0003553571428571429,
+      "loss": 0.2376,
+      "step": 990
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0003550595238095238,
+      "loss": 0.2968,
+      "step": 991
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00035476190476190476,
+      "loss": 0.2985,
+      "step": 992
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00035446428571428575,
+      "loss": 0.1839,
+      "step": 993
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.0003541666666666667,
+      "loss": 0.2029,
+      "step": 994
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00035386904761904763,
+      "loss": 0.308,
+      "step": 995
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.0003535714285714286,
+      "loss": 0.216,
+      "step": 996
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00035327380952380956,
+      "loss": 0.3188,
+      "step": 997
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0003529761904761905,
+      "loss": 0.2276,
+      "step": 998
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00035267857142857144,
+      "loss": 0.2359,
+      "step": 999
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00035238095238095243,
+      "loss": 0.2868,
+      "step": 1000
+    },
+    {
+      "epoch": 9.17,
+      "eval_loss": 0.6840150356292725,
+      "eval_runtime": 37.0156,
+      "eval_samples_per_second": 44.495,
+      "eval_steps_per_second": 2.783,
+      "eval_wer": 0.21149535702849823,
+      "step": 1000
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00035208333333333337,
+      "loss": 0.2791,
+      "step": 1001
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0003517857142857143,
+      "loss": 0.2604,
+      "step": 1002
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.0003514880952380952,
+      "loss": 0.2241,
+      "step": 1003
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.0003511904761904762,
+      "loss": 0.2533,
+      "step": 1004
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.0003508928571428571,
+      "loss": 0.2323,
+      "step": 1005
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00035059523809523806,
+      "loss": 0.3852,
+      "step": 1006
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00035029761904761906,
+      "loss": 0.1816,
+      "step": 1007
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00035,
+      "loss": 0.2424,
+      "step": 1008
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00034970238095238094,
+      "loss": 0.2433,
+      "step": 1009
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.0003494047619047619,
+      "loss": 0.2205,
+      "step": 1010
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00034910714285714287,
+      "loss": 0.2195,
+      "step": 1011
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0003488095238095238,
+      "loss": 0.2785,
+      "step": 1012
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.00034851190476190474,
+      "loss": 0.2081,
+      "step": 1013
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.0003482142857142857,
+      "loss": 0.1919,
+      "step": 1014
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0003479166666666667,
+      "loss": 0.1916,
+      "step": 1015
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0003476190476190476,
+      "loss": 0.2608,
+      "step": 1016
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.00034732142857142855,
+      "loss": 0.2061,
+      "step": 1017
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00034702380952380955,
+      "loss": 0.231,
+      "step": 1018
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0003467261904761905,
+      "loss": 0.2966,
+      "step": 1019
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0003464285714285714,
+      "loss": 0.2182,
+      "step": 1020
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00034613095238095236,
+      "loss": 0.2253,
+      "step": 1021
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00034583333333333335,
+      "loss": 0.2194,
+      "step": 1022
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0003455357142857143,
+      "loss": 0.2211,
+      "step": 1023
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00034523809523809523,
+      "loss": 0.2318,
+      "step": 1024
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00034494047619047617,
+      "loss": 0.2568,
+      "step": 1025
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.00034464285714285716,
+      "loss": 0.2787,
+      "step": 1026
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0003443452380952381,
+      "loss": 0.3128,
+      "step": 1027
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00034404761904761904,
+      "loss": 0.3909,
+      "step": 1028
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00034375,
+      "loss": 0.3002,
+      "step": 1029
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.000343452380952381,
+      "loss": 0.3773,
+      "step": 1030
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.0003431547619047619,
+      "loss": 0.2966,
+      "step": 1031
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00034285714285714285,
+      "loss": 0.2592,
+      "step": 1032
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00034255952380952384,
+      "loss": 0.241,
+      "step": 1033
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0003422619047619048,
+      "loss": 0.2041,
+      "step": 1034
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.0003419642857142857,
+      "loss": 0.28,
+      "step": 1035
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00034166666666666666,
+      "loss": 0.2085,
+      "step": 1036
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00034136904761904765,
+      "loss": 0.1941,
+      "step": 1037
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.0003410714285714286,
+      "loss": 0.231,
+      "step": 1038
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00034077380952380953,
+      "loss": 0.2515,
+      "step": 1039
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00034047619047619047,
+      "loss": 0.1928,
+      "step": 1040
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00034017857142857146,
+      "loss": 0.2551,
+      "step": 1041
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0003398809523809524,
+      "loss": 0.1941,
+      "step": 1042
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00033958333333333334,
+      "loss": 0.2178,
+      "step": 1043
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00033928571428571433,
+      "loss": 0.2424,
+      "step": 1044
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00033898809523809527,
+      "loss": 0.1978,
+      "step": 1045
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0003386904761904762,
+      "loss": 0.2971,
+      "step": 1046
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00033839285714285715,
+      "loss": 0.209,
+      "step": 1047
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00033809523809523814,
+      "loss": 0.3304,
+      "step": 1048
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0003377976190476191,
+      "loss": 0.2405,
+      "step": 1049
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0003375,
+      "loss": 0.2384,
+      "step": 1050
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00033720238095238096,
+      "loss": 0.2111,
+      "step": 1051
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00033690476190476195,
+      "loss": 0.2861,
+      "step": 1052
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.0003366071428571429,
+      "loss": 0.2644,
+      "step": 1053
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0003363095238095238,
+      "loss": 0.3634,
+      "step": 1054
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0003360119047619048,
+      "loss": 0.4161,
+      "step": 1055
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0003357142857142857,
+      "loss": 0.3934,
+      "step": 1056
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.00033541666666666664,
+      "loss": 0.2475,
+      "step": 1057
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.0003351190476190476,
+      "loss": 0.2075,
+      "step": 1058
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0003348214285714286,
+      "loss": 0.2203,
+      "step": 1059
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0003345238095238095,
+      "loss": 0.2575,
+      "step": 1060
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00033422619047619045,
+      "loss": 0.2083,
+      "step": 1061
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.0003339285714285714,
+      "loss": 0.2592,
+      "step": 1062
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0003336309523809524,
+      "loss": 0.272,
+      "step": 1063
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.0003333333333333333,
+      "loss": 0.2289,
+      "step": 1064
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00033303571428571426,
+      "loss": 0.2054,
+      "step": 1065
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00033273809523809525,
+      "loss": 0.2495,
+      "step": 1066
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.0003324404761904762,
+      "loss": 0.248,
+      "step": 1067
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00033214285714285713,
+      "loss": 0.1639,
+      "step": 1068
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00033184523809523807,
+      "loss": 0.2519,
+      "step": 1069
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00033154761904761906,
+      "loss": 0.3246,
+      "step": 1070
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00033125,
+      "loss": 0.244,
+      "step": 1071
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00033095238095238094,
+      "loss": 0.2755,
+      "step": 1072
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0003306547619047619,
+      "loss": 0.2482,
+      "step": 1073
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00033035714285714287,
+      "loss": 0.2851,
+      "step": 1074
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.0003300595238095238,
+      "loss": 0.298,
+      "step": 1075
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00032976190476190475,
+      "loss": 0.2937,
+      "step": 1076
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00032946428571428574,
+      "loss": 0.2559,
+      "step": 1077
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.0003291666666666667,
+      "loss": 0.2288,
+      "step": 1078
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0003288690476190476,
+      "loss": 0.2818,
+      "step": 1079
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00032857142857142856,
+      "loss": 0.3398,
+      "step": 1080
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.00032827380952380955,
+      "loss": 0.3383,
+      "step": 1081
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.0003279761904761905,
+      "loss": 0.2191,
+      "step": 1082
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00032767857142857143,
+      "loss": 0.2869,
+      "step": 1083
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.00032738095238095237,
+      "loss": 0.2246,
+      "step": 1084
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00032708333333333336,
+      "loss": 0.2791,
+      "step": 1085
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0003267857142857143,
+      "loss": 0.2363,
+      "step": 1086
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00032648809523809524,
+      "loss": 0.2835,
+      "step": 1087
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00032619047619047623,
+      "loss": 0.1896,
+      "step": 1088
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00032589285714285717,
+      "loss": 0.28,
+      "step": 1089
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0003255952380952381,
+      "loss": 0.2528,
+      "step": 1090
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00032529761904761905,
+      "loss": 0.1597,
+      "step": 1091
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 0.1638,
+      "step": 1092
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.000324702380952381,
+      "loss": 0.1685,
+      "step": 1093
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0003244047619047619,
+      "loss": 0.1743,
+      "step": 1094
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00032410714285714286,
+      "loss": 0.2036,
+      "step": 1095
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00032380952380952385,
+      "loss": 0.1537,
+      "step": 1096
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0003235119047619048,
+      "loss": 0.1965,
+      "step": 1097
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.0003232142857142857,
+      "loss": 0.1632,
+      "step": 1098
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.0003229166666666667,
+      "loss": 0.1855,
+      "step": 1099
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00032261904761904766,
+      "loss": 0.2245,
+      "step": 1100
+    },
+    {
+      "epoch": 10.09,
+      "eval_loss": 0.6714171171188354,
+      "eval_runtime": 36.7057,
+      "eval_samples_per_second": 44.87,
+      "eval_steps_per_second": 2.806,
+      "eval_wer": 0.1951649055395453,
+      "step": 1100
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.0003223214285714286,
+      "loss": 0.2124,
+      "step": 1101
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00032202380952380953,
+      "loss": 0.1662,
+      "step": 1102
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.00032172619047619053,
+      "loss": 0.1575,
+      "step": 1103
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 0.1855,
+      "step": 1104
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.0003211309523809524,
+      "loss": 0.1469,
+      "step": 1105
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00032083333333333334,
+      "loss": 0.2269,
+      "step": 1106
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00032053571428571434,
+      "loss": 0.1851,
+      "step": 1107
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0003202380952380952,
+      "loss": 0.1973,
+      "step": 1108
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00031994047619047616,
+      "loss": 0.1702,
+      "step": 1109
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.0003196428571428571,
+      "loss": 0.2572,
+      "step": 1110
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0003193452380952381,
+      "loss": 0.2217,
+      "step": 1111
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00031904761904761903,
+      "loss": 0.3484,
+      "step": 1112
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.00031874999999999997,
+      "loss": 0.2573,
+      "step": 1113
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00031845238095238096,
+      "loss": 0.2933,
+      "step": 1114
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.0003181547619047619,
+      "loss": 0.2721,
+      "step": 1115
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00031785714285714284,
+      "loss": 0.2322,
+      "step": 1116
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.0003175595238095238,
+      "loss": 0.209,
+      "step": 1117
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00031726190476190477,
+      "loss": 0.1959,
+      "step": 1118
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0003169642857142857,
+      "loss": 0.2054,
+      "step": 1119
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00031666666666666665,
+      "loss": 0.2067,
+      "step": 1120
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.0003163690476190476,
+      "loss": 0.1761,
+      "step": 1121
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0003160714285714286,
+      "loss": 0.1438,
+      "step": 1122
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.0003157738095238095,
+      "loss": 0.2193,
+      "step": 1123
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00031547619047619046,
+      "loss": 0.1247,
+      "step": 1124
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00031517857142857145,
+      "loss": 0.1663,
+      "step": 1125
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0003148809523809524,
+      "loss": 0.1558,
+      "step": 1126
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00031458333333333333,
+      "loss": 0.2255,
+      "step": 1127
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00031428571428571427,
+      "loss": 0.1876,
+      "step": 1128
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00031398809523809526,
+      "loss": 0.2509,
+      "step": 1129
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0003136904761904762,
+      "loss": 0.2681,
+      "step": 1130
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.00031339285714285714,
+      "loss": 0.1969,
+      "step": 1131
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0003130952380952381,
+      "loss": 0.2223,
+      "step": 1132
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.00031279761904761907,
+      "loss": 0.2032,
+      "step": 1133
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0003125,
+      "loss": 0.2677,
+      "step": 1134
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00031220238095238095,
+      "loss": 0.2092,
+      "step": 1135
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.00031190476190476194,
+      "loss": 0.2098,
+      "step": 1136
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.0003116071428571429,
+      "loss": 0.2574,
+      "step": 1137
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.0003113095238095238,
+      "loss": 0.2726,
+      "step": 1138
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00031101190476190475,
+      "loss": 0.2737,
+      "step": 1139
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.00031071428571428575,
+      "loss": 0.3136,
+      "step": 1140
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0003104166666666667,
+      "loss": 0.1579,
+      "step": 1141
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.0003101190476190476,
+      "loss": 0.2151,
+      "step": 1142
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00030982142857142856,
+      "loss": 0.1319,
+      "step": 1143
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00030952380952380956,
+      "loss": 0.1572,
+      "step": 1144
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0003092261904761905,
+      "loss": 0.1858,
+      "step": 1145
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00030892857142857143,
+      "loss": 0.1836,
+      "step": 1146
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0003086309523809524,
+      "loss": 0.1712,
+      "step": 1147
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00030833333333333337,
+      "loss": 0.1684,
+      "step": 1148
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0003080357142857143,
+      "loss": 0.2143,
+      "step": 1149
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00030773809523809524,
+      "loss": 0.2053,
+      "step": 1150
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00030744047619047624,
+      "loss": 0.2502,
+      "step": 1151
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.0003071428571428572,
+      "loss": 0.2436,
+      "step": 1152
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.0003068452380952381,
+      "loss": 0.2388,
+      "step": 1153
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00030654761904761905,
+      "loss": 0.1775,
+      "step": 1154
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00030625000000000004,
+      "loss": 0.2329,
+      "step": 1155
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.000305952380952381,
+      "loss": 0.2458,
+      "step": 1156
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0003056547619047619,
+      "loss": 0.2579,
+      "step": 1157
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0003053571428571429,
+      "loss": 0.2407,
+      "step": 1158
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00030505952380952385,
+      "loss": 0.1835,
+      "step": 1159
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.0003047619047619048,
+      "loss": 0.2127,
+      "step": 1160
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0003044642857142857,
+      "loss": 0.2242,
+      "step": 1161
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00030416666666666667,
+      "loss": 0.2784,
+      "step": 1162
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.0003038690476190476,
+      "loss": 0.1671,
+      "step": 1163
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00030357142857142855,
+      "loss": 0.2885,
+      "step": 1164
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0003032738095238095,
+      "loss": 0.3359,
+      "step": 1165
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0003029761904761905,
+      "loss": 0.1643,
+      "step": 1166
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.0003026785714285714,
+      "loss": 0.1567,
+      "step": 1167
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.00030238095238095236,
+      "loss": 0.2185,
+      "step": 1168
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.0003020833333333333,
+      "loss": 0.1849,
+      "step": 1169
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.0003017857142857143,
+      "loss": 0.173,
+      "step": 1170
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0003014880952380952,
+      "loss": 0.2386,
+      "step": 1171
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00030119047619047616,
+      "loss": 0.1614,
+      "step": 1172
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00030089285714285716,
+      "loss": 0.2094,
+      "step": 1173
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.0003005952380952381,
+      "loss": 0.1395,
+      "step": 1174
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00030029761904761904,
+      "loss": 0.1441,
+      "step": 1175
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0003,
+      "loss": 0.2072,
+      "step": 1176
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00029970238095238097,
+      "loss": 0.1485,
+      "step": 1177
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0002994047619047619,
+      "loss": 0.2386,
+      "step": 1178
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.00029910714285714284,
+      "loss": 0.1947,
+      "step": 1179
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0002988095238095238,
+      "loss": 0.1924,
+      "step": 1180
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0002985119047619048,
+      "loss": 0.1867,
+      "step": 1181
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0002982142857142857,
+      "loss": 0.1704,
+      "step": 1182
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00029791666666666665,
+      "loss": 0.3194,
+      "step": 1183
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00029761904761904765,
+      "loss": 0.2865,
+      "step": 1184
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.0002973214285714286,
+      "loss": 0.2274,
+      "step": 1185
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0002970238095238095,
+      "loss": 0.193,
+      "step": 1186
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00029672619047619046,
+      "loss": 0.2034,
+      "step": 1187
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.00029642857142857145,
+      "loss": 0.2145,
+      "step": 1188
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0002961309523809524,
+      "loss": 0.2647,
+      "step": 1189
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00029583333333333333,
+      "loss": 0.3336,
+      "step": 1190
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00029553571428571427,
+      "loss": 0.2042,
+      "step": 1191
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00029523809523809526,
+      "loss": 0.1762,
+      "step": 1192
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0002949404761904762,
+      "loss": 0.192,
+      "step": 1193
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00029464285714285714,
+      "loss": 0.1797,
+      "step": 1194
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.00029434523809523813,
+      "loss": 0.2608,
+      "step": 1195
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00029404761904761907,
+      "loss": 0.2411,
+      "step": 1196
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00029375,
+      "loss": 0.2309,
+      "step": 1197
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00029345238095238095,
+      "loss": 0.1886,
+      "step": 1198
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.00029315476190476194,
+      "loss": 0.1988,
+      "step": 1199
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.0002928571428571429,
+      "loss": 0.1394,
+      "step": 1200
+    },
+    {
+      "epoch": 11.01,
+      "eval_loss": 0.6863502264022827,
+      "eval_runtime": 36.6212,
+      "eval_samples_per_second": 44.974,
+      "eval_steps_per_second": 2.813,
+      "eval_wer": 0.19538504642971502,
+      "step": 1200
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0002925595238095238,
+      "loss": 0.1949,
+      "step": 1201
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00029226190476190476,
+      "loss": 0.1418,
+      "step": 1202
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.00029196428571428575,
+      "loss": 0.1319,
+      "step": 1203
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0002916666666666667,
+      "loss": 0.1739,
+      "step": 1204
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00029136904761904763,
+      "loss": 0.1327,
+      "step": 1205
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.0002910714285714286,
+      "loss": 0.1159,
+      "step": 1206
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.00029077380952380956,
+      "loss": 0.152,
+      "step": 1207
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0002904761904761905,
+      "loss": 0.1804,
+      "step": 1208
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00029017857142857144,
+      "loss": 0.1072,
+      "step": 1209
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00028988095238095243,
+      "loss": 0.2076,
+      "step": 1210
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00028958333333333337,
+      "loss": 0.1438,
+      "step": 1211
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0002892857142857143,
+      "loss": 0.1612,
+      "step": 1212
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.0002889880952380952,
+      "loss": 0.1654,
+      "step": 1213
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.0002886904761904762,
+      "loss": 0.1786,
+      "step": 1214
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0002883928571428571,
+      "loss": 0.1734,
+      "step": 1215
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00028809523809523806,
+      "loss": 0.1682,
+      "step": 1216
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00028779761904761906,
+      "loss": 0.202,
+      "step": 1217
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0002875,
+      "loss": 0.1828,
+      "step": 1218
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00028720238095238093,
+      "loss": 0.1843,
+      "step": 1219
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00028690476190476187,
+      "loss": 0.2117,
+      "step": 1220
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.00028660714285714287,
+      "loss": 0.2144,
+      "step": 1221
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0002863095238095238,
+      "loss": 0.1483,
+      "step": 1222
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00028601190476190474,
+      "loss": 0.1576,
+      "step": 1223
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0002857142857142857,
+      "loss": 0.2629,
+      "step": 1224
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0002854166666666667,
+      "loss": 0.1646,
+      "step": 1225
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.0002851190476190476,
+      "loss": 0.217,
+      "step": 1226
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00028482142857142855,
+      "loss": 0.1987,
+      "step": 1227
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.00028452380952380954,
+      "loss": 0.1668,
+      "step": 1228
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0002842261904761905,
+      "loss": 0.1597,
+      "step": 1229
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0002839285714285714,
+      "loss": 0.1058,
+      "step": 1230
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00028363095238095236,
+      "loss": 0.1551,
+      "step": 1231
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00028333333333333335,
+      "loss": 0.1942,
+      "step": 1232
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0002830357142857143,
+      "loss": 0.1844,
+      "step": 1233
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00028273809523809523,
+      "loss": 0.1627,
+      "step": 1234
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00028244047619047617,
+      "loss": 0.154,
+      "step": 1235
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00028214285714285716,
+      "loss": 0.1297,
+      "step": 1236
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.0002818452380952381,
+      "loss": 0.2614,
+      "step": 1237
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00028154761904761904,
+      "loss": 0.2313,
+      "step": 1238
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00028125000000000003,
+      "loss": 0.1651,
+      "step": 1239
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00028095238095238097,
+      "loss": 0.1829,
+      "step": 1240
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0002806547619047619,
+      "loss": 0.2098,
+      "step": 1241
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00028035714285714285,
+      "loss": 0.1887,
+      "step": 1242
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.00028005952380952384,
+      "loss": 0.2172,
+      "step": 1243
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.0002797619047619048,
+      "loss": 0.2585,
+      "step": 1244
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.0002794642857142857,
+      "loss": 0.1443,
+      "step": 1245
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00027916666666666666,
+      "loss": 0.1834,
+      "step": 1246
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00027886904761904765,
+      "loss": 0.2857,
+      "step": 1247
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0002785714285714286,
+      "loss": 0.2706,
+      "step": 1248
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.00027827380952380953,
+      "loss": 0.2763,
+      "step": 1249
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00027797619047619047,
+      "loss": 0.1234,
+      "step": 1250
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00027767857142857146,
+      "loss": 0.1077,
+      "step": 1251
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.0002773809523809524,
+      "loss": 0.1529,
+      "step": 1252
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00027708333333333334,
+      "loss": 0.1292,
+      "step": 1253
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00027678571428571433,
+      "loss": 0.1853,
+      "step": 1254
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00027648809523809527,
+      "loss": 0.1373,
+      "step": 1255
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0002761904761904762,
+      "loss": 0.1704,
+      "step": 1256
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00027589285714285715,
+      "loss": 0.1279,
+      "step": 1257
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00027559523809523814,
+      "loss": 0.1314,
+      "step": 1258
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.0002752976190476191,
+      "loss": 0.1636,
+      "step": 1259
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.000275,
+      "loss": 0.1882,
+      "step": 1260
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00027470238095238096,
+      "loss": 0.1263,
+      "step": 1261
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00027440476190476195,
+      "loss": 0.1763,
+      "step": 1262
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.0002741071428571429,
+      "loss": 0.1508,
+      "step": 1263
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0002738095238095238,
+      "loss": 0.1695,
+      "step": 1264
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0002735119047619048,
+      "loss": 0.1285,
+      "step": 1265
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0002732142857142857,
+      "loss": 0.2407,
+      "step": 1266
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00027291666666666664,
+      "loss": 0.2114,
+      "step": 1267
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.0002726190476190476,
+      "loss": 0.2901,
+      "step": 1268
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0002723214285714286,
+      "loss": 0.2208,
+      "step": 1269
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.0002720238095238095,
+      "loss": 0.2494,
+      "step": 1270
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00027172619047619045,
+      "loss": 0.1735,
+      "step": 1271
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.0002714285714285714,
+      "loss": 0.2889,
+      "step": 1272
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.0002711309523809524,
+      "loss": 0.2026,
+      "step": 1273
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.0002708333333333333,
+      "loss": 0.2269,
+      "step": 1274
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00027053571428571426,
+      "loss": 0.189,
+      "step": 1275
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00027023809523809525,
+      "loss": 0.1365,
+      "step": 1276
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0002699404761904762,
+      "loss": 0.1396,
+      "step": 1277
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00026964285714285713,
+      "loss": 0.1525,
+      "step": 1278
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00026934523809523807,
+      "loss": 0.1592,
+      "step": 1279
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00026904761904761906,
+      "loss": 0.1397,
+      "step": 1280
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00026875,
+      "loss": 0.1986,
+      "step": 1281
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.00026845238095238094,
+      "loss": 0.2033,
+      "step": 1282
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.0002681547619047619,
+      "loss": 0.1507,
+      "step": 1283
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00026785714285714287,
+      "loss": 0.1721,
+      "step": 1284
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0002675595238095238,
+      "loss": 0.1695,
+      "step": 1285
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00026726190476190475,
+      "loss": 0.103,
+      "step": 1286
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00026696428571428574,
+      "loss": 0.1791,
+      "step": 1287
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0002666666666666667,
+      "loss": 0.143,
+      "step": 1288
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.0002663690476190476,
+      "loss": 0.2243,
+      "step": 1289
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.00026607142857142856,
+      "loss": 0.2203,
+      "step": 1290
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00026577380952380955,
+      "loss": 0.1993,
+      "step": 1291
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0002654761904761905,
+      "loss": 0.1208,
+      "step": 1292
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00026517857142857143,
+      "loss": 0.2457,
+      "step": 1293
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00026488095238095237,
+      "loss": 0.1415,
+      "step": 1294
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00026458333333333336,
+      "loss": 0.1852,
+      "step": 1295
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0002642857142857143,
+      "loss": 0.2278,
+      "step": 1296
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00026398809523809524,
+      "loss": 0.2048,
+      "step": 1297
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00026369047619047623,
+      "loss": 0.2362,
+      "step": 1298
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.00026339285714285717,
+      "loss": 0.2281,
+      "step": 1299
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0002630952380952381,
+      "loss": 0.1288,
+      "step": 1300
+    },
+    {
+      "epoch": 11.93,
+      "eval_loss": 0.6695712208747864,
+      "eval_runtime": 36.4005,
+      "eval_samples_per_second": 45.247,
+      "eval_steps_per_second": 2.83,
+      "eval_wer": 0.20170909382004482,
+      "step": 1300
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00026279761904761905,
+      "loss": 0.14,
+      "step": 1301
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00026250000000000004,
+      "loss": 0.1574,
+      "step": 1302
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.000262202380952381,
+      "loss": 0.1975,
+      "step": 1303
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0002619047619047619,
+      "loss": 0.2476,
+      "step": 1304
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00026160714285714285,
+      "loss": 0.1416,
+      "step": 1305
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00026130952380952385,
+      "loss": 0.1595,
+      "step": 1306
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.0002610119047619048,
+      "loss": 0.1912,
+      "step": 1307
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0002607142857142857,
+      "loss": 0.2233,
+      "step": 1308
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.0002604166666666667,
+      "loss": 0.1118,
+      "step": 1309
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00026011904761904766,
+      "loss": 0.1067,
+      "step": 1310
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0002598214285714286,
+      "loss": 0.115,
+      "step": 1311
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00025952380952380953,
+      "loss": 0.144,
+      "step": 1312
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.0002592261904761905,
+      "loss": 0.1352,
+      "step": 1313
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.00025892857142857146,
+      "loss": 0.1781,
+      "step": 1314
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.0002586309523809524,
+      "loss": 0.1279,
+      "step": 1315
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00025833333333333334,
+      "loss": 0.1698,
+      "step": 1316
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00025803571428571434,
+      "loss": 0.1028,
+      "step": 1317
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0002577380952380952,
+      "loss": 0.0987,
+      "step": 1318
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00025744047619047616,
+      "loss": 0.1648,
+      "step": 1319
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0002571428571428571,
+      "loss": 0.1405,
+      "step": 1320
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.0002568452380952381,
+      "loss": 0.1909,
+      "step": 1321
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00025654761904761903,
+      "loss": 0.1348,
+      "step": 1322
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00025624999999999997,
+      "loss": 0.1823,
+      "step": 1323
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.00025595238095238096,
+      "loss": 0.193,
+      "step": 1324
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.0002556547619047619,
+      "loss": 0.1911,
+      "step": 1325
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00025535714285714284,
+      "loss": 0.2173,
+      "step": 1326
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0002550595238095238,
+      "loss": 0.1886,
+      "step": 1327
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00025476190476190477,
+      "loss": 0.2151,
+      "step": 1328
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.0002544642857142857,
+      "loss": 0.2166,
+      "step": 1329
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00025416666666666665,
+      "loss": 0.1235,
+      "step": 1330
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0002538690476190476,
+      "loss": 0.2091,
+      "step": 1331
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0002535714285714286,
+      "loss": 0.2178,
+      "step": 1332
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0002532738095238095,
+      "loss": 0.2451,
+      "step": 1333
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00025297619047619046,
+      "loss": 0.112,
+      "step": 1334
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.00025267857142857145,
+      "loss": 0.1321,
+      "step": 1335
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.0002523809523809524,
+      "loss": 0.1029,
+      "step": 1336
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.0002520833333333333,
+      "loss": 0.1416,
+      "step": 1337
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00025178571428571426,
+      "loss": 0.1463,
+      "step": 1338
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00025148809523809526,
+      "loss": 0.1034,
+      "step": 1339
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.0002511904761904762,
+      "loss": 0.1351,
+      "step": 1340
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00025089285714285713,
+      "loss": 0.1211,
+      "step": 1341
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.0002505952380952381,
+      "loss": 0.1213,
+      "step": 1342
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00025029761904761907,
+      "loss": 0.1272,
+      "step": 1343
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00025,
+      "loss": 0.2052,
+      "step": 1344
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00024970238095238094,
+      "loss": 0.1208,
+      "step": 1345
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00024940476190476194,
+      "loss": 0.1054,
+      "step": 1346
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.0002491071428571429,
+      "loss": 0.1703,
+      "step": 1347
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.0002488095238095238,
+      "loss": 0.2391,
+      "step": 1348
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00024851190476190475,
+      "loss": 0.1497,
+      "step": 1349
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00024821428571428575,
+      "loss": 0.1565,
+      "step": 1350
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.0002479166666666667,
+      "loss": 0.204,
+      "step": 1351
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.0002476190476190476,
+      "loss": 0.2001,
+      "step": 1352
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00024732142857142856,
+      "loss": 0.2262,
+      "step": 1353
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00024702380952380955,
+      "loss": 0.1062,
+      "step": 1354
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0002467261904761905,
+      "loss": 0.2098,
+      "step": 1355
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00024642857142857143,
+      "loss": 0.151,
+      "step": 1356
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0002461309523809524,
+      "loss": 0.1836,
+      "step": 1357
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.0002458333333333333,
+      "loss": 0.3832,
+      "step": 1358
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.0002455357142857143,
+      "loss": 0.1313,
+      "step": 1359
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.00024523809523809524,
+      "loss": 0.1036,
+      "step": 1360
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.0002449404761904762,
+      "loss": 0.1144,
+      "step": 1361
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.0002446428571428571,
+      "loss": 0.1801,
+      "step": 1362
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.0002443452380952381,
+      "loss": 0.1363,
+      "step": 1363
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00024404761904761905,
+      "loss": 0.131,
+      "step": 1364
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.00024375,
+      "loss": 0.1823,
+      "step": 1365
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00024345238095238095,
+      "loss": 0.1294,
+      "step": 1366
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.0002431547619047619,
+      "loss": 0.1978,
+      "step": 1367
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00024285714285714286,
+      "loss": 0.163,
+      "step": 1368
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00024255952380952382,
+      "loss": 0.153,
+      "step": 1369
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00024226190476190476,
+      "loss": 0.113,
+      "step": 1370
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.00024196428571428573,
+      "loss": 0.1363,
+      "step": 1371
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00024166666666666667,
+      "loss": 0.1721,
+      "step": 1372
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00024136904761904763,
+      "loss": 0.1712,
+      "step": 1373
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00024107142857142857,
+      "loss": 0.1585,
+      "step": 1374
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00024077380952380954,
+      "loss": 0.2059,
+      "step": 1375
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00024047619047619048,
+      "loss": 0.2307,
+      "step": 1376
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00024017857142857144,
+      "loss": 0.1444,
+      "step": 1377
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00023988095238095238,
+      "loss": 0.1862,
+      "step": 1378
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00023958333333333335,
+      "loss": 0.2154,
+      "step": 1379
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.0002392857142857143,
+      "loss": 0.1591,
+      "step": 1380
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00023898809523809525,
+      "loss": 0.2057,
+      "step": 1381
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00023869047619047622,
+      "loss": 0.2325,
+      "step": 1382
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.00023839285714285716,
+      "loss": 0.2616,
+      "step": 1383
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.0002380952380952381,
+      "loss": 0.1244,
+      "step": 1384
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00023779761904761903,
+      "loss": 0.1533,
+      "step": 1385
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0002375,
+      "loss": 0.1774,
+      "step": 1386
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00023720238095238094,
+      "loss": 0.1178,
+      "step": 1387
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.0002369047619047619,
+      "loss": 0.1291,
+      "step": 1388
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00023660714285714284,
+      "loss": 0.1371,
+      "step": 1389
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0002363095238095238,
+      "loss": 0.1334,
+      "step": 1390
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00023601190476190477,
+      "loss": 0.1364,
+      "step": 1391
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0002357142857142857,
+      "loss": 0.1195,
+      "step": 1392
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00023541666666666668,
+      "loss": 0.1139,
+      "step": 1393
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00023511904761904762,
+      "loss": 0.1623,
+      "step": 1394
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00023482142857142858,
+      "loss": 0.1182,
+      "step": 1395
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00023452380952380952,
+      "loss": 0.1153,
+      "step": 1396
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.0002342261904761905,
+      "loss": 0.1796,
+      "step": 1397
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00023392857142857143,
+      "loss": 0.1492,
+      "step": 1398
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0002336309523809524,
+      "loss": 0.1594,
+      "step": 1399
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00023333333333333333,
+      "loss": 0.1568,
+      "step": 1400
+    },
+    {
+      "epoch": 12.84,
+      "eval_loss": 0.6468299031257629,
+      "eval_runtime": 36.6604,
+      "eval_samples_per_second": 44.926,
+      "eval_steps_per_second": 2.81,
+      "eval_wer": 0.1842979506884406,
+      "step": 1400
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.0002330357142857143,
+      "loss": 0.2115,
+      "step": 1401
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00023273809523809524,
+      "loss": 0.1733,
+      "step": 1402
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0002324404761904762,
+      "loss": 0.2107,
+      "step": 1403
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00023214285714285717,
+      "loss": 0.1567,
+      "step": 1404
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.0002318452380952381,
+      "loss": 0.1368,
+      "step": 1405
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00023154761904761907,
+      "loss": 0.1364,
+      "step": 1406
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00023125,
+      "loss": 0.2124,
+      "step": 1407
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00023095238095238098,
+      "loss": 0.2395,
+      "step": 1408
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00023065476190476191,
+      "loss": 0.1024,
+      "step": 1409
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00023035714285714285,
+      "loss": 0.1436,
+      "step": 1410
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.0002300595238095238,
+      "loss": 0.1906,
+      "step": 1411
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00022976190476190476,
+      "loss": 0.1522,
+      "step": 1412
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.0002294642857142857,
+      "loss": 0.1353,
+      "step": 1413
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00022916666666666666,
+      "loss": 0.1127,
+      "step": 1414
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00022886904761904763,
+      "loss": 0.2034,
+      "step": 1415
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00022857142857142857,
+      "loss": 0.1715,
+      "step": 1416
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00022827380952380953,
+      "loss": 0.2118,
+      "step": 1417
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00022797619047619047,
+      "loss": 0.0937,
+      "step": 1418
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00022767857142857144,
+      "loss": 0.1018,
+      "step": 1419
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00022738095238095238,
+      "loss": 0.1135,
+      "step": 1420
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00022708333333333334,
+      "loss": 0.1267,
+      "step": 1421
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00022678571428571428,
+      "loss": 0.1508,
+      "step": 1422
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00022648809523809525,
+      "loss": 0.1363,
+      "step": 1423
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.00022619047619047618,
+      "loss": 0.1145,
+      "step": 1424
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00022589285714285715,
+      "loss": 0.1299,
+      "step": 1425
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.00022559523809523812,
+      "loss": 0.1418,
+      "step": 1426
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00022529761904761906,
+      "loss": 0.0964,
+      "step": 1427
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 0.1454,
+      "step": 1428
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00022470238095238096,
+      "loss": 0.1608,
+      "step": 1429
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00022440476190476193,
+      "loss": 0.1438,
+      "step": 1430
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00022410714285714286,
+      "loss": 0.129,
+      "step": 1431
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00022380952380952383,
+      "loss": 0.0879,
+      "step": 1432
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00022351190476190477,
+      "loss": 0.1167,
+      "step": 1433
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00022321428571428573,
+      "loss": 0.1736,
+      "step": 1434
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.00022291666666666667,
+      "loss": 0.0996,
+      "step": 1435
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.0002226190476190476,
+      "loss": 0.1095,
+      "step": 1436
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00022232142857142855,
+      "loss": 0.1721,
+      "step": 1437
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00022202380952380952,
+      "loss": 0.175,
+      "step": 1438
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00022172619047619048,
+      "loss": 0.183,
+      "step": 1439
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00022142857142857142,
+      "loss": 0.1674,
+      "step": 1440
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.0002211309523809524,
+      "loss": 0.2182,
+      "step": 1441
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00022083333333333333,
+      "loss": 0.1832,
+      "step": 1442
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.0002205357142857143,
+      "loss": 0.1166,
+      "step": 1443
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00022023809523809523,
+      "loss": 0.1084,
+      "step": 1444
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.0002199404761904762,
+      "loss": 0.1335,
+      "step": 1445
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.00021964285714285713,
+      "loss": 0.1491,
+      "step": 1446
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0002193452380952381,
+      "loss": 0.1297,
+      "step": 1447
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00021904761904761904,
+      "loss": 0.1244,
+      "step": 1448
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00021875,
+      "loss": 0.1045,
+      "step": 1449
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00021845238095238097,
+      "loss": 0.0786,
+      "step": 1450
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.0002181547619047619,
+      "loss": 0.1038,
+      "step": 1451
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00021785714285714287,
+      "loss": 0.1676,
+      "step": 1452
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.0002175595238095238,
+      "loss": 0.1288,
+      "step": 1453
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00021726190476190478,
+      "loss": 0.1336,
+      "step": 1454
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00021696428571428572,
+      "loss": 0.11,
+      "step": 1455
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00021666666666666668,
+      "loss": 0.1587,
+      "step": 1456
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00021636904761904762,
+      "loss": 0.1601,
+      "step": 1457
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.0002160714285714286,
+      "loss": 0.1811,
+      "step": 1458
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00021577380952380953,
+      "loss": 0.1386,
+      "step": 1459
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.0002154761904761905,
+      "loss": 0.1419,
+      "step": 1460
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.00021517857142857146,
+      "loss": 0.1536,
+      "step": 1461
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.0002148809523809524,
+      "loss": 0.0938,
+      "step": 1462
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00021458333333333334,
+      "loss": 0.1763,
+      "step": 1463
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00021428571428571427,
+      "loss": 0.232,
+      "step": 1464
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00021398809523809524,
+      "loss": 0.1602,
+      "step": 1465
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.00021369047619047618,
+      "loss": 0.2694,
+      "step": 1466
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00021339285714285714,
+      "loss": 0.2889,
+      "step": 1467
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.00021309523809523808,
+      "loss": 0.0837,
+      "step": 1468
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00021279761904761905,
+      "loss": 0.126,
+      "step": 1469
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.0002125,
+      "loss": 0.1299,
+      "step": 1470
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.00021220238095238095,
+      "loss": 0.1365,
+      "step": 1471
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0002119047619047619,
+      "loss": 0.0967,
+      "step": 1472
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00021160714285714286,
+      "loss": 0.1193,
+      "step": 1473
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.00021130952380952382,
+      "loss": 0.0853,
+      "step": 1474
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00021101190476190476,
+      "loss": 0.1248,
+      "step": 1475
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00021071428571428573,
+      "loss": 0.1366,
+      "step": 1476
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00021041666666666667,
+      "loss": 0.1377,
+      "step": 1477
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00021011904761904763,
+      "loss": 0.1385,
+      "step": 1478
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.00020982142857142857,
+      "loss": 0.1632,
+      "step": 1479
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00020952380952380954,
+      "loss": 0.157,
+      "step": 1480
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00020922619047619048,
+      "loss": 0.1555,
+      "step": 1481
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.00020892857142857144,
+      "loss": 0.1179,
+      "step": 1482
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00020863095238095238,
+      "loss": 0.1182,
+      "step": 1483
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00020833333333333335,
+      "loss": 0.1812,
+      "step": 1484
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.0002080357142857143,
+      "loss": 0.2239,
+      "step": 1485
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.00020773809523809525,
+      "loss": 0.1251,
+      "step": 1486
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.00020744047619047622,
+      "loss": 0.1767,
+      "step": 1487
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00020714285714285716,
+      "loss": 0.1284,
+      "step": 1488
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.0002068452380952381,
+      "loss": 0.1663,
+      "step": 1489
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 0.00020654761904761903,
+      "loss": 0.1667,
+      "step": 1490
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.00020625,
+      "loss": 0.1841,
+      "step": 1491
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.00020595238095238094,
+      "loss": 0.2747,
+      "step": 1492
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 0.0002056547619047619,
+      "loss": 0.115,
+      "step": 1493
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 0.00020535714285714284,
+      "loss": 0.1718,
+      "step": 1494
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.0002050595238095238,
+      "loss": 0.1571,
+      "step": 1495
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.00020476190476190477,
+      "loss": 0.1371,
+      "step": 1496
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 0.0002044642857142857,
+      "loss": 0.0953,
+      "step": 1497
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 0.00020416666666666668,
+      "loss": 0.1224,
+      "step": 1498
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.00020386904761904762,
+      "loss": 0.114,
+      "step": 1499
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.00020357142857142858,
+      "loss": 0.1269,
+      "step": 1500
+    },
+    {
+      "epoch": 13.76,
+      "eval_loss": 0.673614501953125,
+      "eval_runtime": 36.7343,
+      "eval_samples_per_second": 44.835,
+      "eval_steps_per_second": 2.804,
+      "eval_wer": 0.1965457893051553,
+      "step": 1500
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 0.00020327380952380952,
+      "loss": 0.1228,
+      "step": 1501
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.0002029761904761905,
+      "loss": 0.1247,
+      "step": 1502
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 0.00020267857142857143,
+      "loss": 0.151,
+      "step": 1503
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 0.0002023809523809524,
+      "loss": 0.1346,
+      "step": 1504
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.00020208333333333333,
+      "loss": 0.1368,
+      "step": 1505
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.0002017857142857143,
+      "loss": 0.1749,
+      "step": 1506
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.00020148809523809523,
+      "loss": 0.1633,
+      "step": 1507
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.0002011904761904762,
+      "loss": 0.1752,
+      "step": 1508
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.00020089285714285717,
+      "loss": 0.1149,
+      "step": 1509
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 0.0002005952380952381,
+      "loss": 0.2126,
+      "step": 1510
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.00020029761904761907,
+      "loss": 0.1729,
+      "step": 1511
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 0.0002,
+      "loss": 0.1919,
+      "step": 1512
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.00019970238095238098,
+      "loss": 0.1953,
+      "step": 1513
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.00019940476190476191,
+      "loss": 0.1007,
+      "step": 1514
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 0.00019910714285714285,
+      "loss": 0.1579,
+      "step": 1515
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.0001988095238095238,
+      "loss": 0.2561,
+      "step": 1516
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.00019851190476190476,
+      "loss": 0.1974,
+      "step": 1517
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 0.0001982142857142857,
+      "loss": 0.091,
+      "step": 1518
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00019791666666666666,
+      "loss": 0.129,
+      "step": 1519
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00019761904761904763,
+      "loss": 0.0888,
+      "step": 1520
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.00019732142857142857,
+      "loss": 0.0606,
+      "step": 1521
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 0.00019702380952380953,
+      "loss": 0.1118,
+      "step": 1522
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 0.00019672619047619047,
+      "loss": 0.102,
+      "step": 1523
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 0.00019642857142857144,
+      "loss": 0.1546,
+      "step": 1524
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.00019613095238095238,
+      "loss": 0.152,
+      "step": 1525
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.00019583333333333334,
+      "loss": 0.1793,
+      "step": 1526
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.00019553571428571428,
+      "loss": 0.1228,
+      "step": 1527
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.00019523809523809525,
+      "loss": 0.1196,
+      "step": 1528
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 0.00019494047619047618,
+      "loss": 0.069,
+      "step": 1529
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 0.00019464285714285715,
+      "loss": 0.1271,
+      "step": 1530
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.00019434523809523812,
+      "loss": 0.14,
+      "step": 1531
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.00019404761904761905,
+      "loss": 0.1332,
+      "step": 1532
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.00019375000000000002,
+      "loss": 0.125,
+      "step": 1533
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 0.00019345238095238096,
+      "loss": 0.1207,
+      "step": 1534
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.00019315476190476192,
+      "loss": 0.1329,
+      "step": 1535
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 0.00019285714285714286,
+      "loss": 0.1237,
+      "step": 1536
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 0.00019255952380952383,
+      "loss": 0.1078,
+      "step": 1537
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 0.00019226190476190477,
+      "loss": 0.1049,
+      "step": 1538
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 0.00019196428571428573,
+      "loss": 0.1286,
+      "step": 1539
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 0.00019166666666666667,
+      "loss": 0.1353,
+      "step": 1540
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.0001913690476190476,
+      "loss": 0.0915,
+      "step": 1541
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 0.00019107142857142855,
+      "loss": 0.1373,
+      "step": 1542
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 0.00019077380952380952,
+      "loss": 0.1359,
+      "step": 1543
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.00019047619047619048,
+      "loss": 0.17,
+      "step": 1544
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.00019017857142857142,
+      "loss": 0.1758,
+      "step": 1545
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 0.00018988095238095239,
+      "loss": 0.2215,
+      "step": 1546
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 0.00018958333333333332,
+      "loss": 0.1304,
+      "step": 1547
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 0.0001892857142857143,
+      "loss": 0.1621,
+      "step": 1548
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.00018898809523809523,
+      "loss": 0.1197,
+      "step": 1549
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.0001886904761904762,
+      "loss": 0.137,
+      "step": 1550
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 0.00018839285714285713,
+      "loss": 0.2186,
+      "step": 1551
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.0001880952380952381,
+      "loss": 0.0905,
+      "step": 1552
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 0.00018779761904761904,
+      "loss": 0.0568,
+      "step": 1553
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 0.0001875,
+      "loss": 0.1072,
+      "step": 1554
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.00018720238095238097,
+      "loss": 0.1276,
+      "step": 1555
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.0001869047619047619,
+      "loss": 0.0774,
+      "step": 1556
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.00018660714285714287,
+      "loss": 0.1033,
+      "step": 1557
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 0.0001863095238095238,
+      "loss": 0.1196,
+      "step": 1558
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.00018601190476190478,
+      "loss": 0.098,
+      "step": 1559
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 0.00018571428571428572,
+      "loss": 0.0992,
+      "step": 1560
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.00018541666666666668,
+      "loss": 0.1902,
+      "step": 1561
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 0.00018511904761904762,
+      "loss": 0.1159,
+      "step": 1562
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.0001848214285714286,
+      "loss": 0.1246,
+      "step": 1563
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.00018452380952380953,
+      "loss": 0.0791,
+      "step": 1564
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 0.0001842261904761905,
+      "loss": 0.1391,
+      "step": 1565
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 0.00018392857142857146,
+      "loss": 0.1708,
+      "step": 1566
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 0.0001836309523809524,
+      "loss": 0.1009,
+      "step": 1567
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.117,
+      "step": 1568
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.00018303571428571427,
+      "loss": 0.0986,
+      "step": 1569
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.00018273809523809524,
+      "loss": 0.1158,
+      "step": 1570
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.00018244047619047618,
+      "loss": 0.1628,
+      "step": 1571
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 0.00018214285714285714,
+      "loss": 0.1329,
+      "step": 1572
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.00018184523809523808,
+      "loss": 0.0917,
+      "step": 1573
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.00018154761904761905,
+      "loss": 0.1559,
+      "step": 1574
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 0.00018125,
+      "loss": 0.1625,
+      "step": 1575
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 0.00018095238095238095,
+      "loss": 0.2786,
+      "step": 1576
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0001806547619047619,
+      "loss": 0.1004,
+      "step": 1577
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 0.00018035714285714286,
+      "loss": 0.1035,
+      "step": 1578
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 0.00018005952380952382,
+      "loss": 0.1154,
+      "step": 1579
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.00017976190476190476,
+      "loss": 0.1125,
+      "step": 1580
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.00017946428571428573,
+      "loss": 0.1166,
+      "step": 1581
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.00017916666666666667,
+      "loss": 0.0739,
+      "step": 1582
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 0.00017886904761904763,
+      "loss": 0.0833,
+      "step": 1583
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 0.00017857142857142857,
+      "loss": 0.0757,
+      "step": 1584
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.00017827380952380954,
+      "loss": 0.0681,
+      "step": 1585
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.00017797619047619048,
+      "loss": 0.128,
+      "step": 1586
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 0.00017767857142857144,
+      "loss": 0.1038,
+      "step": 1587
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.00017738095238095238,
+      "loss": 0.1885,
+      "step": 1588
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.00017708333333333335,
+      "loss": 0.1078,
+      "step": 1589
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 0.0001767857142857143,
+      "loss": 0.139,
+      "step": 1590
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 0.00017648809523809525,
+      "loss": 0.1435,
+      "step": 1591
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.00017619047619047622,
+      "loss": 0.1215,
+      "step": 1592
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.00017589285714285716,
+      "loss": 0.1369,
+      "step": 1593
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.0001755952380952381,
+      "loss": 0.1147,
+      "step": 1594
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 0.00017529761904761903,
+      "loss": 0.1293,
+      "step": 1595
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 0.000175,
+      "loss": 0.1992,
+      "step": 1596
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 0.00017470238095238094,
+      "loss": 0.1596,
+      "step": 1597
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 0.0001744047619047619,
+      "loss": 0.113,
+      "step": 1598
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 0.00017410714285714284,
+      "loss": 0.1502,
+      "step": 1599
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.0001738095238095238,
+      "loss": 0.1101,
+      "step": 1600
+    },
+    {
+      "epoch": 14.68,
+      "eval_loss": 0.6688647270202637,
+      "eval_runtime": 36.549,
+      "eval_samples_per_second": 45.063,
+      "eval_steps_per_second": 2.818,
+      "eval_wer": 0.19150256163944926,
+      "step": 1600
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 0.00017351190476190477,
+      "loss": 0.2092,
+      "step": 1601
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.0001732142857142857,
+      "loss": 0.1306,
+      "step": 1602
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 0.00017291666666666668,
+      "loss": 0.1139,
+      "step": 1603
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.00017261904761904762,
+      "loss": 0.0936,
+      "step": 1604
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.00017232142857142858,
+      "loss": 0.0998,
+      "step": 1605
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.00017202380952380952,
+      "loss": 0.1037,
+      "step": 1606
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 0.0001717261904761905,
+      "loss": 0.1007,
+      "step": 1607
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 0.00017142857142857143,
+      "loss": 0.1396,
+      "step": 1608
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.0001711309523809524,
+      "loss": 0.118,
+      "step": 1609
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.00017083333333333333,
+      "loss": 0.1365,
+      "step": 1610
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.0001705357142857143,
+      "loss": 0.1289,
+      "step": 1611
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 0.00017023809523809523,
+      "loss": 0.1594,
+      "step": 1612
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 0.0001699404761904762,
+      "loss": 0.1334,
+      "step": 1613
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.00016964285714285717,
+      "loss": 0.1756,
+      "step": 1614
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 0.0001693452380952381,
+      "loss": 0.1599,
+      "step": 1615
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.00016904761904761907,
+      "loss": 0.1237,
+      "step": 1616
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.00016875,
+      "loss": 0.1413,
+      "step": 1617
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.00016845238095238097,
+      "loss": 0.1499,
+      "step": 1618
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 0.0001681547619047619,
+      "loss": 0.1337,
+      "step": 1619
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.00016785714285714285,
+      "loss": 0.1375,
+      "step": 1620
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 0.0001675595238095238,
+      "loss": 0.1279,
+      "step": 1621
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 0.00016726190476190476,
+      "loss": 0.1517,
+      "step": 1622
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.0001669642857142857,
+      "loss": 0.1419,
+      "step": 1623
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 0.00016666666666666666,
+      "loss": 0.1906,
+      "step": 1624
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 0.00016636904761904763,
+      "loss": 0.1064,
+      "step": 1625
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.00016607142857142857,
+      "loss": 0.1322,
+      "step": 1626
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.00016577380952380953,
+      "loss": 0.0819,
+      "step": 1627
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.00016547619047619047,
+      "loss": 0.1126,
+      "step": 1628
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.00016517857142857144,
+      "loss": 0.0938,
+      "step": 1629
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 0.00016488095238095237,
+      "loss": 0.0992,
+      "step": 1630
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.00016458333333333334,
+      "loss": 0.1247,
+      "step": 1631
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.00016428571428571428,
+      "loss": 0.1414,
+      "step": 1632
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 0.00016398809523809524,
+      "loss": 0.1601,
+      "step": 1633
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 0.00016369047619047618,
+      "loss": 0.1517,
+      "step": 1634
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.00016339285714285715,
+      "loss": 0.1709,
+      "step": 1635
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 0.00016309523809523812,
+      "loss": 0.0885,
+      "step": 1636
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 0.00016279761904761905,
+      "loss": 0.1446,
+      "step": 1637
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 0.1007,
+      "step": 1638
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 0.00016220238095238096,
+      "loss": 0.1018,
+      "step": 1639
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 0.00016190476190476192,
+      "loss": 0.1171,
+      "step": 1640
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.00016160714285714286,
+      "loss": 0.1211,
+      "step": 1641
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.00016130952380952383,
+      "loss": 0.1413,
+      "step": 1642
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 0.00016101190476190477,
+      "loss": 0.0957,
+      "step": 1643
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 0.00016071428571428573,
+      "loss": 0.0991,
+      "step": 1644
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 0.00016041666666666667,
+      "loss": 0.1057,
+      "step": 1645
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 0.0001601190476190476,
+      "loss": 0.1226,
+      "step": 1646
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.00015982142857142855,
+      "loss": 0.1054,
+      "step": 1647
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 0.00015952380952380951,
+      "loss": 0.0746,
+      "step": 1648
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 0.00015922619047619048,
+      "loss": 0.1715,
+      "step": 1649
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.00015892857142857142,
+      "loss": 0.1269,
+      "step": 1650
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 0.00015863095238095239,
+      "loss": 0.1494,
+      "step": 1651
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.00015833333333333332,
+      "loss": 0.0959,
+      "step": 1652
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.0001580357142857143,
+      "loss": 0.089,
+      "step": 1653
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.00015773809523809523,
+      "loss": 0.1809,
+      "step": 1654
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 0.0001574404761904762,
+      "loss": 0.1713,
+      "step": 1655
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.00015714285714285713,
+      "loss": 0.0911,
+      "step": 1656
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 0.0001568452380952381,
+      "loss": 0.1865,
+      "step": 1657
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 0.00015654761904761904,
+      "loss": 0.1093,
+      "step": 1658
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.00015625,
+      "loss": 0.108,
+      "step": 1659
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 0.00015595238095238097,
+      "loss": 0.279,
+      "step": 1660
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.0001556547619047619,
+      "loss": 0.062,
+      "step": 1661
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 0.00015535714285714287,
+      "loss": 0.0691,
+      "step": 1662
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.0001550595238095238,
+      "loss": 0.1567,
+      "step": 1663
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.00015476190476190478,
+      "loss": 0.0955,
+      "step": 1664
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.00015446428571428572,
+      "loss": 0.0718,
+      "step": 1665
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.00015416666666666668,
+      "loss": 0.1116,
+      "step": 1666
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 0.00015386904761904762,
+      "loss": 0.0813,
+      "step": 1667
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.0001535714285714286,
+      "loss": 0.1078,
+      "step": 1668
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 0.00015327380952380953,
+      "loss": 0.1098,
+      "step": 1669
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.0001529761904761905,
+      "loss": 0.0802,
+      "step": 1670
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.00015267857142857146,
+      "loss": 0.1187,
+      "step": 1671
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 0.0001523809523809524,
+      "loss": 0.0898,
+      "step": 1672
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 0.00015208333333333333,
+      "loss": 0.1225,
+      "step": 1673
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 0.00015178571428571427,
+      "loss": 0.1324,
+      "step": 1674
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.00015148809523809524,
+      "loss": 0.1091,
+      "step": 1675
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.00015119047619047618,
+      "loss": 0.1412,
+      "step": 1676
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.00015089285714285714,
+      "loss": 0.1368,
+      "step": 1677
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.00015059523809523808,
+      "loss": 0.0599,
+      "step": 1678
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 0.00015029761904761905,
+      "loss": 0.1246,
+      "step": 1679
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.00015,
+      "loss": 0.0989,
+      "step": 1680
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 0.00014970238095238095,
+      "loss": 0.1071,
+      "step": 1681
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.0001494047619047619,
+      "loss": 0.1598,
+      "step": 1682
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 0.00014910714285714286,
+      "loss": 0.1463,
+      "step": 1683
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 0.00014880952380952382,
+      "loss": 0.1487,
+      "step": 1684
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.00014851190476190476,
+      "loss": 0.1055,
+      "step": 1685
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 0.00014821428571428573,
+      "loss": 0.0822,
+      "step": 1686
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 0.00014791666666666667,
+      "loss": 0.1216,
+      "step": 1687
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.00014761904761904763,
+      "loss": 0.0489,
+      "step": 1688
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.00014732142857142857,
+      "loss": 0.0922,
+      "step": 1689
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.00014702380952380954,
+      "loss": 0.0532,
+      "step": 1690
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 0.00014672619047619048,
+      "loss": 0.0787,
+      "step": 1691
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.00014642857142857144,
+      "loss": 0.1188,
+      "step": 1692
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 0.00014613095238095238,
+      "loss": 0.0916,
+      "step": 1693
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 0.00014583333333333335,
+      "loss": 0.1027,
+      "step": 1694
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 0.0001455357142857143,
+      "loss": 0.088,
+      "step": 1695
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.00014523809523809525,
+      "loss": 0.0986,
+      "step": 1696
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.00014494047619047622,
+      "loss": 0.0915,
+      "step": 1697
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 0.00014464285714285715,
+      "loss": 0.1159,
+      "step": 1698
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.0001443452380952381,
+      "loss": 0.1902,
+      "step": 1699
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 0.00014404761904761903,
+      "loss": 0.1388,
+      "step": 1700
+    },
+    {
+      "epoch": 15.6,
+      "eval_loss": 0.6690436005592346,
+      "eval_runtime": 36.8379,
+      "eval_samples_per_second": 44.709,
+      "eval_steps_per_second": 2.796,
+      "eval_wer": 0.178234069804675,
+      "step": 1700
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.00014375,
+      "loss": 0.0945,
+      "step": 1701
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.00014345238095238094,
+      "loss": 0.1124,
+      "step": 1702
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 0.0001431547619047619,
+      "loss": 0.1366,
+      "step": 1703
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.00014285714285714284,
+      "loss": 0.0965,
+      "step": 1704
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 0.0001425595238095238,
+      "loss": 0.1108,
+      "step": 1705
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.00014226190476190477,
+      "loss": 0.117,
+      "step": 1706
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.0001419642857142857,
+      "loss": 0.1422,
+      "step": 1707
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 0.00014166666666666668,
+      "loss": 0.1125,
+      "step": 1708
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.00014136904761904762,
+      "loss": 0.2592,
+      "step": 1709
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 0.00014107142857142858,
+      "loss": 0.156,
+      "step": 1710
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.00014077380952380952,
+      "loss": 0.0956,
+      "step": 1711
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 0.00014047619047619049,
+      "loss": 0.09,
+      "step": 1712
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.00014017857142857142,
+      "loss": 0.0994,
+      "step": 1713
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.0001398809523809524,
+      "loss": 0.0763,
+      "step": 1714
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.00013958333333333333,
+      "loss": 0.0863,
+      "step": 1715
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 0.0001392857142857143,
+      "loss": 0.0928,
+      "step": 1716
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 0.00013898809523809523,
+      "loss": 0.0654,
+      "step": 1717
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.0001386904761904762,
+      "loss": 0.1179,
+      "step": 1718
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 0.00013839285714285717,
+      "loss": 0.0994,
+      "step": 1719
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.0001380952380952381,
+      "loss": 0.1124,
+      "step": 1720
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.00013779761904761907,
+      "loss": 0.0884,
+      "step": 1721
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 0.0001375,
+      "loss": 0.1646,
+      "step": 1722
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 0.00013720238095238097,
+      "loss": 0.0793,
+      "step": 1723
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.0001369047619047619,
+      "loss": 0.1362,
+      "step": 1724
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.00013660714285714285,
+      "loss": 0.0705,
+      "step": 1725
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.0001363095238095238,
+      "loss": 0.1266,
+      "step": 1726
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 0.00013601190476190476,
+      "loss": 0.1071,
+      "step": 1727
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 0.0001357142857142857,
+      "loss": 0.1134,
+      "step": 1728
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.00013541666666666666,
+      "loss": 0.0701,
+      "step": 1729
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 0.00013511904761904763,
+      "loss": 0.1348,
+      "step": 1730
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 0.00013482142857142856,
+      "loss": 0.1776,
+      "step": 1731
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.00013452380952380953,
+      "loss": 0.2107,
+      "step": 1732
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 0.00013422619047619047,
+      "loss": 0.0709,
+      "step": 1733
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 0.00013392857142857144,
+      "loss": 0.1542,
+      "step": 1734
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.00013363095238095237,
+      "loss": 0.1311,
+      "step": 1735
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.0729,
+      "step": 1736
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.00013303571428571428,
+      "loss": 0.096,
+      "step": 1737
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.00013273809523809524,
+      "loss": 0.1067,
+      "step": 1738
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.00013244047619047618,
+      "loss": 0.0956,
+      "step": 1739
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 0.00013214285714285715,
+      "loss": 0.0983,
+      "step": 1740
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.00013184523809523811,
+      "loss": 0.1505,
+      "step": 1741
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 0.00013154761904761905,
+      "loss": 0.0834,
+      "step": 1742
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.00013125000000000002,
+      "loss": 0.1417,
+      "step": 1743
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 0.00013095238095238096,
+      "loss": 0.1448,
+      "step": 1744
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.00013065476190476192,
+      "loss": 0.0706,
+      "step": 1745
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 0.00013035714285714286,
+      "loss": 0.0663,
+      "step": 1746
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 0.00013005952380952383,
+      "loss": 0.1197,
+      "step": 1747
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 0.00012976190476190477,
+      "loss": 0.0855,
+      "step": 1748
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.00012946428571428573,
+      "loss": 0.1345,
+      "step": 1749
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.00012916666666666667,
+      "loss": 0.0666,
+      "step": 1750
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.0001288690476190476,
+      "loss": 0.1267,
+      "step": 1751
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 0.00012857142857142855,
+      "loss": 0.081,
+      "step": 1752
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.00012827380952380951,
+      "loss": 0.0718,
+      "step": 1753
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 0.00012797619047619048,
+      "loss": 0.1174,
+      "step": 1754
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 0.00012767857142857142,
+      "loss": 0.1198,
+      "step": 1755
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.00012738095238095238,
+      "loss": 0.1056,
+      "step": 1756
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.00012708333333333332,
+      "loss": 0.105,
+      "step": 1757
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 0.0001267857142857143,
+      "loss": 0.086,
+      "step": 1758
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.00012648809523809523,
+      "loss": 0.129,
+      "step": 1759
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 0.0001261904761904762,
+      "loss": 0.0731,
+      "step": 1760
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.00012589285714285713,
+      "loss": 0.101,
+      "step": 1761
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.0001255952380952381,
+      "loss": 0.0858,
+      "step": 1762
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.00012529761904761904,
+      "loss": 0.113,
+      "step": 1763
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.000125,
+      "loss": 0.1445,
+      "step": 1764
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.00012470238095238097,
+      "loss": 0.1058,
+      "step": 1765
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 0.0001244047619047619,
+      "loss": 0.1318,
+      "step": 1766
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 0.00012410714285714287,
+      "loss": 0.1386,
+      "step": 1767
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.0001238095238095238,
+      "loss": 0.2089,
+      "step": 1768
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 0.00012351190476190478,
+      "loss": 0.1946,
+      "step": 1769
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.00012321428571428572,
+      "loss": 0.1108,
+      "step": 1770
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.00012291666666666665,
+      "loss": 0.0632,
+      "step": 1771
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 0.00012261904761904762,
+      "loss": 0.0662,
+      "step": 1772
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.00012232142857142856,
+      "loss": 0.0505,
+      "step": 1773
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.00012202380952380953,
+      "loss": 0.1315,
+      "step": 1774
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.00012172619047619048,
+      "loss": 0.1036,
+      "step": 1775
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 0.00012142857142857143,
+      "loss": 0.0682,
+      "step": 1776
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 0.00012113095238095238,
+      "loss": 0.1203,
+      "step": 1777
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.00012083333333333333,
+      "loss": 0.0905,
+      "step": 1778
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.00012053571428571429,
+      "loss": 0.1007,
+      "step": 1779
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 0.00012023809523809524,
+      "loss": 0.163,
+      "step": 1780
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.00011994047619047619,
+      "loss": 0.0525,
+      "step": 1781
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.00011964285714285716,
+      "loss": 0.1342,
+      "step": 1782
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 0.00011934523809523811,
+      "loss": 0.1066,
+      "step": 1783
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 0.00011904761904761905,
+      "loss": 0.0884,
+      "step": 1784
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.00011875,
+      "loss": 0.0691,
+      "step": 1785
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.00011845238095238095,
+      "loss": 0.1113,
+      "step": 1786
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.0001181547619047619,
+      "loss": 0.1485,
+      "step": 1787
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 0.00011785714285714286,
+      "loss": 0.0884,
+      "step": 1788
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.00011755952380952381,
+      "loss": 0.1138,
+      "step": 1789
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.00011726190476190476,
+      "loss": 0.0589,
+      "step": 1790
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 0.00011696428571428571,
+      "loss": 0.1474,
+      "step": 1791
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 0.00011666666666666667,
+      "loss": 0.1702,
+      "step": 1792
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.00011636904761904762,
+      "loss": 0.2386,
+      "step": 1793
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 0.00011607142857142858,
+      "loss": 0.0561,
+      "step": 1794
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 0.00011577380952380954,
+      "loss": 0.0518,
+      "step": 1795
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 0.00011547619047619049,
+      "loss": 0.1007,
+      "step": 1796
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.00011517857142857143,
+      "loss": 0.1163,
+      "step": 1797
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.00011488095238095238,
+      "loss": 0.0908,
+      "step": 1798
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.00011458333333333333,
+      "loss": 0.0642,
+      "step": 1799
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.00011428571428571428,
+      "loss": 0.0739,
+      "step": 1800
+    },
+    {
+      "epoch": 16.51,
+      "eval_loss": 0.636353075504303,
+      "eval_runtime": 36.8832,
+      "eval_samples_per_second": 44.654,
+      "eval_steps_per_second": 2.793,
+      "eval_wer": 0.1733909702209414,
+      "step": 1800
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 0.00011398809523809524,
+      "loss": 0.1161,
+      "step": 1801
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 0.00011369047619047619,
+      "loss": 0.1029,
+      "step": 1802
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.00011339285714285714,
+      "loss": 0.0847,
+      "step": 1803
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 0.00011309523809523809,
+      "loss": 0.0892,
+      "step": 1804
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 0.00011279761904761906,
+      "loss": 0.084,
+      "step": 1805
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.00011250000000000001,
+      "loss": 0.0959,
+      "step": 1806
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 0.00011220238095238096,
+      "loss": 0.0757,
+      "step": 1807
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.00011190476190476191,
+      "loss": 0.0454,
+      "step": 1808
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 0.00011160714285714287,
+      "loss": 0.1051,
+      "step": 1809
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.0001113095238095238,
+      "loss": 0.1138,
+      "step": 1810
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.00011101190476190476,
+      "loss": 0.0513,
+      "step": 1811
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.00011071428571428571,
+      "loss": 0.1142,
+      "step": 1812
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 0.00011041666666666666,
+      "loss": 0.1438,
+      "step": 1813
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.00011011904761904761,
+      "loss": 0.0993,
+      "step": 1814
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.00010982142857142857,
+      "loss": 0.1016,
+      "step": 1815
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 0.00010952380952380952,
+      "loss": 0.1222,
+      "step": 1816
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 0.00010922619047619049,
+      "loss": 0.1219,
+      "step": 1817
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.00010892857142857144,
+      "loss": 0.2112,
+      "step": 1818
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 0.00010863095238095239,
+      "loss": 0.2107,
+      "step": 1819
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.00010833333333333334,
+      "loss": 0.0926,
+      "step": 1820
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 0.0001080357142857143,
+      "loss": 0.0608,
+      "step": 1821
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.00010773809523809525,
+      "loss": 0.0786,
+      "step": 1822
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.0001074404761904762,
+      "loss": 0.0934,
+      "step": 1823
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 0.00010714285714285714,
+      "loss": 0.0728,
+      "step": 1824
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 0.00010684523809523809,
+      "loss": 0.09,
+      "step": 1825
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.00010654761904761904,
+      "loss": 0.1064,
+      "step": 1826
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 0.00010625,
+      "loss": 0.1093,
+      "step": 1827
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 0.00010595238095238095,
+      "loss": 0.1165,
+      "step": 1828
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.00010565476190476191,
+      "loss": 0.1184,
+      "step": 1829
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.00010535714285714286,
+      "loss": 0.1474,
+      "step": 1830
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 0.00010505952380952382,
+      "loss": 0.1378,
+      "step": 1831
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.00010476190476190477,
+      "loss": 0.1067,
+      "step": 1832
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 0.00010446428571428572,
+      "loss": 0.107,
+      "step": 1833
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00010416666666666667,
+      "loss": 0.1386,
+      "step": 1834
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00010386904761904763,
+      "loss": 0.1018,
+      "step": 1835
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.00010357142857142858,
+      "loss": 0.1078,
+      "step": 1836
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 0.00010327380952380952,
+      "loss": 0.0957,
+      "step": 1837
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 0.00010297619047619047,
+      "loss": 0.0925,
+      "step": 1838
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.00010267857142857142,
+      "loss": 0.1759,
+      "step": 1839
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 0.00010238095238095239,
+      "loss": 0.1382,
+      "step": 1840
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 0.00010208333333333334,
+      "loss": 0.1441,
+      "step": 1841
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.00010178571428571429,
+      "loss": 0.1077,
+      "step": 1842
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 0.00010148809523809524,
+      "loss": 0.1217,
+      "step": 1843
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 0.0001011904761904762,
+      "loss": 0.1661,
+      "step": 1844
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 0.00010089285714285715,
+      "loss": 0.0629,
+      "step": 1845
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.0001005952380952381,
+      "loss": 0.0667,
+      "step": 1846
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.00010029761904761905,
+      "loss": 0.0838,
+      "step": 1847
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.0001,
+      "loss": 0.095,
+      "step": 1848
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 9.970238095238096e-05,
+      "loss": 0.0978,
+      "step": 1849
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 9.94047619047619e-05,
+      "loss": 0.1134,
+      "step": 1850
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 9.910714285714285e-05,
+      "loss": 0.1552,
+      "step": 1851
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 9.880952380952381e-05,
+      "loss": 0.1465,
+      "step": 1852
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 9.851190476190477e-05,
+      "loss": 0.1127,
+      "step": 1853
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 9.821428571428572e-05,
+      "loss": 0.0991,
+      "step": 1854
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 9.791666666666667e-05,
+      "loss": 0.0906,
+      "step": 1855
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 9.761904761904762e-05,
+      "loss": 0.0847,
+      "step": 1856
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 9.732142857142858e-05,
+      "loss": 0.0729,
+      "step": 1857
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 9.702380952380953e-05,
+      "loss": 0.1175,
+      "step": 1858
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 9.672619047619048e-05,
+      "loss": 0.0731,
+      "step": 1859
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 9.642857142857143e-05,
+      "loss": 0.0511,
+      "step": 1860
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 9.613095238095238e-05,
+      "loss": 0.0825,
+      "step": 1861
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 9.583333333333334e-05,
+      "loss": 0.0768,
+      "step": 1862
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 9.553571428571427e-05,
+      "loss": 0.09,
+      "step": 1863
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 9.523809523809524e-05,
+      "loss": 0.1126,
+      "step": 1864
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 9.494047619047619e-05,
+      "loss": 0.1019,
+      "step": 1865
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 9.464285714285715e-05,
+      "loss": 0.1264,
+      "step": 1866
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 9.43452380952381e-05,
+      "loss": 0.07,
+      "step": 1867
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 9.404761904761905e-05,
+      "loss": 0.0657,
+      "step": 1868
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 9.375e-05,
+      "loss": 0.1125,
+      "step": 1869
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 9.345238095238095e-05,
+      "loss": 0.1469,
+      "step": 1870
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 9.31547619047619e-05,
+      "loss": 0.0735,
+      "step": 1871
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 9.285714285714286e-05,
+      "loss": 0.1193,
+      "step": 1872
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 9.255952380952381e-05,
+      "loss": 0.1006,
+      "step": 1873
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 9.226190476190476e-05,
+      "loss": 0.0742,
+      "step": 1874
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 9.196428571428573e-05,
+      "loss": 0.0634,
+      "step": 1875
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 9.166666666666667e-05,
+      "loss": 0.0904,
+      "step": 1876
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 9.136904761904762e-05,
+      "loss": 0.1332,
+      "step": 1877
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 9.107142857142857e-05,
+      "loss": 0.1588,
+      "step": 1878
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 9.077380952380952e-05,
+      "loss": 0.0653,
+      "step": 1879
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 9.047619047619048e-05,
+      "loss": 0.0711,
+      "step": 1880
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 9.017857142857143e-05,
+      "loss": 0.0951,
+      "step": 1881
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 8.988095238095238e-05,
+      "loss": 0.0535,
+      "step": 1882
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 8.958333333333333e-05,
+      "loss": 0.0702,
+      "step": 1883
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 8.928571428571429e-05,
+      "loss": 0.0847,
+      "step": 1884
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 8.898809523809524e-05,
+      "loss": 0.0955,
+      "step": 1885
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 8.869047619047619e-05,
+      "loss": 0.133,
+      "step": 1886
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 8.839285714285716e-05,
+      "loss": 0.1136,
+      "step": 1887
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 8.809523809523811e-05,
+      "loss": 0.1,
+      "step": 1888
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 8.779761904761905e-05,
+      "loss": 0.1026,
+      "step": 1889
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 8.75e-05,
+      "loss": 0.0865,
+      "step": 1890
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 8.720238095238095e-05,
+      "loss": 0.0939,
+      "step": 1891
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 8.69047619047619e-05,
+      "loss": 0.0931,
+      "step": 1892
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 8.660714285714286e-05,
+      "loss": 0.1059,
+      "step": 1893
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 8.630952380952381e-05,
+      "loss": 0.1058,
+      "step": 1894
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 8.601190476190476e-05,
+      "loss": 0.0489,
+      "step": 1895
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 8.571428571428571e-05,
+      "loss": 0.0869,
+      "step": 1896
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 8.541666666666666e-05,
+      "loss": 0.1203,
+      "step": 1897
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 8.511904761904762e-05,
+      "loss": 0.0807,
+      "step": 1898
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 8.482142857142858e-05,
+      "loss": 0.087,
+      "step": 1899
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 8.452380952380954e-05,
+      "loss": 0.0897,
+      "step": 1900
+    },
+    {
+      "epoch": 17.43,
+      "eval_loss": 0.6480305790901184,
+      "eval_runtime": 36.7987,
+      "eval_samples_per_second": 44.757,
+      "eval_steps_per_second": 2.799,
+      "eval_wer": 0.17479186679474865,
+      "step": 1900
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 8.422619047619049e-05,
+      "loss": 0.0981,
+      "step": 1901
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 8.392857142857143e-05,
+      "loss": 0.0952,
+      "step": 1902
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 8.363095238095238e-05,
+      "loss": 0.1314,
+      "step": 1903
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 8.333333333333333e-05,
+      "loss": 0.0645,
+      "step": 1904
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 8.303571428571428e-05,
+      "loss": 0.1099,
+      "step": 1905
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 8.273809523809524e-05,
+      "loss": 0.0836,
+      "step": 1906
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 8.244047619047619e-05,
+      "loss": 0.0844,
+      "step": 1907
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 8.214285714285714e-05,
+      "loss": 0.1074,
+      "step": 1908
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 8.184523809523809e-05,
+      "loss": 0.066,
+      "step": 1909
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 8.154761904761906e-05,
+      "loss": 0.0932,
+      "step": 1910
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 8.125000000000001e-05,
+      "loss": 0.0805,
+      "step": 1911
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 8.095238095238096e-05,
+      "loss": 0.0981,
+      "step": 1912
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 8.065476190476191e-05,
+      "loss": 0.074,
+      "step": 1913
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 8.035714285714287e-05,
+      "loss": 0.1012,
+      "step": 1914
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 8.00595238095238e-05,
+      "loss": 0.0715,
+      "step": 1915
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 7.976190476190476e-05,
+      "loss": 0.0869,
+      "step": 1916
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 7.946428571428571e-05,
+      "loss": 0.1156,
+      "step": 1917
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 7.916666666666666e-05,
+      "loss": 0.1124,
+      "step": 1918
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 7.886904761904761e-05,
+      "loss": 0.1004,
+      "step": 1919
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 7.857142857142857e-05,
+      "loss": 0.1127,
+      "step": 1920
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 7.827380952380952e-05,
+      "loss": 0.0584,
+      "step": 1921
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 7.797619047619048e-05,
+      "loss": 0.0848,
+      "step": 1922
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 7.767857142857144e-05,
+      "loss": 0.1273,
+      "step": 1923
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 7.738095238095239e-05,
+      "loss": 0.1349,
+      "step": 1924
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 7.708333333333334e-05,
+      "loss": 0.0898,
+      "step": 1925
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 7.67857142857143e-05,
+      "loss": 0.1073,
+      "step": 1926
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 7.648809523809525e-05,
+      "loss": 0.1604,
+      "step": 1927
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 7.61904761904762e-05,
+      "loss": 0.1785,
+      "step": 1928
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 7.589285714285714e-05,
+      "loss": 0.0548,
+      "step": 1929
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 7.559523809523809e-05,
+      "loss": 0.0928,
+      "step": 1930
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 7.529761904761904e-05,
+      "loss": 0.0711,
+      "step": 1931
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 7.5e-05,
+      "loss": 0.0662,
+      "step": 1932
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 7.470238095238095e-05,
+      "loss": 0.0697,
+      "step": 1933
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 7.440476190476191e-05,
+      "loss": 0.0748,
+      "step": 1934
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 7.410714285714286e-05,
+      "loss": 0.0825,
+      "step": 1935
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 7.380952380952382e-05,
+      "loss": 0.0786,
+      "step": 1936
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 7.351190476190477e-05,
+      "loss": 0.0621,
+      "step": 1937
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 7.321428571428572e-05,
+      "loss": 0.1623,
+      "step": 1938
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 7.291666666666667e-05,
+      "loss": 0.116,
+      "step": 1939
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 7.261904761904762e-05,
+      "loss": 0.1189,
+      "step": 1940
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 7.232142857142858e-05,
+      "loss": 0.0676,
+      "step": 1941
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 7.202380952380952e-05,
+      "loss": 0.1002,
+      "step": 1942
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 7.172619047619047e-05,
+      "loss": 0.1061,
+      "step": 1943
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 7.142857142857142e-05,
+      "loss": 0.0815,
+      "step": 1944
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 7.113095238095239e-05,
+      "loss": 0.0864,
+      "step": 1945
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 7.083333333333334e-05,
+      "loss": 0.1093,
+      "step": 1946
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 7.053571428571429e-05,
+      "loss": 0.1162,
+      "step": 1947
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 7.023809523809524e-05,
+      "loss": 0.0918,
+      "step": 1948
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 6.99404761904762e-05,
+      "loss": 0.1159,
+      "step": 1949
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 6.964285714285715e-05,
+      "loss": 0.1043,
+      "step": 1950
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 6.93452380952381e-05,
+      "loss": 0.1428,
+      "step": 1951
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 6.904761904761905e-05,
+      "loss": 0.0792,
+      "step": 1952
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 6.875e-05,
+      "loss": 0.0718,
+      "step": 1953
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 6.845238095238096e-05,
+      "loss": 0.0371,
+      "step": 1954
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 6.81547619047619e-05,
+      "loss": 0.0364,
+      "step": 1955
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 6.785714285714285e-05,
+      "loss": 0.1064,
+      "step": 1956
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 6.755952380952381e-05,
+      "loss": 0.1271,
+      "step": 1957
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 6.726190476190477e-05,
+      "loss": 0.0836,
+      "step": 1958
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 6.696428571428572e-05,
+      "loss": 0.1375,
+      "step": 1959
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.1149,
+      "step": 1960
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 6.636904761904762e-05,
+      "loss": 0.1345,
+      "step": 1961
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 6.607142857142857e-05,
+      "loss": 0.1497,
+      "step": 1962
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 6.577380952380953e-05,
+      "loss": 0.0635,
+      "step": 1963
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 6.547619047619048e-05,
+      "loss": 0.0405,
+      "step": 1964
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 6.517857142857143e-05,
+      "loss": 0.0641,
+      "step": 1965
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 6.488095238095238e-05,
+      "loss": 0.0873,
+      "step": 1966
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 6.458333333333334e-05,
+      "loss": 0.0987,
+      "step": 1967
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 6.428571428571427e-05,
+      "loss": 0.0818,
+      "step": 1968
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 6.398809523809524e-05,
+      "loss": 0.0501,
+      "step": 1969
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 6.369047619047619e-05,
+      "loss": 0.0817,
+      "step": 1970
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 6.339285714285714e-05,
+      "loss": 0.0699,
+      "step": 1971
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 6.30952380952381e-05,
+      "loss": 0.0936,
+      "step": 1972
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 6.279761904761905e-05,
+      "loss": 0.1194,
+      "step": 1973
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 6.25e-05,
+      "loss": 0.0769,
+      "step": 1974
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 6.220238095238095e-05,
+      "loss": 0.0863,
+      "step": 1975
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 6.19047619047619e-05,
+      "loss": 0.0809,
+      "step": 1976
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 6.160714285714286e-05,
+      "loss": 0.0479,
+      "step": 1977
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 6.130952380952381e-05,
+      "loss": 0.0576,
+      "step": 1978
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 6.101190476190476e-05,
+      "loss": 0.1438,
+      "step": 1979
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 6.0714285714285715e-05,
+      "loss": 0.1105,
+      "step": 1980
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 6.041666666666667e-05,
+      "loss": 0.1015,
+      "step": 1981
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 6.011904761904762e-05,
+      "loss": 0.0884,
+      "step": 1982
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 5.982142857142858e-05,
+      "loss": 0.0889,
+      "step": 1983
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 5.9523809523809524e-05,
+      "loss": 0.1118,
+      "step": 1984
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 5.9226190476190476e-05,
+      "loss": 0.1157,
+      "step": 1985
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 5.892857142857143e-05,
+      "loss": 0.1445,
+      "step": 1986
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 5.863095238095238e-05,
+      "loss": 0.2139,
+      "step": 1987
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 5.833333333333333e-05,
+      "loss": 0.1033,
+      "step": 1988
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 5.803571428571429e-05,
+      "loss": 0.0555,
+      "step": 1989
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 5.7738095238095244e-05,
+      "loss": 0.0635,
+      "step": 1990
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 5.744047619047619e-05,
+      "loss": 0.0768,
+      "step": 1991
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 5.714285714285714e-05,
+      "loss": 0.0708,
+      "step": 1992
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 5.6845238095238094e-05,
+      "loss": 0.0716,
+      "step": 1993
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 5.6547619047619046e-05,
+      "loss": 0.0827,
+      "step": 1994
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 5.6250000000000005e-05,
+      "loss": 0.0707,
+      "step": 1995
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 5.595238095238096e-05,
+      "loss": 0.1142,
+      "step": 1996
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 5.56547619047619e-05,
+      "loss": 0.0799,
+      "step": 1997
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 5.5357142857142855e-05,
+      "loss": 0.0646,
+      "step": 1998
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 5.505952380952381e-05,
+      "loss": 0.0667,
+      "step": 1999
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 5.476190476190476e-05,
+      "loss": 0.0795,
+      "step": 2000
+    },
+    {
+      "epoch": 18.35,
+      "eval_loss": 0.6355569958686829,
+      "eval_runtime": 36.3335,
+      "eval_samples_per_second": 45.33,
+      "eval_steps_per_second": 2.835,
+      "eval_wer": 0.16946845981428113,
+      "step": 2000
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 5.446428571428572e-05,
+      "loss": 0.0836,
+      "step": 2001
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 5.416666666666667e-05,
+      "loss": 0.0855,
+      "step": 2002
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 5.386904761904762e-05,
+      "loss": 0.0638,
+      "step": 2003
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 5.357142857142857e-05,
+      "loss": 0.093,
+      "step": 2004
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 5.327380952380952e-05,
+      "loss": 0.0781,
+      "step": 2005
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 5.297619047619047e-05,
+      "loss": 0.1032,
+      "step": 2006
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 5.267857142857143e-05,
+      "loss": 0.0754,
+      "step": 2007
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 5.2380952380952384e-05,
+      "loss": 0.0874,
+      "step": 2008
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 5.208333333333334e-05,
+      "loss": 0.1203,
+      "step": 2009
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 5.178571428571429e-05,
+      "loss": 0.175,
+      "step": 2010
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 5.1488095238095234e-05,
+      "loss": 0.0995,
+      "step": 2011
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 5.1190476190476193e-05,
+      "loss": 0.1214,
+      "step": 2012
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 5.0892857142857146e-05,
+      "loss": 0.0648,
+      "step": 2013
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 5.05952380952381e-05,
+      "loss": 0.0969,
+      "step": 2014
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 5.029761904761905e-05,
+      "loss": 0.0798,
+      "step": 2015
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 5e-05,
+      "loss": 0.0641,
+      "step": 2016
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 4.970238095238095e-05,
+      "loss": 0.0607,
+      "step": 2017
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 4.940476190476191e-05,
+      "loss": 0.058,
+      "step": 2018
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 4.910714285714286e-05,
+      "loss": 0.08,
+      "step": 2019
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 4.880952380952381e-05,
+      "loss": 0.0812,
+      "step": 2020
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 4.8511904761904764e-05,
+      "loss": 0.0707,
+      "step": 2021
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 4.8214285714285716e-05,
+      "loss": 0.0712,
+      "step": 2022
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 4.791666666666667e-05,
+      "loss": 0.0826,
+      "step": 2023
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 4.761904761904762e-05,
+      "loss": 0.0673,
+      "step": 2024
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 4.732142857142857e-05,
+      "loss": 0.0507,
+      "step": 2025
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 4.7023809523809525e-05,
+      "loss": 0.185,
+      "step": 2026
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 4.672619047619048e-05,
+      "loss": 0.1038,
+      "step": 2027
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 4.642857142857143e-05,
+      "loss": 0.0968,
+      "step": 2028
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 4.613095238095238e-05,
+      "loss": 0.0972,
+      "step": 2029
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 4.5833333333333334e-05,
+      "loss": 0.0769,
+      "step": 2030
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 4.5535714285714286e-05,
+      "loss": 0.1079,
+      "step": 2031
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 4.523809523809524e-05,
+      "loss": 0.1168,
+      "step": 2032
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 4.494047619047619e-05,
+      "loss": 0.0554,
+      "step": 2033
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 4.464285714285714e-05,
+      "loss": 0.116,
+      "step": 2034
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 4.4345238095238095e-05,
+      "loss": 0.1561,
+      "step": 2035
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 4.4047619047619054e-05,
+      "loss": 0.1166,
+      "step": 2036
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 4.375e-05,
+      "loss": 0.1488,
+      "step": 2037
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 4.345238095238095e-05,
+      "loss": 0.0781,
+      "step": 2038
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 4.3154761904761904e-05,
+      "loss": 0.0462,
+      "step": 2039
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 4.2857142857142856e-05,
+      "loss": 0.051,
+      "step": 2040
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 4.255952380952381e-05,
+      "loss": 0.0418,
+      "step": 2041
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 4.226190476190477e-05,
+      "loss": 0.0922,
+      "step": 2042
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 4.196428571428571e-05,
+      "loss": 0.0862,
+      "step": 2043
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 4.1666666666666665e-05,
+      "loss": 0.0866,
+      "step": 2044
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 4.136904761904762e-05,
+      "loss": 0.0852,
+      "step": 2045
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 4.107142857142857e-05,
+      "loss": 0.0459,
+      "step": 2046
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 4.077380952380953e-05,
+      "loss": 0.0874,
+      "step": 2047
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 4.047619047619048e-05,
+      "loss": 0.0818,
+      "step": 2048
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 4.017857142857143e-05,
+      "loss": 0.1007,
+      "step": 2049
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 3.988095238095238e-05,
+      "loss": 0.0855,
+      "step": 2050
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 3.958333333333333e-05,
+      "loss": 0.0631,
+      "step": 2051
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 3.928571428571428e-05,
+      "loss": 0.0807,
+      "step": 2052
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 3.898809523809524e-05,
+      "loss": 0.1081,
+      "step": 2053
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 3.8690476190476195e-05,
+      "loss": 0.137,
+      "step": 2054
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 3.839285714285715e-05,
+      "loss": 0.0707,
+      "step": 2055
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 3.80952380952381e-05,
+      "loss": 0.0526,
+      "step": 2056
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 3.7797619047619044e-05,
+      "loss": 0.0926,
+      "step": 2057
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 3.75e-05,
+      "loss": 0.1162,
+      "step": 2058
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 3.7202380952380956e-05,
+      "loss": 0.0954,
+      "step": 2059
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 3.690476190476191e-05,
+      "loss": 0.1053,
+      "step": 2060
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 3.660714285714286e-05,
+      "loss": 0.1704,
+      "step": 2061
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 3.630952380952381e-05,
+      "loss": 0.1157,
+      "step": 2062
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 3.601190476190476e-05,
+      "loss": 0.0659,
+      "step": 2063
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 3.571428571428571e-05,
+      "loss": 0.1095,
+      "step": 2064
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 3.541666666666667e-05,
+      "loss": 0.0532,
+      "step": 2065
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 3.511904761904762e-05,
+      "loss": 0.071,
+      "step": 2066
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 3.4821428571428574e-05,
+      "loss": 0.0824,
+      "step": 2067
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 3.4523809523809526e-05,
+      "loss": 0.0958,
+      "step": 2068
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 3.422619047619048e-05,
+      "loss": 0.088,
+      "step": 2069
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 3.3928571428571424e-05,
+      "loss": 0.15,
+      "step": 2070
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 3.363095238095238e-05,
+      "loss": 0.126,
+      "step": 2071
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.0591,
+      "step": 2072
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 3.303571428571429e-05,
+      "loss": 0.0479,
+      "step": 2073
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 3.273809523809524e-05,
+      "loss": 0.0767,
+      "step": 2074
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 3.244047619047619e-05,
+      "loss": 0.0861,
+      "step": 2075
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 3.214285714285714e-05,
+      "loss": 0.0689,
+      "step": 2076
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 3.1845238095238096e-05,
+      "loss": 0.0604,
+      "step": 2077
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 3.154761904761905e-05,
+      "loss": 0.0707,
+      "step": 2078
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 3.125e-05,
+      "loss": 0.0837,
+      "step": 2079
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 3.095238095238095e-05,
+      "loss": 0.0842,
+      "step": 2080
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 3.0654761904761905e-05,
+      "loss": 0.0696,
+      "step": 2081
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 3.0357142857142857e-05,
+      "loss": 0.0978,
+      "step": 2082
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 3.005952380952381e-05,
+      "loss": 0.0768,
+      "step": 2083
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.9761904761904762e-05,
+      "loss": 0.0602,
+      "step": 2084
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.9464285714285714e-05,
+      "loss": 0.0518,
+      "step": 2085
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.9166666666666666e-05,
+      "loss": 0.0879,
+      "step": 2086
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.8869047619047622e-05,
+      "loss": 0.0623,
+      "step": 2087
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 0.1092,
+      "step": 2088
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.8273809523809523e-05,
+      "loss": 0.0798,
+      "step": 2089
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.797619047619048e-05,
+      "loss": 0.0767,
+      "step": 2090
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.7678571428571428e-05,
+      "loss": 0.107,
+      "step": 2091
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.738095238095238e-05,
+      "loss": 0.0637,
+      "step": 2092
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.7083333333333335e-05,
+      "loss": 0.1391,
+      "step": 2093
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 2.6785714285714284e-05,
+      "loss": 0.1613,
+      "step": 2094
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 2.6488095238095237e-05,
+      "loss": 0.1396,
+      "step": 2095
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2.6190476190476192e-05,
+      "loss": 0.0893,
+      "step": 2096
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 2.5892857142857144e-05,
+      "loss": 0.0672,
+      "step": 2097
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 2.5595238095238097e-05,
+      "loss": 0.0681,
+      "step": 2098
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 2.529761904761905e-05,
+      "loss": 0.0472,
+      "step": 2099
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 2.5e-05,
+      "loss": 0.0823,
+      "step": 2100
+    },
+    {
+      "epoch": 19.27,
+      "eval_loss": 0.6381826400756836,
+      "eval_runtime": 36.8508,
+      "eval_samples_per_second": 44.694,
+      "eval_steps_per_second": 2.795,
+      "eval_wer": 0.16846781940441882,
+      "step": 2100
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2.4702380952380953e-05,
+      "loss": 0.0771,
+      "step": 2101
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2.4404761904761906e-05,
+      "loss": 0.0986,
+      "step": 2102
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 2.4107142857142858e-05,
+      "loss": 0.0623,
+      "step": 2103
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 2.380952380952381e-05,
+      "loss": 0.0533,
+      "step": 2104
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 2.3511904761904762e-05,
+      "loss": 0.0735,
+      "step": 2105
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 2.3214285714285715e-05,
+      "loss": 0.0825,
+      "step": 2106
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 2.2916666666666667e-05,
+      "loss": 0.0804,
+      "step": 2107
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 2.261904761904762e-05,
+      "loss": 0.0386,
+      "step": 2108
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 2.232142857142857e-05,
+      "loss": 0.0503,
+      "step": 2109
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 2.2023809523809527e-05,
+      "loss": 0.096,
+      "step": 2110
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 2.1726190476190476e-05,
+      "loss": 0.0934,
+      "step": 2111
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 2.1428571428571428e-05,
+      "loss": 0.0629,
+      "step": 2112
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 2.1130952380952384e-05,
+      "loss": 0.1387,
+      "step": 2113
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 2.0833333333333333e-05,
+      "loss": 0.1105,
+      "step": 2114
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 2.0535714285714285e-05,
+      "loss": 0.1027,
+      "step": 2115
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 2.023809523809524e-05,
+      "loss": 0.1025,
+      "step": 2116
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.994047619047619e-05,
+      "loss": 0.0798,
+      "step": 2117
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.964285714285714e-05,
+      "loss": 0.1065,
+      "step": 2118
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.9345238095238097e-05,
+      "loss": 0.1064,
+      "step": 2119
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.904761904761905e-05,
+      "loss": 0.0863,
+      "step": 2120
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.875e-05,
+      "loss": 0.08,
+      "step": 2121
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.8452380952380954e-05,
+      "loss": 0.0492,
+      "step": 2122
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.8154761904761906e-05,
+      "loss": 0.0425,
+      "step": 2123
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.7857142857142855e-05,
+      "loss": 0.0471,
+      "step": 2124
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.755952380952381e-05,
+      "loss": 0.1172,
+      "step": 2125
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.7261904761904763e-05,
+      "loss": 0.0696,
+      "step": 2126
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.6964285714285712e-05,
+      "loss": 0.0558,
+      "step": 2127
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.0858,
+      "step": 2128
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.636904761904762e-05,
+      "loss": 0.1149,
+      "step": 2129
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.607142857142857e-05,
+      "loss": 0.0495,
+      "step": 2130
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.5773809523809524e-05,
+      "loss": 0.1075,
+      "step": 2131
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.5476190476190476e-05,
+      "loss": 0.0984,
+      "step": 2132
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.5178571428571429e-05,
+      "loss": 0.0769,
+      "step": 2133
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.4880952380952381e-05,
+      "loss": 0.0844,
+      "step": 2134
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.4583333333333333e-05,
+      "loss": 0.043,
+      "step": 2135
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.0487,
+      "step": 2136
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.398809523809524e-05,
+      "loss": 0.0803,
+      "step": 2137
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.369047619047619e-05,
+      "loss": 0.068,
+      "step": 2138
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.3392857142857142e-05,
+      "loss": 0.1138,
+      "step": 2139
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.3095238095238096e-05,
+      "loss": 0.0654,
+      "step": 2140
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.2797619047619048e-05,
+      "loss": 0.0612,
+      "step": 2141
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.25e-05,
+      "loss": 0.077,
+      "step": 2142
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.2202380952380953e-05,
+      "loss": 0.0798,
+      "step": 2143
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.1904761904761905e-05,
+      "loss": 0.0549,
+      "step": 2144
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.1607142857142857e-05,
+      "loss": 0.1398,
+      "step": 2145
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.130952380952381e-05,
+      "loss": 0.1274,
+      "step": 2146
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.1011904761904764e-05,
+      "loss": 0.0733,
+      "step": 2147
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.0714285714285714e-05,
+      "loss": 0.1126,
+      "step": 2148
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.0416666666666666e-05,
+      "loss": 0.1228,
+      "step": 2149
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.011904761904762e-05,
+      "loss": 0.0837,
+      "step": 2150
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 9.82142857142857e-06,
+      "loss": 0.0424,
+      "step": 2151
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 9.523809523809525e-06,
+      "loss": 0.0918,
+      "step": 2152
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 9.226190476190477e-06,
+      "loss": 0.0414,
+      "step": 2153
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 8.928571428571428e-06,
+      "loss": 0.0876,
+      "step": 2154
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 8.630952380952381e-06,
+      "loss": 0.0931,
+      "step": 2155
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.0784,
+      "step": 2156
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 8.035714285714284e-06,
+      "loss": 0.0839,
+      "step": 2157
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 7.738095238095238e-06,
+      "loss": 0.1082,
+      "step": 2158
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 7.4404761904761905e-06,
+      "loss": 0.0608,
+      "step": 2159
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 7.142857142857143e-06,
+      "loss": 0.1029,
+      "step": 2160
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 6.845238095238095e-06,
+      "loss": 0.09,
+      "step": 2161
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 6.547619047619048e-06,
+      "loss": 0.0542,
+      "step": 2162
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 6.25e-06,
+      "loss": 0.1051,
+      "step": 2163
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 5.9523809523809525e-06,
+      "loss": 0.1139,
+      "step": 2164
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 5.654761904761905e-06,
+      "loss": 0.1148,
+      "step": 2165
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 5.357142857142857e-06,
+      "loss": 0.0877,
+      "step": 2166
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 5.05952380952381e-06,
+      "loss": 0.0984,
+      "step": 2167
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 4.761904761904762e-06,
+      "loss": 0.082,
+      "step": 2168
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 4.464285714285714e-06,
+      "loss": 0.1127,
+      "step": 2169
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 0.0601,
+      "step": 2170
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 3.869047619047619e-06,
+      "loss": 0.1223,
+      "step": 2171
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 3.5714285714285714e-06,
+      "loss": 0.0218,
+      "step": 2172
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 3.273809523809524e-06,
+      "loss": 0.0562,
+      "step": 2173
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 2.9761904761904763e-06,
+      "loss": 0.0542,
+      "step": 2174
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 2.6785714285714285e-06,
+      "loss": 0.0911,
+      "step": 2175
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 2.380952380952381e-06,
+      "loss": 0.1105,
+      "step": 2176
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 2.0833333333333334e-06,
+      "loss": 0.054,
+      "step": 2177
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.7857142857142857e-06,
+      "loss": 0.1336,
+      "step": 2178
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.4880952380952381e-06,
+      "loss": 0.0786,
+      "step": 2179
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.1904761904761906e-06,
+      "loss": 0.1093,
+      "step": 2180
+    },
+    {
+      "epoch": 20.0,
+      "step": 2180,
+      "total_flos": 8.762439478066282e+18,
+      "train_loss": 1.8342130993945747,
+      "train_runtime": 2886.3244,
+      "train_samples_per_second": 24.1,
+      "train_steps_per_second": 0.755
+    }
+  ],
+  "max_steps": 2180,
+  "num_train_epochs": 20,
+  "total_flos": 8.762439478066282e+18,
+  "trial_name": null,
+  "trial_params": null
+}