diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5010 @@
+{
+  "best_metric": 5.988565044524036,
+  "best_model_checkpoint": "./checkpoint-1000",
+  "epoch": 70.54673721340389,
+  "eval_steps": 1000,
+  "global_step": 20000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.800000000000001e-07,
+      "loss": 0.9499,
+      "step": 25
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400000000000001e-07,
+      "loss": 0.6078,
+      "step": 50
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.44e-06,
+      "loss": 0.2291,
+      "step": 75
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.94e-06,
+      "loss": 0.188,
+      "step": 100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.4400000000000004e-06,
+      "loss": 0.1732,
+      "step": 125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.92e-06,
+      "loss": 0.1665,
+      "step": 150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4200000000000007e-06,
+      "loss": 0.1677,
+      "step": 175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.920000000000001e-06,
+      "loss": 0.1551,
+      "step": 200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.42e-06,
+      "loss": 0.1526,
+      "step": 225
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.92e-06,
+      "loss": 0.1524,
+      "step": 250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.420000000000001e-06,
+      "loss": 0.1491,
+      "step": 275
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 5.92e-06,
+      "loss": 0.1179,
+      "step": 300
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 6.42e-06,
+      "loss": 0.1142,
+      "step": 325
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 6.92e-06,
+      "loss": 0.107,
+      "step": 350
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.420000000000001e-06,
+      "loss": 0.1063,
+      "step": 375
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 7.92e-06,
+      "loss": 0.122,
+      "step": 400
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.42e-06,
+      "loss": 0.1165,
+      "step": 425
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.900000000000001e-06,
+      "loss": 0.1099,
+      "step": 450
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 9.4e-06,
+      "loss": 0.1193,
+      "step": 475
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 9.9e-06,
+      "loss": 0.1174,
+      "step": 500
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 9.98974358974359e-06,
+      "loss": 0.1223,
+      "step": 525
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.976923076923077e-06,
+      "loss": 0.1154,
+      "step": 550
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.964102564102564e-06,
+      "loss": 0.1031,
+      "step": 575
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.951282051282051e-06,
+      "loss": 0.0681,
+      "step": 600
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.93846153846154e-06,
+      "loss": 0.0697,
+      "step": 625
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 9.925641025641026e-06,
+      "loss": 0.0719,
+      "step": 650
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 9.912820512820513e-06,
+      "loss": 0.0683,
+      "step": 675
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.9e-06,
+      "loss": 0.0739,
+      "step": 700
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 9.887179487179489e-06,
+      "loss": 0.0779,
+      "step": 725
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 9.874358974358975e-06,
+      "loss": 0.0742,
+      "step": 750
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 9.861538461538462e-06,
+      "loss": 0.0661,
+      "step": 775
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 9.848717948717949e-06,
+      "loss": 0.071,
+      "step": 800
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 9.835897435897438e-06,
+      "loss": 0.0744,
+      "step": 825
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.823076923076924e-06,
+      "loss": 0.0724,
+      "step": 850
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 9.810256410256411e-06,
+      "loss": 0.0423,
+      "step": 875
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 9.797435897435898e-06,
+      "loss": 0.0387,
+      "step": 900
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 9.784615384615387e-06,
+      "loss": 0.0439,
+      "step": 925
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 9.771794871794873e-06,
+      "loss": 0.0416,
+      "step": 950
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 9.75897435897436e-06,
+      "loss": 0.0415,
+      "step": 975
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 9.746153846153847e-06,
+      "loss": 0.0424,
+      "step": 1000
+    },
+    {
+      "epoch": 3.53,
+      "eval_loss": 0.18793095648288727,
+      "eval_runtime": 2384.2986,
+      "eval_samples_per_second": 3.805,
+      "eval_steps_per_second": 0.238,
+      "eval_wer": 5.988565044524036,
+      "step": 1000
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 9.733333333333334e-06,
+      "loss": 0.0397,
+      "step": 1025
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 9.720512820512822e-06,
+      "loss": 0.0451,
+      "step": 1050
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 9.707692307692308e-06,
+      "loss": 0.0428,
+      "step": 1075
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 9.694871794871794e-06,
+      "loss": 0.0488,
+      "step": 1100
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 9.682051282051283e-06,
+      "loss": 0.0455,
+      "step": 1125
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.66923076923077e-06,
+      "loss": 0.0325,
+      "step": 1150
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 9.656410256410257e-06,
+      "loss": 0.0215,
+      "step": 1175
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 9.643589743589743e-06,
+      "loss": 0.0243,
+      "step": 1200
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 9.630769230769232e-06,
+      "loss": 0.0224,
+      "step": 1225
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 9.617948717948719e-06,
+      "loss": 0.024,
+      "step": 1250
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 9.605128205128206e-06,
+      "loss": 0.0253,
+      "step": 1275
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 9.592307692307692e-06,
+      "loss": 0.0266,
+      "step": 1300
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 9.579487179487181e-06,
+      "loss": 0.0244,
+      "step": 1325
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 9.566666666666668e-06,
+      "loss": 0.0244,
+      "step": 1350
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 9.553846153846155e-06,
+      "loss": 0.0271,
+      "step": 1375
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 9.541025641025641e-06,
+      "loss": 0.0286,
+      "step": 1400
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 9.52820512820513e-06,
+      "loss": 0.0245,
+      "step": 1425
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 9.515384615384617e-06,
+      "loss": 0.0153,
+      "step": 1450
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 9.502564102564103e-06,
+      "loss": 0.0152,
+      "step": 1475
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 9.48974358974359e-06,
+      "loss": 0.0164,
+      "step": 1500
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 9.476923076923079e-06,
+      "loss": 0.0153,
+      "step": 1525
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.464102564102566e-06,
+      "loss": 0.0159,
+      "step": 1550
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 9.451282051282052e-06,
+      "loss": 0.0148,
+      "step": 1575
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 9.43846153846154e-06,
+      "loss": 0.0192,
+      "step": 1600
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 9.425641025641026e-06,
+      "loss": 0.0147,
+      "step": 1625
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 9.412820512820515e-06,
+      "loss": 0.0189,
+      "step": 1650
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 9.4e-06,
+      "loss": 0.0177,
+      "step": 1675
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 9.387179487179488e-06,
+      "loss": 0.0185,
+      "step": 1700
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 9.374358974358975e-06,
+      "loss": 0.0106,
+      "step": 1725
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 9.361538461538462e-06,
+      "loss": 0.0101,
+      "step": 1750
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 9.348717948717949e-06,
+      "loss": 0.0106,
+      "step": 1775
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 9.335897435897436e-06,
+      "loss": 0.0092,
+      "step": 1800
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 9.323076923076924e-06,
+      "loss": 0.0109,
+      "step": 1825
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 9.310256410256411e-06,
+      "loss": 0.0114,
+      "step": 1850
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 9.297435897435898e-06,
+      "loss": 0.0124,
+      "step": 1875
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.284615384615385e-06,
+      "loss": 0.0126,
+      "step": 1900
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.271794871794873e-06,
+      "loss": 0.0119,
+      "step": 1925
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.25897435897436e-06,
+      "loss": 0.0099,
+      "step": 1950
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.246153846153847e-06,
+      "loss": 0.0143,
+      "step": 1975
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 9.233333333333334e-06,
+      "loss": 0.0093,
+      "step": 2000
+    },
+    {
+      "epoch": 7.05,
+      "eval_loss": 0.24930579960346222,
+      "eval_runtime": 2307.1499,
+      "eval_samples_per_second": 3.932,
+      "eval_steps_per_second": 0.246,
+      "eval_wer": 6.274438931423126,
+      "step": 2000
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 9.220512820512822e-06,
+      "loss": 0.0074,
+      "step": 2025
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 9.207692307692309e-06,
+      "loss": 0.0079,
+      "step": 2050
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 9.194871794871796e-06,
+      "loss": 0.0062,
+      "step": 2075
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 9.182051282051283e-06,
+      "loss": 0.0081,
+      "step": 2100
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 9.169230769230771e-06,
+      "loss": 0.0095,
+      "step": 2125
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 9.156410256410258e-06,
+      "loss": 0.0086,
+      "step": 2150
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 9.143589743589745e-06,
+      "loss": 0.0092,
+      "step": 2175
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 9.130769230769232e-06,
+      "loss": 0.0095,
+      "step": 2200
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 9.117948717948718e-06,
+      "loss": 0.0109,
+      "step": 2225
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 9.105128205128207e-06,
+      "loss": 0.0091,
+      "step": 2250
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 9.092307692307692e-06,
+      "loss": 0.0082,
+      "step": 2275
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 9.07948717948718e-06,
+      "loss": 0.0058,
+      "step": 2300
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 9.066666666666667e-06,
+      "loss": 0.0074,
+      "step": 2325
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 9.053846153846154e-06,
+      "loss": 0.0072,
+      "step": 2350
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 9.041025641025641e-06,
+      "loss": 0.0058,
+      "step": 2375
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 9.028205128205128e-06,
+      "loss": 0.0063,
+      "step": 2400
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 9.015384615384616e-06,
+      "loss": 0.0074,
+      "step": 2425
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 9.002564102564103e-06,
+      "loss": 0.0073,
+      "step": 2450
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 8.98974358974359e-06,
+      "loss": 0.0074,
+      "step": 2475
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 8.976923076923077e-06,
+      "loss": 0.0085,
+      "step": 2500
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 8.964102564102565e-06,
+      "loss": 0.0064,
+      "step": 2525
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 8.951282051282052e-06,
+      "loss": 0.0066,
+      "step": 2550
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 8.938461538461539e-06,
+      "loss": 0.0052,
+      "step": 2575
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.925641025641026e-06,
+      "loss": 0.0066,
+      "step": 2600
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 8.912820512820514e-06,
+      "loss": 0.0082,
+      "step": 2625
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 8.900000000000001e-06,
+      "loss": 0.006,
+      "step": 2650
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 8.887179487179488e-06,
+      "loss": 0.0057,
+      "step": 2675
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 8.874358974358975e-06,
+      "loss": 0.0072,
+      "step": 2700
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 8.861538461538463e-06,
+      "loss": 0.0079,
+      "step": 2725
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.84871794871795e-06,
+      "loss": 0.0056,
+      "step": 2750
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 8.835897435897437e-06,
+      "loss": 0.0055,
+      "step": 2775
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 8.823076923076924e-06,
+      "loss": 0.0064,
+      "step": 2800
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 8.81025641025641e-06,
+      "loss": 0.0073,
+      "step": 2825
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 8.7974358974359e-06,
+      "loss": 0.0053,
+      "step": 2850
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 8.784615384615386e-06,
+      "loss": 0.0062,
+      "step": 2875
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 8.771794871794873e-06,
+      "loss": 0.0065,
+      "step": 2900
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 8.75897435897436e-06,
+      "loss": 0.0064,
+      "step": 2925
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 8.746153846153846e-06,
+      "loss": 0.0069,
+      "step": 2950
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 8.733333333333333e-06,
+      "loss": 0.0074,
+      "step": 2975
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 8.72051282051282e-06,
+      "loss": 0.0049,
+      "step": 3000
+    },
+    {
+      "epoch": 10.58,
+      "eval_loss": 0.2656032145023346,
+      "eval_runtime": 2304.9596,
+      "eval_samples_per_second": 3.936,
+      "eval_steps_per_second": 0.246,
+      "eval_wer": 6.350014786580357,
+      "step": 3000
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 8.707692307692309e-06,
+      "loss": 0.0054,
+      "step": 3025
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 8.694871794871795e-06,
+      "loss": 0.0056,
+      "step": 3050
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 8.682051282051282e-06,
+      "loss": 0.0056,
+      "step": 3075
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 8.669230769230769e-06,
+      "loss": 0.005,
+      "step": 3100
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 8.656410256410258e-06,
+      "loss": 0.0048,
+      "step": 3125
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 8.643589743589744e-06,
+      "loss": 0.0038,
+      "step": 3150
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 8.630769230769231e-06,
+      "loss": 0.0041,
+      "step": 3175
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 8.617948717948718e-06,
+      "loss": 0.0037,
+      "step": 3200
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 8.605128205128207e-06,
+      "loss": 0.0038,
+      "step": 3225
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 8.592307692307693e-06,
+      "loss": 0.0035,
+      "step": 3250
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 8.57948717948718e-06,
+      "loss": 0.0042,
+      "step": 3275
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 8.566666666666667e-06,
+      "loss": 0.0048,
+      "step": 3300
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 8.553846153846156e-06,
+      "loss": 0.0051,
+      "step": 3325
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 8.541025641025642e-06,
+      "loss": 0.0067,
+      "step": 3350
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 8.52820512820513e-06,
+      "loss": 0.0058,
+      "step": 3375
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 8.515384615384616e-06,
+      "loss": 0.0052,
+      "step": 3400
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 8.502564102564103e-06,
+      "loss": 0.0043,
+      "step": 3425
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 8.489743589743591e-06,
+      "loss": 0.0033,
+      "step": 3450
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 8.476923076923078e-06,
+      "loss": 0.004,
+      "step": 3475
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 8.464102564102565e-06,
+      "loss": 0.0047,
+      "step": 3500
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 8.451282051282052e-06,
+      "loss": 0.0043,
+      "step": 3525
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 8.438461538461539e-06,
+      "loss": 0.0037,
+      "step": 3550
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 8.425641025641026e-06,
+      "loss": 0.0024,
+      "step": 3575
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 8.412820512820512e-06,
+      "loss": 0.0048,
+      "step": 3600
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.0053,
+      "step": 3625
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 8.387179487179488e-06,
+      "loss": 0.0043,
+      "step": 3650
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 8.374358974358975e-06,
+      "loss": 0.0045,
+      "step": 3675
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 8.361538461538461e-06,
+      "loss": 0.0044,
+      "step": 3700
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 8.34871794871795e-06,
+      "loss": 0.008,
+      "step": 3725
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 8.335897435897437e-06,
+      "loss": 0.0047,
+      "step": 3750
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 8.323076923076924e-06,
+      "loss": 0.0051,
+      "step": 3775
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 8.31025641025641e-06,
+      "loss": 0.0047,
+      "step": 3800
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 8.297435897435899e-06,
+      "loss": 0.0038,
+      "step": 3825
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 8.284615384615386e-06,
+      "loss": 0.0049,
+      "step": 3850
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 8.271794871794873e-06,
+      "loss": 0.0037,
+      "step": 3875
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 8.25897435897436e-06,
+      "loss": 0.0029,
+      "step": 3900
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 8.246153846153848e-06,
+      "loss": 0.0029,
+      "step": 3925
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 8.233333333333335e-06,
+      "loss": 0.0048,
+      "step": 3950
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 8.220512820512822e-06,
+      "loss": 0.0059,
+      "step": 3975
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 8.207692307692308e-06,
+      "loss": 0.0031,
+      "step": 4000
+    },
+    {
+      "epoch": 14.11,
+      "eval_loss": 0.285030722618103,
+      "eval_runtime": 2298.8806,
+      "eval_samples_per_second": 3.946,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.425590641737587,
+      "step": 4000
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 8.194871794871795e-06,
+      "loss": 0.0032,
+      "step": 4025
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 8.182051282051284e-06,
+      "loss": 0.0029,
+      "step": 4050
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.16923076923077e-06,
+      "loss": 0.0024,
+      "step": 4075
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.156410256410257e-06,
+      "loss": 0.0027,
+      "step": 4100
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.143589743589744e-06,
+      "loss": 0.0045,
+      "step": 4125
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.130769230769231e-06,
+      "loss": 0.0033,
+      "step": 4150
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.117948717948718e-06,
+      "loss": 0.0043,
+      "step": 4175
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.105128205128205e-06,
+      "loss": 0.0043,
+      "step": 4200
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 8.092307692307693e-06,
+      "loss": 0.0042,
+      "step": 4225
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 8.07948717948718e-06,
+      "loss": 0.0039,
+      "step": 4250
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 8.066666666666667e-06,
+      "loss": 0.0042,
+      "step": 4275
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 8.053846153846154e-06,
+      "loss": 0.0051,
+      "step": 4300
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 8.041025641025642e-06,
+      "loss": 0.0034,
+      "step": 4325
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 8.028205128205129e-06,
+      "loss": 0.0045,
+      "step": 4350
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 8.015384615384616e-06,
+      "loss": 0.0035,
+      "step": 4375
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 8.002564102564103e-06,
+      "loss": 0.0037,
+      "step": 4400
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 7.989743589743591e-06,
+      "loss": 0.006,
+      "step": 4425
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 7.976923076923078e-06,
+      "loss": 0.0035,
+      "step": 4450
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 7.964102564102565e-06,
+      "loss": 0.0038,
+      "step": 4475
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 7.951282051282052e-06,
+      "loss": 0.0051,
+      "step": 4500
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 7.93846153846154e-06,
+      "loss": 0.0038,
+      "step": 4525
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 7.925641025641027e-06,
+      "loss": 0.003,
+      "step": 4550
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 7.912820512820514e-06,
+      "loss": 0.0031,
+      "step": 4575
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 7.9e-06,
+      "loss": 0.0031,
+      "step": 4600
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 7.887179487179487e-06,
+      "loss": 0.0031,
+      "step": 4625
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 7.874358974358976e-06,
+      "loss": 0.0031,
+      "step": 4650
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 7.861538461538463e-06,
+      "loss": 0.0024,
+      "step": 4675
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 7.84871794871795e-06,
+      "loss": 0.002,
+      "step": 4700
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 7.835897435897436e-06,
+      "loss": 0.0028,
+      "step": 4725
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 7.823076923076923e-06,
+      "loss": 0.0023,
+      "step": 4750
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.81025641025641e-06,
+      "loss": 0.0027,
+      "step": 4775
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.797435897435897e-06,
+      "loss": 0.003,
+      "step": 4800
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.784615384615385e-06,
+      "loss": 0.0025,
+      "step": 4825
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.771794871794872e-06,
+      "loss": 0.0035,
+      "step": 4850
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.758974358974359e-06,
+      "loss": 0.0034,
+      "step": 4875
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 7.746153846153846e-06,
+      "loss": 0.0029,
+      "step": 4900
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 7.733333333333334e-06,
+      "loss": 0.0021,
+      "step": 4925
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 7.720512820512821e-06,
+      "loss": 0.0037,
+      "step": 4950
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 7.707692307692308e-06,
+      "loss": 0.0046,
+      "step": 4975
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 7.694871794871795e-06,
+      "loss": 0.0029,
+      "step": 5000
+    },
+    {
+      "epoch": 17.64,
+      "eval_loss": 0.3012606203556061,
+      "eval_runtime": 2293.1472,
+      "eval_samples_per_second": 3.956,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.583314165543983,
+      "step": 5000
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 7.682051282051283e-06,
+      "loss": 0.0033,
+      "step": 5025
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 7.66923076923077e-06,
+      "loss": 0.0036,
+      "step": 5050
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 7.656410256410257e-06,
+      "loss": 0.0034,
+      "step": 5075
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 7.643589743589744e-06,
+      "loss": 0.0042,
+      "step": 5100
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 7.630769230769232e-06,
+      "loss": 0.0032,
+      "step": 5125
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 7.617948717948719e-06,
+      "loss": 0.002,
+      "step": 5150
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 7.605128205128206e-06,
+      "loss": 0.0025,
+      "step": 5175
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 7.592307692307693e-06,
+      "loss": 0.0027,
+      "step": 5200
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 7.5794871794871805e-06,
+      "loss": 0.0024,
+      "step": 5225
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 7.566666666666667e-06,
+      "loss": 0.0039,
+      "step": 5250
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 7.553846153846155e-06,
+      "loss": 0.0035,
+      "step": 5275
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 7.541025641025642e-06,
+      "loss": 0.0017,
+      "step": 5300
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 7.528205128205129e-06,
+      "loss": 0.0018,
+      "step": 5325
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 7.5153846153846155e-06,
+      "loss": 0.0026,
+      "step": 5350
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 7.502564102564102e-06,
+      "loss": 0.004,
+      "step": 5375
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 7.48974358974359e-06,
+      "loss": 0.0022,
+      "step": 5400
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 7.476923076923077e-06,
+      "loss": 0.0026,
+      "step": 5425
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 7.4641025641025645e-06,
+      "loss": 0.0019,
+      "step": 5450
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 7.451282051282051e-06,
+      "loss": 0.0033,
+      "step": 5475
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 7.438461538461539e-06,
+      "loss": 0.0024,
+      "step": 5500
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 7.425641025641026e-06,
+      "loss": 0.0021,
+      "step": 5525
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 7.4128205128205135e-06,
+      "loss": 0.0048,
+      "step": 5550
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 7.4e-06,
+      "loss": 0.0039,
+      "step": 5575
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 7.387179487179488e-06,
+      "loss": 0.0038,
+      "step": 5600
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 7.374358974358975e-06,
+      "loss": 0.0039,
+      "step": 5625
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 7.362051282051282e-06,
+      "loss": 0.0049,
+      "step": 5650
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 7.34923076923077e-06,
+      "loss": 0.0046,
+      "step": 5675
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 7.336410256410257e-06,
+      "loss": 0.0035,
+      "step": 5700
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 7.323589743589744e-06,
+      "loss": 0.0042,
+      "step": 5725
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 7.310769230769231e-06,
+      "loss": 0.0021,
+      "step": 5750
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 7.297948717948719e-06,
+      "loss": 0.003,
+      "step": 5775
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 7.285128205128206e-06,
+      "loss": 0.004,
+      "step": 5800
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 7.272307692307693e-06,
+      "loss": 0.0027,
+      "step": 5825
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 7.25948717948718e-06,
+      "loss": 0.0031,
+      "step": 5850
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 7.246666666666667e-06,
+      "loss": 0.0029,
+      "step": 5875
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 7.233846153846155e-06,
+      "loss": 0.0026,
+      "step": 5900
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 7.2210256410256415e-06,
+      "loss": 0.0026,
+      "step": 5925
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 7.208205128205129e-06,
+      "loss": 0.0041,
+      "step": 5950
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 7.195384615384616e-06,
+      "loss": 0.0042,
+      "step": 5975
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 7.182564102564104e-06,
+      "loss": 0.0053,
+      "step": 6000
+    },
+    {
+      "epoch": 21.16,
+      "eval_loss": 0.2810451090335846,
+      "eval_runtime": 2294.5608,
+      "eval_samples_per_second": 3.954,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.796898104031808,
+      "step": 6000
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 7.1697435897435905e-06,
+      "loss": 0.0026,
+      "step": 6025
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 7.156923076923078e-06,
+      "loss": 0.0043,
+      "step": 6050
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 7.144102564102565e-06,
+      "loss": 0.003,
+      "step": 6075
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 7.131282051282053e-06,
+      "loss": 0.0026,
+      "step": 6100
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 7.118461538461539e-06,
+      "loss": 0.0031,
+      "step": 6125
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 7.1056410256410255e-06,
+      "loss": 0.0026,
+      "step": 6150
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 7.092820512820513e-06,
+      "loss": 0.0031,
+      "step": 6175
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 7.08e-06,
+      "loss": 0.003,
+      "step": 6200
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 7.067179487179488e-06,
+      "loss": 0.003,
+      "step": 6225
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 7.0543589743589745e-06,
+      "loss": 0.0021,
+      "step": 6250
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 7.041538461538462e-06,
+      "loss": 0.002,
+      "step": 6275
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 7.028717948717949e-06,
+      "loss": 0.0016,
+      "step": 6300
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 7.015897435897437e-06,
+      "loss": 0.003,
+      "step": 6325
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 7.0030769230769235e-06,
+      "loss": 0.0025,
+      "step": 6350
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 6.990256410256411e-06,
+      "loss": 0.0021,
+      "step": 6375
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 6.977435897435898e-06,
+      "loss": 0.0023,
+      "step": 6400
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 6.964615384615386e-06,
+      "loss": 0.0027,
+      "step": 6425
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 6.9517948717948725e-06,
+      "loss": 0.0024,
+      "step": 6450
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 6.938974358974359e-06,
+      "loss": 0.0024,
+      "step": 6475
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 6.926153846153847e-06,
+      "loss": 0.0022,
+      "step": 6500
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 6.913333333333334e-06,
+      "loss": 0.0036,
+      "step": 6525
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 6.9005128205128214e-06,
+      "loss": 0.0016,
+      "step": 6550
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 6.887692307692308e-06,
+      "loss": 0.002,
+      "step": 6575
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 6.874871794871796e-06,
+      "loss": 0.0037,
+      "step": 6600
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 6.862051282051283e-06,
+      "loss": 0.004,
+      "step": 6625
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 6.8492307692307704e-06,
+      "loss": 0.0038,
+      "step": 6650
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 6.836410256410257e-06,
+      "loss": 0.0034,
+      "step": 6675
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 6.823589743589745e-06,
+      "loss": 0.0024,
+      "step": 6700
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 6.810769230769231e-06,
+      "loss": 0.0035,
+      "step": 6725
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 6.797948717948718e-06,
+      "loss": 0.0025,
+      "step": 6750
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 6.785128205128205e-06,
+      "loss": 0.0037,
+      "step": 6775
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 6.772307692307692e-06,
+      "loss": 0.0033,
+      "step": 6800
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 6.75948717948718e-06,
+      "loss": 0.0023,
+      "step": 6825
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 6.746666666666667e-06,
+      "loss": 0.003,
+      "step": 6850
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 6.733846153846154e-06,
+      "loss": 0.002,
+      "step": 6875
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 6.721025641025641e-06,
+      "loss": 0.0019,
+      "step": 6900
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 6.708205128205129e-06,
+      "loss": 0.0016,
+      "step": 6925
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 6.695384615384616e-06,
+      "loss": 0.0035,
+      "step": 6950
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 6.682564102564103e-06,
+      "loss": 0.004,
+      "step": 6975
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 6.66974358974359e-06,
+      "loss": 0.0033,
+      "step": 7000
+    },
+    {
+      "epoch": 24.69,
+      "eval_loss": 0.3072134852409363,
+      "eval_runtime": 2292.4759,
+      "eval_samples_per_second": 3.957,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.890546446291854,
+      "step": 7000
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 6.656923076923078e-06,
+      "loss": 0.0042,
+      "step": 7025
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 6.644102564102565e-06,
+      "loss": 0.0031,
+      "step": 7050
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 6.6312820512820515e-06,
+      "loss": 0.0039,
+      "step": 7075
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 6.618461538461539e-06,
+      "loss": 0.0027,
+      "step": 7100
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 6.605641025641026e-06,
+      "loss": 0.0033,
+      "step": 7125
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 6.592820512820514e-06,
+      "loss": 0.0018,
+      "step": 7150
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 0.0025,
+      "step": 7175
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 6.567179487179488e-06,
+      "loss": 0.0022,
+      "step": 7200
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 6.554358974358975e-06,
+      "loss": 0.0018,
+      "step": 7225
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 6.541538461538463e-06,
+      "loss": 0.0028,
+      "step": 7250
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 6.5287179487179495e-06,
+      "loss": 0.0018,
+      "step": 7275
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 6.515897435897437e-06,
+      "loss": 0.0017,
+      "step": 7300
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 6.503076923076923e-06,
+      "loss": 0.0016,
+      "step": 7325
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 6.49025641025641e-06,
+      "loss": 0.0028,
+      "step": 7350
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 6.477435897435898e-06,
+      "loss": 0.0018,
+      "step": 7375
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 6.4646153846153845e-06,
+      "loss": 0.0013,
+      "step": 7400
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 6.451794871794872e-06,
+      "loss": 0.0015,
+      "step": 7425
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 6.438974358974359e-06,
+      "loss": 0.0015,
+      "step": 7450
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 6.426153846153847e-06,
+      "loss": 0.002,
+      "step": 7475
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 6.4133333333333335e-06,
+      "loss": 0.0018,
+      "step": 7500
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 6.400512820512821e-06,
+      "loss": 0.0027,
+      "step": 7525
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 6.387692307692308e-06,
+      "loss": 0.0008,
+      "step": 7550
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 6.374871794871796e-06,
+      "loss": 0.0011,
+      "step": 7575
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 6.3620512820512825e-06,
+      "loss": 0.0028,
+      "step": 7600
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 6.34923076923077e-06,
+      "loss": 0.0019,
+      "step": 7625
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 6.336410256410257e-06,
+      "loss": 0.0025,
+      "step": 7650
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 6.323589743589744e-06,
+      "loss": 0.0013,
+      "step": 7675
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 6.3107692307692315e-06,
+      "loss": 0.001,
+      "step": 7700
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 6.297948717948718e-06,
+      "loss": 0.0008,
+      "step": 7725
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 6.285128205128206e-06,
+      "loss": 0.0014,
+      "step": 7750
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 6.272307692307693e-06,
+      "loss": 0.0009,
+      "step": 7775
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 6.2594871794871805e-06,
+      "loss": 0.0018,
+      "step": 7800
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 6.246666666666667e-06,
+      "loss": 0.0012,
+      "step": 7825
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 6.233846153846155e-06,
+      "loss": 0.0011,
+      "step": 7850
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 6.221025641025642e-06,
+      "loss": 0.0016,
+      "step": 7875
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 6.2082051282051294e-06,
+      "loss": 0.0019,
+      "step": 7900
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 6.195384615384616e-06,
+      "loss": 0.0024,
+      "step": 7925
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 6.182564102564102e-06,
+      "loss": 0.0015,
+      "step": 7950
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 6.16974358974359e-06,
+      "loss": 0.0011,
+      "step": 7975
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 6.156923076923077e-06,
+      "loss": 0.0007,
+      "step": 8000
+    },
+    {
+      "epoch": 28.22,
+      "eval_loss": 0.32103443145751953,
+      "eval_runtime": 2300.2969,
+      "eval_samples_per_second": 3.944,
+      "eval_steps_per_second": 0.246,
+      "eval_wer": 6.731179969112476,
+      "step": 8000
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 6.144102564102564e-06,
+      "loss": 0.0011,
+      "step": 8025
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 6.131282051282051e-06,
+      "loss": 0.0011,
+      "step": 8050
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 6.118461538461539e-06,
+      "loss": 0.0009,
+      "step": 8075
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 6.105641025641026e-06,
+      "loss": 0.0011,
+      "step": 8100
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 6.092820512820513e-06,
+      "loss": 0.0005,
+      "step": 8125
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 6.08e-06,
+      "loss": 0.0016,
+      "step": 8150
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 6.067179487179488e-06,
+      "loss": 0.0013,
+      "step": 8175
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 6.054358974358975e-06,
+      "loss": 0.0014,
+      "step": 8200
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 6.041538461538462e-06,
+      "loss": 0.0013,
+      "step": 8225
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 6.028717948717949e-06,
+      "loss": 0.0006,
+      "step": 8250
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 6.015897435897436e-06,
+      "loss": 0.0005,
+      "step": 8275
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 6.003076923076924e-06,
+      "loss": 0.0004,
+      "step": 8300
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 5.9902564102564105e-06,
+      "loss": 0.0004,
+      "step": 8325
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 5.977435897435898e-06,
+      "loss": 0.0006,
+      "step": 8350
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 5.964615384615385e-06,
+      "loss": 0.0015,
+      "step": 8375
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 5.951794871794873e-06,
+      "loss": 0.0018,
+      "step": 8400
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 5.9389743589743595e-06,
+      "loss": 0.0022,
+      "step": 8425
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 5.926153846153847e-06,
+      "loss": 0.0028,
+      "step": 8450
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 5.913333333333334e-06,
+      "loss": 0.002,
+      "step": 8475
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 5.900512820512822e-06,
+      "loss": 0.0012,
+      "step": 8500
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 5.8876923076923085e-06,
+      "loss": 0.0022,
+      "step": 8525
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 5.8748717948717945e-06,
+      "loss": 0.002,
+      "step": 8550
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 5.862051282051282e-06,
+      "loss": 0.001,
+      "step": 8575
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 5.849230769230769e-06,
+      "loss": 0.001,
+      "step": 8600
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 5.836410256410257e-06,
+      "loss": 0.0021,
+      "step": 8625
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 5.8235897435897435e-06,
+      "loss": 0.0011,
+      "step": 8650
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 5.810769230769231e-06,
+      "loss": 0.0014,
+      "step": 8675
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 5.797948717948718e-06,
+      "loss": 0.0015,
+      "step": 8700
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 5.785128205128206e-06,
+      "loss": 0.0009,
+      "step": 8725
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 5.7723076923076925e-06,
+      "loss": 0.0023,
+      "step": 8750
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 5.75948717948718e-06,
+      "loss": 0.001,
+      "step": 8775
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 5.746666666666667e-06,
+      "loss": 0.0018,
+      "step": 8800
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 5.733846153846155e-06,
+      "loss": 0.0022,
+      "step": 8825
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 5.7210256410256415e-06,
+      "loss": 0.0016,
+      "step": 8850
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 5.708205128205129e-06,
+      "loss": 0.0015,
+      "step": 8875
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 5.695384615384616e-06,
+      "loss": 0.0012,
+      "step": 8900
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 5.682564102564103e-06,
+      "loss": 0.0016,
+      "step": 8925
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 5.6697435897435905e-06,
+      "loss": 0.0017,
+      "step": 8950
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 5.656923076923077e-06,
+      "loss": 0.0014,
+      "step": 8975
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 5.644102564102565e-06,
+      "loss": 0.0021,
+      "step": 9000
+    },
+    {
+      "epoch": 31.75,
+      "eval_loss": 0.33106353878974915,
+      "eval_runtime": 2298.0967,
+      "eval_samples_per_second": 3.948,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.992409555416818,
+      "step": 9000
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 5.631282051282052e-06,
+      "loss": 0.0023,
+      "step": 9025
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 5.6184615384615395e-06,
+      "loss": 0.0022,
+      "step": 9050
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 5.605641025641026e-06,
+      "loss": 0.0014,
+      "step": 9075
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 5.592820512820514e-06,
+      "loss": 0.0009,
+      "step": 9100
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 5.580000000000001e-06,
+      "loss": 0.0013,
+      "step": 9125
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 5.567179487179487e-06,
+      "loss": 0.002,
+      "step": 9150
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 5.5543589743589744e-06,
+      "loss": 0.0023,
+      "step": 9175
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 5.541538461538461e-06,
+      "loss": 0.0019,
+      "step": 9200
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 5.528717948717949e-06,
+      "loss": 0.0036,
+      "step": 9225
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 5.515897435897436e-06,
+      "loss": 0.002,
+      "step": 9250
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 5.5030769230769234e-06,
+      "loss": 0.0024,
+      "step": 9275
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 5.49025641025641e-06,
+      "loss": 0.0028,
+      "step": 9300
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 5.477435897435898e-06,
+      "loss": 0.002,
+      "step": 9325
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 5.464615384615385e-06,
+      "loss": 0.0027,
+      "step": 9350
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 5.451794871794872e-06,
+      "loss": 0.0029,
+      "step": 9375
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 5.438974358974359e-06,
+      "loss": 0.0029,
+      "step": 9400
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 5.426153846153847e-06,
+      "loss": 0.003,
+      "step": 9425
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 5.413333333333334e-06,
+      "loss": 0.0021,
+      "step": 9450
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 5.400512820512821e-06,
+      "loss": 0.0021,
+      "step": 9475
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 5.387692307692308e-06,
+      "loss": 0.0011,
+      "step": 9500
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 5.374871794871795e-06,
+      "loss": 0.0013,
+      "step": 9525
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 5.362051282051283e-06,
+      "loss": 0.0018,
+      "step": 9550
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 5.3492307692307696e-06,
+      "loss": 0.0021,
+      "step": 9575
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 5.336410256410257e-06,
+      "loss": 0.0023,
+      "step": 9600
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 5.323589743589744e-06,
+      "loss": 0.0017,
+      "step": 9625
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 5.310769230769232e-06,
+      "loss": 0.0017,
+      "step": 9650
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 5.2979487179487185e-06,
+      "loss": 0.0015,
+      "step": 9675
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 5.285128205128206e-06,
+      "loss": 0.0022,
+      "step": 9700
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 5.272307692307693e-06,
+      "loss": 0.0009,
+      "step": 9725
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 5.259487179487181e-06,
+      "loss": 0.0012,
+      "step": 9750
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 5.246666666666667e-06,
+      "loss": 0.0012,
+      "step": 9775
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 5.2338461538461535e-06,
+      "loss": 0.0008,
+      "step": 9800
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 5.221025641025641e-06,
+      "loss": 0.0016,
+      "step": 9825
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 5.208205128205128e-06,
+      "loss": 0.001,
+      "step": 9850
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 5.195384615384616e-06,
+      "loss": 0.0013,
+      "step": 9875
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 5.1825641025641025e-06,
+      "loss": 0.0014,
+      "step": 9900
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 5.16974358974359e-06,
+      "loss": 0.001,
+      "step": 9925
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 5.156923076923077e-06,
+      "loss": 0.0012,
+      "step": 9950
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 5.144102564102565e-06,
+      "loss": 0.0008,
+      "step": 9975
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 5.1312820512820515e-06,
+      "loss": 0.0006,
+      "step": 10000
+    },
+    {
+      "epoch": 35.27,
+      "eval_loss": 0.31875914335250854,
+      "eval_runtime": 2312.6787,
+      "eval_samples_per_second": 3.923,
+      "eval_steps_per_second": 0.245,
+      "eval_wer": 6.729537015739494,
+      "step": 10000
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 5.118461538461539e-06,
+      "loss": 0.0008,
+      "step": 10025
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 5.105641025641026e-06,
+      "loss": 0.0023,
+      "step": 10050
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 5.092820512820514e-06,
+      "loss": 0.0007,
+      "step": 10075
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 5.0800000000000005e-06,
+      "loss": 0.0018,
+      "step": 10100
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 5.067179487179487e-06,
+      "loss": 0.0012,
+      "step": 10125
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 5.054358974358975e-06,
+      "loss": 0.0005,
+      "step": 10150
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 5.041538461538462e-06,
+      "loss": 0.0004,
+      "step": 10175
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 5.0287179487179495e-06,
+      "loss": 0.0006,
+      "step": 10200
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 5.015897435897436e-06,
+      "loss": 0.0007,
+      "step": 10225
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 5.003076923076924e-06,
+      "loss": 0.0003,
+      "step": 10250
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 4.990256410256411e-06,
+      "loss": 0.0003,
+      "step": 10275
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 4.977435897435898e-06,
+      "loss": 0.0002,
+      "step": 10300
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 4.964615384615385e-06,
+      "loss": 0.0002,
+      "step": 10325
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 4.951794871794872e-06,
+      "loss": 0.0002,
+      "step": 10350
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 4.93897435897436e-06,
+      "loss": 0.0003,
+      "step": 10375
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 4.926153846153847e-06,
+      "loss": 0.0009,
+      "step": 10400
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 4.9133333333333334e-06,
+      "loss": 0.001,
+      "step": 10425
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 4.900512820512821e-06,
+      "loss": 0.0004,
+      "step": 10450
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 4.887692307692308e-06,
+      "loss": 0.0004,
+      "step": 10475
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 4.874871794871796e-06,
+      "loss": 0.0009,
+      "step": 10500
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 4.8620512820512824e-06,
+      "loss": 0.0005,
+      "step": 10525
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 4.849230769230769e-06,
+      "loss": 0.0009,
+      "step": 10550
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 4.836410256410257e-06,
+      "loss": 0.0002,
+      "step": 10575
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 4.823589743589744e-06,
+      "loss": 0.0002,
+      "step": 10600
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 4.8107692307692314e-06,
+      "loss": 0.0002,
+      "step": 10625
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 4.797948717948718e-06,
+      "loss": 0.0002,
+      "step": 10650
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 4.785128205128206e-06,
+      "loss": 0.0002,
+      "step": 10675
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 4.772307692307693e-06,
+      "loss": 0.0003,
+      "step": 10700
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 4.7594871794871796e-06,
+      "loss": 0.0002,
+      "step": 10725
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 4.746666666666667e-06,
+      "loss": 0.0002,
+      "step": 10750
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 4.733846153846154e-06,
+      "loss": 0.0001,
+      "step": 10775
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 4.721025641025642e-06,
+      "loss": 0.0001,
+      "step": 10800
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 4.7082051282051286e-06,
+      "loss": 0.0001,
+      "step": 10825
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 4.695384615384615e-06,
+      "loss": 0.0001,
+      "step": 10850
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 4.682564102564103e-06,
+      "loss": 0.0001,
+      "step": 10875
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 4.66974358974359e-06,
+      "loss": 0.0001,
+      "step": 10900
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 4.6569230769230776e-06,
+      "loss": 0.0002,
+      "step": 10925
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 4.644102564102564e-06,
+      "loss": 0.0001,
+      "step": 10950
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 4.631282051282052e-06,
+      "loss": 0.001,
+      "step": 10975
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 4.618461538461539e-06,
+      "loss": 0.0002,
+      "step": 11000
+    },
+    {
+      "epoch": 38.8,
+      "eval_loss": 0.33356669545173645,
+      "eval_runtime": 2302.5441,
+      "eval_samples_per_second": 3.94,
+      "eval_steps_per_second": 0.246,
+      "eval_wer": 6.658890020701212,
+      "step": 11000
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 4.605641025641026e-06,
+      "loss": 0.0002,
+      "step": 11025
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 4.592820512820513e-06,
+      "loss": 0.0002,
+      "step": 11050
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 4.58e-06,
+      "loss": 0.0001,
+      "step": 11075
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 4.567179487179488e-06,
+      "loss": 0.0001,
+      "step": 11100
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 4.554358974358975e-06,
+      "loss": 0.0007,
+      "step": 11125
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 4.5415384615384615e-06,
+      "loss": 0.0001,
+      "step": 11150
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 4.528717948717949e-06,
+      "loss": 0.0002,
+      "step": 11175
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 4.515897435897436e-06,
+      "loss": 0.0001,
+      "step": 11200
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 4.503076923076924e-06,
+      "loss": 0.0002,
+      "step": 11225
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 4.4902564102564105e-06,
+      "loss": 0.0003,
+      "step": 11250
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 4.477435897435898e-06,
+      "loss": 0.0006,
+      "step": 11275
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 4.464615384615385e-06,
+      "loss": 0.0009,
+      "step": 11300
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 4.451794871794872e-06,
+      "loss": 0.0008,
+      "step": 11325
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 4.4389743589743595e-06,
+      "loss": 0.0002,
+      "step": 11350
+    },
+    {
+      "epoch": 40.12,
+      "learning_rate": 4.426153846153846e-06,
+      "loss": 0.0002,
+      "step": 11375
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 4.413333333333334e-06,
+      "loss": 0.0011,
+      "step": 11400
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 4.400512820512821e-06,
+      "loss": 0.0002,
+      "step": 11425
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 4.387692307692308e-06,
+      "loss": 0.0003,
+      "step": 11450
+    },
+    {
+      "epoch": 40.48,
+      "learning_rate": 4.374871794871795e-06,
+      "loss": 0.0003,
+      "step": 11475
+    },
+    {
+      "epoch": 40.56,
+      "learning_rate": 4.362051282051282e-06,
+      "loss": 0.0003,
+      "step": 11500
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 4.34923076923077e-06,
+      "loss": 0.001,
+      "step": 11525
+    },
+    {
+      "epoch": 40.74,
+      "learning_rate": 4.336410256410257e-06,
+      "loss": 0.001,
+      "step": 11550
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 4.323589743589744e-06,
+      "loss": 0.0018,
+      "step": 11575
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 4.310769230769231e-06,
+      "loss": 0.0034,
+      "step": 11600
+    },
+    {
+      "epoch": 41.01,
+      "learning_rate": 4.297948717948718e-06,
+      "loss": 0.0023,
+      "step": 11625
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 4.285128205128206e-06,
+      "loss": 0.0007,
+      "step": 11650
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 4.2723076923076925e-06,
+      "loss": 0.0004,
+      "step": 11675
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 4.25948717948718e-06,
+      "loss": 0.0004,
+      "step": 11700
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 4.246666666666667e-06,
+      "loss": 0.0003,
+      "step": 11725
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 4.233846153846154e-06,
+      "loss": 0.0005,
+      "step": 11750
+    },
+    {
+      "epoch": 41.53,
+      "learning_rate": 4.2210256410256414e-06,
+      "loss": 0.0005,
+      "step": 11775
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 4.208205128205128e-06,
+      "loss": 0.0015,
+      "step": 11800
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 4.195384615384616e-06,
+      "loss": 0.0019,
+      "step": 11825
+    },
+    {
+      "epoch": 41.8,
+      "learning_rate": 4.182564102564103e-06,
+      "loss": 0.0021,
+      "step": 11850
+    },
+    {
+      "epoch": 41.89,
+      "learning_rate": 4.1697435897435904e-06,
+      "loss": 0.0011,
+      "step": 11875
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 4.156923076923077e-06,
+      "loss": 0.0007,
+      "step": 11900
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 4.144102564102564e-06,
+      "loss": 0.0003,
+      "step": 11925
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 4.131282051282052e-06,
+      "loss": 0.0005,
+      "step": 11950
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 4.118461538461539e-06,
+      "loss": 0.0013,
+      "step": 11975
+    },
+    {
+      "epoch": 42.33,
+      "learning_rate": 4.105641025641026e-06,
+      "loss": 0.0004,
+      "step": 12000
+    },
+    {
+      "epoch": 42.33,
+      "eval_loss": 0.3464820981025696,
+      "eval_runtime": 2300.3502,
+      "eval_samples_per_second": 3.944,
+      "eval_steps_per_second": 0.246,
+      "eval_wer": 6.90861893339467,
+      "step": 12000
+    },
+    {
+      "epoch": 42.42,
+      "learning_rate": 4.092820512820513e-06,
+      "loss": 0.0007,
+      "step": 12025
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 4.08e-06,
+      "loss": 0.0013,
+      "step": 12050
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 4.0671794871794876e-06,
+      "loss": 0.0018,
+      "step": 12075
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 4.054358974358974e-06,
+      "loss": 0.0007,
+      "step": 12100
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 4.041538461538462e-06,
+      "loss": 0.0021,
+      "step": 12125
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 4.028717948717949e-06,
+      "loss": 0.0022,
+      "step": 12150
+    },
+    {
+      "epoch": 42.95,
+      "learning_rate": 4.0158974358974366e-06,
+      "loss": 0.0016,
+      "step": 12175
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 4.003076923076923e-06,
+      "loss": 0.0032,
+      "step": 12200
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 3.990256410256411e-06,
+      "loss": 0.0008,
+      "step": 12225
+    },
+    {
+      "epoch": 43.21,
+      "learning_rate": 3.977435897435898e-06,
+      "loss": 0.0017,
+      "step": 12250
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 3.964615384615385e-06,
+      "loss": 0.0022,
+      "step": 12275
+    },
+    {
+      "epoch": 43.39,
+      "learning_rate": 3.951794871794872e-06,
+      "loss": 0.003,
+      "step": 12300
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 3.938974358974359e-06,
+      "loss": 0.0023,
+      "step": 12325
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 3.926153846153846e-06,
+      "loss": 0.0022,
+      "step": 12350
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 3.913333333333334e-06,
+      "loss": 0.0013,
+      "step": 12375
+    },
+    {
+      "epoch": 43.74,
+      "learning_rate": 3.9005128205128205e-06,
+      "loss": 0.0012,
+      "step": 12400
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 3.887692307692308e-06,
+      "loss": 0.0009,
+      "step": 12425
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 3.874871794871795e-06,
+      "loss": 0.0015,
+      "step": 12450
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 3.862051282051283e-06,
+      "loss": 0.002,
+      "step": 12475
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 3.8492307692307695e-06,
+      "loss": 0.001,
+      "step": 12500
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 3.836410256410257e-06,
+      "loss": 0.0008,
+      "step": 12525
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 3.823589743589744e-06,
+      "loss": 0.0004,
+      "step": 12550
+    },
+    {
+      "epoch": 44.36,
+      "learning_rate": 3.8107692307692313e-06,
+      "loss": 0.0008,
+      "step": 12575
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 3.7979487179487185e-06,
+      "loss": 0.0019,
+      "step": 12600
+    },
+    {
+      "epoch": 44.53,
+      "learning_rate": 3.7851282051282058e-06,
+      "loss": 0.0005,
+      "step": 12625
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 3.772307692307693e-06,
+      "loss": 0.0007,
+      "step": 12650
+    },
+    {
+      "epoch": 44.71,
+      "learning_rate": 3.7594871794871794e-06,
+      "loss": 0.0003,
+      "step": 12675
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 3.7466666666666667e-06,
+      "loss": 0.0005,
+      "step": 12700
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 3.733846153846154e-06,
+      "loss": 0.0008,
+      "step": 12725
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 3.721025641025641e-06,
+      "loss": 0.0009,
+      "step": 12750
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 3.7082051282051284e-06,
+      "loss": 0.0003,
+      "step": 12775
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 3.6953846153846156e-06,
+      "loss": 0.0006,
+      "step": 12800
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 3.682564102564103e-06,
+      "loss": 0.0004,
+      "step": 12825
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 3.66974358974359e-06,
+      "loss": 0.0008,
+      "step": 12850
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 3.6569230769230774e-06,
+      "loss": 0.0008,
+      "step": 12875
+    },
+    {
+      "epoch": 45.5,
+      "learning_rate": 3.6441025641025646e-06,
+      "loss": 0.0013,
+      "step": 12900
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 3.631282051282052e-06,
+      "loss": 0.0004,
+      "step": 12925
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 3.618461538461539e-06,
+      "loss": 0.001,
+      "step": 12950
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 3.6056410256410255e-06,
+      "loss": 0.0013,
+      "step": 12975
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 3.5928205128205128e-06,
+      "loss": 0.0004,
+      "step": 13000
+    },
+    {
+      "epoch": 45.86,
+      "eval_loss": 0.334043949842453,
+      "eval_runtime": 2299.1497,
+      "eval_samples_per_second": 3.946,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.992409555416818,
+      "step": 13000
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 3.58e-06,
+      "loss": 0.0004,
+      "step": 13025
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 3.5671794871794873e-06,
+      "loss": 0.0004,
+      "step": 13050
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 3.5543589743589745e-06,
+      "loss": 0.0002,
+      "step": 13075
+    },
+    {
+      "epoch": 46.21,
+      "learning_rate": 3.5415384615384618e-06,
+      "loss": 0.0006,
+      "step": 13100
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 3.528717948717949e-06,
+      "loss": 0.0005,
+      "step": 13125
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 3.5158974358974363e-06,
+      "loss": 0.0006,
+      "step": 13150
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 3.5030769230769235e-06,
+      "loss": 0.0002,
+      "step": 13175
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 3.4902564102564108e-06,
+      "loss": 0.0002,
+      "step": 13200
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 3.477435897435898e-06,
+      "loss": 0.0005,
+      "step": 13225
+    },
+    {
+      "epoch": 46.74,
+      "learning_rate": 3.4646153846153853e-06,
+      "loss": 0.0002,
+      "step": 13250
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 3.4517948717948717e-06,
+      "loss": 0.0006,
+      "step": 13275
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 3.438974358974359e-06,
+      "loss": 0.0005,
+      "step": 13300
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 3.426153846153846e-06,
+      "loss": 0.0009,
+      "step": 13325
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 3.4133333333333334e-06,
+      "loss": 0.0002,
+      "step": 13350
+    },
+    {
+      "epoch": 47.18,
+      "learning_rate": 3.4005128205128207e-06,
+      "loss": 0.0003,
+      "step": 13375
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 3.387692307692308e-06,
+      "loss": 0.0002,
+      "step": 13400
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 3.374871794871795e-06,
+      "loss": 0.0002,
+      "step": 13425
+    },
+    {
+      "epoch": 47.44,
+      "learning_rate": 3.3620512820512824e-06,
+      "loss": 0.0002,
+      "step": 13450
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 3.3492307692307696e-06,
+      "loss": 0.0006,
+      "step": 13475
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 3.336410256410257e-06,
+      "loss": 0.0004,
+      "step": 13500
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 3.323589743589744e-06,
+      "loss": 0.0003,
+      "step": 13525
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 3.3107692307692314e-06,
+      "loss": 0.0002,
+      "step": 13550
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 3.297948717948718e-06,
+      "loss": 0.0002,
+      "step": 13575
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 3.285128205128205e-06,
+      "loss": 0.0001,
+      "step": 13600
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 3.2723076923076923e-06,
+      "loss": 0.0002,
+      "step": 13625
+    },
+    {
+      "epoch": 48.15,
+      "learning_rate": 3.2594871794871795e-06,
+      "loss": 0.0001,
+      "step": 13650
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 3.2466666666666668e-06,
+      "loss": 0.0002,
+      "step": 13675
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 3.233846153846154e-06,
+      "loss": 0.0001,
+      "step": 13700
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 3.2210256410256413e-06,
+      "loss": 0.0001,
+      "step": 13725
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 3.2082051282051285e-06,
+      "loss": 0.0001,
+      "step": 13750
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 3.1953846153846158e-06,
+      "loss": 0.0001,
+      "step": 13775
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 3.182564102564103e-06,
+      "loss": 0.0001,
+      "step": 13800
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 3.1697435897435903e-06,
+      "loss": 0.0001,
+      "step": 13825
+    },
+    {
+      "epoch": 48.85,
+      "learning_rate": 3.1569230769230775e-06,
+      "loss": 0.0001,
+      "step": 13850
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 3.144102564102564e-06,
+      "loss": 0.0001,
+      "step": 13875
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 3.131282051282051e-06,
+      "loss": 0.0001,
+      "step": 13900
+    },
+    {
+      "epoch": 49.12,
+      "learning_rate": 3.1184615384615384e-06,
+      "loss": 0.0001,
+      "step": 13925
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 3.1056410256410257e-06,
+      "loss": 0.0001,
+      "step": 13950
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 3.092820512820513e-06,
+      "loss": 0.0001,
+      "step": 13975
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 3.08e-06,
+      "loss": 0.0001,
+      "step": 14000
+    },
+    {
+      "epoch": 49.38,
+      "eval_loss": 0.3607153296470642,
+      "eval_runtime": 2306.9954,
+      "eval_samples_per_second": 3.932,
+      "eval_steps_per_second": 0.246,
+      "eval_wer": 6.8198994512535736,
+      "step": 14000
+    },
+    {
+      "epoch": 49.47,
+      "learning_rate": 3.0671794871794874e-06,
+      "loss": 0.0001,
+      "step": 14025
+    },
+    {
+      "epoch": 49.56,
+      "learning_rate": 3.0543589743589747e-06,
+      "loss": 0.0001,
+      "step": 14050
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 3.041538461538462e-06,
+      "loss": 0.0001,
+      "step": 14075
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 3.028717948717949e-06,
+      "loss": 0.0001,
+      "step": 14100
+    },
+    {
+      "epoch": 49.82,
+      "learning_rate": 3.0158974358974364e-06,
+      "loss": 0.0001,
+      "step": 14125
+    },
+    {
+      "epoch": 49.91,
+      "learning_rate": 3.0030769230769236e-06,
+      "loss": 0.0001,
+      "step": 14150
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 2.99025641025641e-06,
+      "loss": 0.0001,
+      "step": 14175
+    },
+    {
+      "epoch": 50.09,
+      "learning_rate": 2.9774358974358973e-06,
+      "loss": 0.0001,
+      "step": 14200
+    },
+    {
+      "epoch": 50.18,
+      "learning_rate": 2.9646153846153845e-06,
+      "loss": 0.0001,
+      "step": 14225
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 2.951794871794872e-06,
+      "loss": 0.0001,
+      "step": 14250
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 2.938974358974359e-06,
+      "loss": 0.0001,
+      "step": 14275
+    },
+    {
+      "epoch": 50.44,
+      "learning_rate": 2.9261538461538463e-06,
+      "loss": 0.0001,
+      "step": 14300
+    },
+    {
+      "epoch": 50.53,
+      "learning_rate": 2.9133333333333335e-06,
+      "loss": 0.0001,
+      "step": 14325
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 2.9005128205128208e-06,
+      "loss": 0.0001,
+      "step": 14350
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 2.887692307692308e-06,
+      "loss": 0.0001,
+      "step": 14375
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 2.8748717948717953e-06,
+      "loss": 0.0001,
+      "step": 14400
+    },
+    {
+      "epoch": 50.88,
+      "learning_rate": 2.8620512820512825e-06,
+      "loss": 0.0001,
+      "step": 14425
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 2.8492307692307698e-06,
+      "loss": 0.0001,
+      "step": 14450
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 2.836410256410257e-06,
+      "loss": 0.0001,
+      "step": 14475
+    },
+    {
+      "epoch": 51.15,
+      "learning_rate": 2.8235897435897434e-06,
+      "loss": 0.0001,
+      "step": 14500
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 2.8107692307692307e-06,
+      "loss": 0.0001,
+      "step": 14525
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 2.797948717948718e-06,
+      "loss": 0.0001,
+      "step": 14550
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 2.785128205128205e-06,
+      "loss": 0.0001,
+      "step": 14575
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 2.7723076923076924e-06,
+      "loss": 0.0001,
+      "step": 14600
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 2.7594871794871797e-06,
+      "loss": 0.0001,
+      "step": 14625
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 2.746666666666667e-06,
+      "loss": 0.0001,
+      "step": 14650
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 2.733846153846154e-06,
+      "loss": 0.0001,
+      "step": 14675
+    },
+    {
+      "epoch": 51.85,
+      "learning_rate": 2.7210256410256414e-06,
+      "loss": 0.0001,
+      "step": 14700
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 2.7082051282051287e-06,
+      "loss": 0.0001,
+      "step": 14725
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 2.695384615384616e-06,
+      "loss": 0.0001,
+      "step": 14750
+    },
+    {
+      "epoch": 52.12,
+      "learning_rate": 2.682564102564103e-06,
+      "loss": 0.0001,
+      "step": 14775
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 2.6697435897435896e-06,
+      "loss": 0.0001,
+      "step": 14800
+    },
+    {
+      "epoch": 52.29,
+      "learning_rate": 2.656923076923077e-06,
+      "loss": 0.0001,
+      "step": 14825
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 2.644102564102564e-06,
+      "loss": 0.0001,
+      "step": 14850
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 2.6312820512820513e-06,
+      "loss": 0.0001,
+      "step": 14875
+    },
+    {
+      "epoch": 52.56,
+      "learning_rate": 2.6184615384615385e-06,
+      "loss": 0.0001,
+      "step": 14900
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 2.605641025641026e-06,
+      "loss": 0.0001,
+      "step": 14925
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 2.592820512820513e-06,
+      "loss": 0.0001,
+      "step": 14950
+    },
+    {
+      "epoch": 52.82,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 0.0001,
+      "step": 14975
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 2.5671794871794875e-06,
+      "loss": 0.0001,
+      "step": 15000
+    },
+    {
+      "epoch": 52.91,
+      "eval_loss": 0.37789127230644226,
+      "eval_runtime": 2291.031,
+      "eval_samples_per_second": 3.96,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.6112443728846975,
+      "step": 15000
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 2.5543589743589748e-06,
+      "loss": 0.0001,
+      "step": 15025
+    },
+    {
+      "epoch": 53.09,
+      "learning_rate": 2.541538461538462e-06,
+      "loss": 0.0001,
+      "step": 15050
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 2.5287179487179493e-06,
+      "loss": 0.0001,
+      "step": 15075
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 2.5158974358974357e-06,
+      "loss": 0.0001,
+      "step": 15100
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 2.503076923076923e-06,
+      "loss": 0.0001,
+      "step": 15125
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 2.4902564102564106e-06,
+      "loss": 0.0001,
+      "step": 15150
+    },
+    {
+      "epoch": 53.53,
+      "learning_rate": 2.4774358974358974e-06,
+      "loss": 0.0001,
+      "step": 15175
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 2.4646153846153847e-06,
+      "loss": 0.0001,
+      "step": 15200
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 2.451794871794872e-06,
+      "loss": 0.0001,
+      "step": 15225
+    },
+    {
+      "epoch": 53.79,
+      "learning_rate": 2.438974358974359e-06,
+      "loss": 0.0001,
+      "step": 15250
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 2.4261538461538464e-06,
+      "loss": 0.0001,
+      "step": 15275
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 2.4133333333333337e-06,
+      "loss": 0.0001,
+      "step": 15300
+    },
+    {
+      "epoch": 54.06,
+      "learning_rate": 2.4005128205128205e-06,
+      "loss": 0.0001,
+      "step": 15325
+    },
+    {
+      "epoch": 54.14,
+      "learning_rate": 2.3876923076923077e-06,
+      "loss": 0.0001,
+      "step": 15350
+    },
+    {
+      "epoch": 54.23,
+      "learning_rate": 2.374871794871795e-06,
+      "loss": 0.0001,
+      "step": 15375
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 2.3620512820512822e-06,
+      "loss": 0.0001,
+      "step": 15400
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 2.3492307692307695e-06,
+      "loss": 0.0001,
+      "step": 15425
+    },
+    {
+      "epoch": 54.5,
+      "learning_rate": 2.3364102564102567e-06,
+      "loss": 0.0001,
+      "step": 15450
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 2.3235897435897436e-06,
+      "loss": 0.0001,
+      "step": 15475
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 2.310769230769231e-06,
+      "loss": 0.0001,
+      "step": 15500
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 2.297948717948718e-06,
+      "loss": 0.0001,
+      "step": 15525
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 2.2851282051282053e-06,
+      "loss": 0.0001,
+      "step": 15550
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 2.2723076923076925e-06,
+      "loss": 0.0001,
+      "step": 15575
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 2.25948717948718e-06,
+      "loss": 0.0001,
+      "step": 15600
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 2.2466666666666666e-06,
+      "loss": 0.0001,
+      "step": 15625
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 2.233846153846154e-06,
+      "loss": 0.0001,
+      "step": 15650
+    },
+    {
+      "epoch": 55.29,
+      "learning_rate": 2.221025641025641e-06,
+      "loss": 0.0001,
+      "step": 15675
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 2.2082051282051284e-06,
+      "loss": 0.0001,
+      "step": 15700
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 2.1953846153846156e-06,
+      "loss": 0.0001,
+      "step": 15725
+    },
+    {
+      "epoch": 55.56,
+      "learning_rate": 2.182564102564103e-06,
+      "loss": 0.0001,
+      "step": 15750
+    },
+    {
+      "epoch": 55.64,
+      "learning_rate": 2.1697435897435897e-06,
+      "loss": 0.0001,
+      "step": 15775
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 2.156923076923077e-06,
+      "loss": 0.0001,
+      "step": 15800
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 2.144102564102564e-06,
+      "loss": 0.0001,
+      "step": 15825
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 2.1312820512820514e-06,
+      "loss": 0.0001,
+      "step": 15850
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 2.1184615384615387e-06,
+      "loss": 0.0001,
+      "step": 15875
+    },
+    {
+      "epoch": 56.08,
+      "learning_rate": 2.105641025641026e-06,
+      "loss": 0.0,
+      "step": 15900
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 2.092820512820513e-06,
+      "loss": 0.0,
+      "step": 15925
+    },
+    {
+      "epoch": 56.26,
+      "learning_rate": 2.08e-06,
+      "loss": 0.0,
+      "step": 15950
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 2.0671794871794872e-06,
+      "loss": 0.0,
+      "step": 15975
+    },
+    {
+      "epoch": 56.44,
+      "learning_rate": 2.0543589743589745e-06,
+      "loss": 0.0,
+      "step": 16000
+    },
+    {
+      "epoch": 56.44,
+      "eval_loss": 0.3884251117706299,
+      "eval_runtime": 2296.6001,
+      "eval_samples_per_second": 3.95,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.550455098084316,
+      "step": 16000
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 2.0415384615384617e-06,
+      "loss": 0.0,
+      "step": 16025
+    },
+    {
+      "epoch": 56.61,
+      "learning_rate": 2.028717948717949e-06,
+      "loss": 0.0001,
+      "step": 16050
+    },
+    {
+      "epoch": 56.7,
+      "learning_rate": 2.0158974358974362e-06,
+      "loss": 0.0001,
+      "step": 16075
+    },
+    {
+      "epoch": 56.79,
+      "learning_rate": 2.003076923076923e-06,
+      "loss": 0.0,
+      "step": 16100
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 1.9902564102564103e-06,
+      "loss": 0.0,
+      "step": 16125
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 1.9774358974358976e-06,
+      "loss": 0.0,
+      "step": 16150
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 1.964615384615385e-06,
+      "loss": 0.0,
+      "step": 16175
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 1.951794871794872e-06,
+      "loss": 0.0,
+      "step": 16200
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 1.9389743589743593e-06,
+      "loss": 0.0,
+      "step": 16225
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 1.926153846153846e-06,
+      "loss": 0.0,
+      "step": 16250
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 1.9133333333333334e-06,
+      "loss": 0.0,
+      "step": 16275
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 1.9005128205128206e-06,
+      "loss": 0.0,
+      "step": 16300
+    },
+    {
+      "epoch": 57.58,
+      "learning_rate": 1.8876923076923079e-06,
+      "loss": 0.0,
+      "step": 16325
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 1.8748717948717951e-06,
+      "loss": 0.0,
+      "step": 16350
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 1.8620512820512824e-06,
+      "loss": 0.0,
+      "step": 16375
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 1.8492307692307692e-06,
+      "loss": 0.0,
+      "step": 16400
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 1.8364102564102564e-06,
+      "loss": 0.0,
+      "step": 16425
+    },
+    {
+      "epoch": 58.02,
+      "learning_rate": 1.8235897435897437e-06,
+      "loss": 0.0,
+      "step": 16450
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 1.810769230769231e-06,
+      "loss": 0.0,
+      "step": 16475
+    },
+    {
+      "epoch": 58.2,
+      "learning_rate": 1.7979487179487182e-06,
+      "loss": 0.0,
+      "step": 16500
+    },
+    {
+      "epoch": 58.29,
+      "learning_rate": 1.7851282051282054e-06,
+      "loss": 0.0,
+      "step": 16525
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 1.7723076923076922e-06,
+      "loss": 0.0,
+      "step": 16550
+    },
+    {
+      "epoch": 58.47,
+      "learning_rate": 1.7594871794871795e-06,
+      "loss": 0.0,
+      "step": 16575
+    },
+    {
+      "epoch": 58.55,
+      "learning_rate": 1.7466666666666667e-06,
+      "loss": 0.0,
+      "step": 16600
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 1.733846153846154e-06,
+      "loss": 0.0,
+      "step": 16625
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 1.7210256410256412e-06,
+      "loss": 0.0,
+      "step": 16650
+    },
+    {
+      "epoch": 58.82,
+      "learning_rate": 1.7082051282051285e-06,
+      "loss": 0.0,
+      "step": 16675
+    },
+    {
+      "epoch": 58.91,
+      "learning_rate": 1.6953846153846153e-06,
+      "loss": 0.0,
+      "step": 16700
+    },
+    {
+      "epoch": 58.99,
+      "learning_rate": 1.6825641025641026e-06,
+      "loss": 0.0,
+      "step": 16725
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 1.6697435897435898e-06,
+      "loss": 0.0,
+      "step": 16750
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 1.656923076923077e-06,
+      "loss": 0.0,
+      "step": 16775
+    },
+    {
+      "epoch": 59.26,
+      "learning_rate": 1.6441025641025643e-06,
+      "loss": 0.0,
+      "step": 16800
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 1.6312820512820516e-06,
+      "loss": 0.0,
+      "step": 16825
+    },
+    {
+      "epoch": 59.44,
+      "learning_rate": 1.6184615384615384e-06,
+      "loss": 0.0,
+      "step": 16850
+    },
+    {
+      "epoch": 59.52,
+      "learning_rate": 1.6056410256410256e-06,
+      "loss": 0.0,
+      "step": 16875
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 1.5928205128205129e-06,
+      "loss": 0.0,
+      "step": 16900
+    },
+    {
+      "epoch": 59.7,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 0.0,
+      "step": 16925
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 1.5671794871794874e-06,
+      "loss": 0.0,
+      "step": 16950
+    },
+    {
+      "epoch": 59.88,
+      "learning_rate": 1.5543589743589746e-06,
+      "loss": 0.0,
+      "step": 16975
+    },
+    {
+      "epoch": 59.96,
+      "learning_rate": 1.5415384615384614e-06,
+      "loss": 0.0,
+      "step": 17000
+    },
+    {
+      "epoch": 59.96,
+      "eval_loss": 0.39664292335510254,
+      "eval_runtime": 2297.1123,
+      "eval_samples_per_second": 3.949,
+      "eval_steps_per_second": 0.247,
+      "eval_wer": 6.489665823283936,
+      "step": 17000
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 1.5287179487179487e-06,
+      "loss": 0.0,
+      "step": 17025
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 1.515897435897436e-06,
+      "loss": 0.0,
+      "step": 17050
+    },
+    {
+      "epoch": 60.23,
+      "learning_rate": 1.5030769230769232e-06,
+      "loss": 0.0,
+      "step": 17075
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 1.4902564102564104e-06,
+      "loss": 0.0,
+      "step": 17100
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 1.4774358974358977e-06,
+      "loss": 0.0,
+      "step": 17125
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 1.4646153846153845e-06,
+      "loss": 0.0,
+      "step": 17150
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 1.4517948717948718e-06,
+      "loss": 0.0,
+      "step": 17175
+    },
+    {
+      "epoch": 60.67,
+      "learning_rate": 1.438974358974359e-06,
+      "loss": 0.0,
+      "step": 17200
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 1.4261538461538462e-06,
+      "loss": 0.0,
+      "step": 17225
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 1.4133333333333335e-06,
+      "loss": 0.0,
+      "step": 17250
+    },
+    {
+      "epoch": 60.93,
+      "learning_rate": 1.4005128205128207e-06,
+      "loss": 0.0,
+      "step": 17275
+    },
+    {
+      "epoch": 61.02,
+      "learning_rate": 1.3876923076923076e-06,
+      "loss": 0.0,
+      "step": 17300
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 1.3748717948717948e-06,
+      "loss": 0.0,
+      "step": 17325
+    },
+    {
+      "epoch": 61.2,
+      "learning_rate": 1.362051282051282e-06,
+      "loss": 0.0,
+      "step": 17350
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 1.3492307692307693e-06,
+      "loss": 0.0,
+      "step": 17375
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 1.3364102564102566e-06,
+      "loss": 0.0,
+      "step": 17400
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 1.3235897435897438e-06,
+      "loss": 0.0,
+      "step": 17425
+    },
+    {
+      "epoch": 61.55,
+      "learning_rate": 1.3107692307692308e-06,
+      "loss": 0.0,
+      "step": 17450
+    },
+    {
+      "epoch": 61.64,
+      "learning_rate": 1.2979487179487179e-06,
+      "loss": 0.0,
+      "step": 17475
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 1.2851282051282051e-06,
+      "loss": 0.0,
+      "step": 17500
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 1.2723076923076924e-06,
+      "loss": 0.0,
+      "step": 17525
+    },
+    {
+      "epoch": 61.9,
+      "learning_rate": 1.2594871794871796e-06,
+      "loss": 0.0,
+      "step": 17550
+    },
+    {
+      "epoch": 61.99,
+      "learning_rate": 1.2466666666666667e-06,
+      "loss": 0.0,
+      "step": 17575
+    },
+    {
+      "epoch": 62.08,
+      "learning_rate": 1.233846153846154e-06,
+      "loss": 0.0,
+      "step": 17600
+    },
+    {
+      "epoch": 62.17,
+      "learning_rate": 1.2210256410256412e-06,
+      "loss": 0.0,
+      "step": 17625
+    },
+    {
+      "epoch": 62.26,
+      "learning_rate": 1.2082051282051282e-06,
+      "loss": 0.0,
+      "step": 17650
+    },
+    {
+      "epoch": 62.35,
+      "learning_rate": 1.1953846153846154e-06,
+      "loss": 0.0,
+      "step": 17675
+    },
+    {
+      "epoch": 62.43,
+      "learning_rate": 1.1825641025641027e-06,
+      "loss": 0.0,
+      "step": 17700
+    },
+    {
+      "epoch": 62.52,
+      "learning_rate": 1.1697435897435897e-06,
+      "loss": 0.0,
+      "step": 17725
+    },
+    {
+      "epoch": 62.61,
+      "learning_rate": 1.156923076923077e-06,
+      "loss": 0.0,
+      "step": 17750
+    },
+    {
+      "epoch": 62.7,
+      "learning_rate": 1.1441025641025642e-06,
+      "loss": 0.0,
+      "step": 17775
+    },
+    {
+      "epoch": 62.79,
+      "learning_rate": 1.1312820512820513e-06,
+      "loss": 0.0,
+      "step": 17800
+    },
+    {
+      "epoch": 62.87,
+      "learning_rate": 1.1184615384615385e-06,
+      "loss": 0.0,
+      "step": 17825
+    },
+    {
+      "epoch": 62.96,
+      "learning_rate": 1.1056410256410258e-06,
+      "loss": 0.0,
+      "step": 17850
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 1.0928205128205128e-06,
+      "loss": 0.0,
+      "step": 17875
+    },
+    {
+      "epoch": 63.14,
+      "learning_rate": 1.08e-06,
+      "loss": 0.0,
+      "step": 17900
+    },
+    {
+      "epoch": 63.23,
+      "learning_rate": 1.0671794871794873e-06,
+      "loss": 0.0,
+      "step": 17925
+    },
+    {
+      "epoch": 63.32,
+      "learning_rate": 1.0543589743589743e-06,
+      "loss": 0.0,
+      "step": 17950
+    },
+    {
+      "epoch": 63.4,
+      "learning_rate": 1.0415384615384616e-06,
+      "loss": 0.0,
+      "step": 17975
+    },
+    {
+      "epoch": 63.49,
+      "learning_rate": 1.0287179487179488e-06,
+      "loss": 0.0,
+      "step": 18000
+    },
+    {
+      "epoch": 63.49,
+      "eval_loss": 0.4039324223995209,
+      "eval_runtime": 2287.4753,
+      "eval_samples_per_second": 3.966,
+      "eval_steps_per_second": 0.248,
+      "eval_wer": 6.465021522689186,
+      "step": 18000
+    },
+    {
+      "epoch": 63.58,
+      "learning_rate": 1.0158974358974359e-06,
+      "loss": 0.0,
+      "step": 18025
+    },
+    {
+      "epoch": 63.67,
+      "learning_rate": 1.003076923076923e-06,
+      "loss": 0.0,
+      "step": 18050
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 9.902564102564103e-07,
+      "loss": 0.0,
+      "step": 18075
+    },
+    {
+      "epoch": 63.84,
+      "learning_rate": 9.774358974358974e-07,
+      "loss": 0.0,
+      "step": 18100
+    },
+    {
+      "epoch": 63.93,
+      "learning_rate": 9.646153846153846e-07,
+      "loss": 0.0,
+      "step": 18125
+    },
+    {
+      "epoch": 64.02,
+      "learning_rate": 9.517948717948719e-07,
+      "loss": 0.0,
+      "step": 18150
+    },
+    {
+      "epoch": 64.11,
+      "learning_rate": 9.38974358974359e-07,
+      "loss": 0.0,
+      "step": 18175
+    },
+    {
+      "epoch": 64.2,
+      "learning_rate": 9.261538461538462e-07,
+      "loss": 0.0,
+      "step": 18200
+    },
+    {
+      "epoch": 64.29,
+      "learning_rate": 9.133333333333334e-07,
+      "loss": 0.0,
+      "step": 18225
+    },
+    {
+      "epoch": 64.37,
+      "learning_rate": 9.005128205128206e-07,
+      "loss": 0.0,
+      "step": 18250
+    },
+    {
+      "epoch": 64.46,
+      "learning_rate": 8.876923076923077e-07,
+      "loss": 0.0,
+      "step": 18275
+    },
+    {
+      "epoch": 64.55,
+      "learning_rate": 8.748717948717949e-07,
+      "loss": 0.0,
+      "step": 18300
+    },
+    {
+      "epoch": 64.64,
+      "learning_rate": 8.620512820512822e-07,
+      "loss": 0.0,
+      "step": 18325
+    },
+    {
+      "epoch": 64.73,
+      "learning_rate": 8.492307692307692e-07,
+      "loss": 0.0,
+      "step": 18350
+    },
+    {
+      "epoch": 64.81,
+      "learning_rate": 8.364102564102565e-07,
+      "loss": 0.0,
+      "step": 18375
+    },
+    {
+      "epoch": 64.9,
+      "learning_rate": 8.235897435897437e-07,
+      "loss": 0.0,
+      "step": 18400
+    },
+    {
+      "epoch": 64.99,
+      "learning_rate": 8.107692307692308e-07,
+      "loss": 0.0,
+      "step": 18425
+    },
+    {
+      "epoch": 65.08,
+      "learning_rate": 7.97948717948718e-07,
+      "loss": 0.0,
+      "step": 18450
+    },
+    {
+      "epoch": 65.17,
+      "learning_rate": 7.851282051282053e-07,
+      "loss": 0.0,
+      "step": 18475
+    },
+    {
+      "epoch": 65.26,
+      "learning_rate": 7.723076923076923e-07,
+      "loss": 0.0,
+      "step": 18500
+    },
+    {
+      "epoch": 65.34,
+      "learning_rate": 7.594871794871795e-07,
+      "loss": 0.0,
+      "step": 18525
+    },
+    {
+      "epoch": 65.43,
+      "learning_rate": 7.466666666666668e-07,
+      "loss": 0.0,
+      "step": 18550
+    },
+    {
+      "epoch": 65.52,
+      "learning_rate": 7.338461538461538e-07,
+      "loss": 0.0,
+      "step": 18575
+    },
+    {
+      "epoch": 65.61,
+      "learning_rate": 7.210256410256411e-07,
+      "loss": 0.0,
+      "step": 18600
+    },
+    {
+      "epoch": 65.7,
+      "learning_rate": 7.082051282051283e-07,
+      "loss": 0.0,
+      "step": 18625
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 6.953846153846154e-07,
+      "loss": 0.0,
+      "step": 18650
+    },
+    {
+      "epoch": 65.87,
+      "learning_rate": 6.825641025641026e-07,
+      "loss": 0.0,
+      "step": 18675
+    },
+    {
+      "epoch": 65.96,
+      "learning_rate": 6.697435897435899e-07,
+      "loss": 0.0,
+      "step": 18700
+    },
+    {
+      "epoch": 66.05,
+      "learning_rate": 6.569230769230769e-07,
+      "loss": 0.0,
+      "step": 18725
+    },
+    {
+      "epoch": 66.14,
+      "learning_rate": 6.441025641025641e-07,
+      "loss": 0.0,
+      "step": 18750
+    },
+    {
+      "epoch": 66.23,
+      "learning_rate": 6.312820512820514e-07,
+      "loss": 0.0,
+      "step": 18775
+    },
+    {
+      "epoch": 66.31,
+      "learning_rate": 6.184615384615385e-07,
+      "loss": 0.0,
+      "step": 18800
+    },
+    {
+      "epoch": 66.4,
+      "learning_rate": 6.056410256410257e-07,
+      "loss": 0.0,
+      "step": 18825
+    },
+    {
+      "epoch": 66.49,
+      "learning_rate": 5.928205128205128e-07,
+      "loss": 0.0,
+      "step": 18850
+    },
+    {
+      "epoch": 66.58,
+      "learning_rate": 5.800000000000001e-07,
+      "loss": 0.0,
+      "step": 18875
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 5.671794871794872e-07,
+      "loss": 0.0,
+      "step": 18900
+    },
+    {
+      "epoch": 66.75,
+      "learning_rate": 5.543589743589743e-07,
+      "loss": 0.0,
+      "step": 18925
+    },
+    {
+      "epoch": 66.84,
+      "learning_rate": 5.415384615384616e-07,
+      "loss": 0.0,
+      "step": 18950
+    },
+    {
+      "epoch": 66.93,
+      "learning_rate": 5.287179487179487e-07,
+      "loss": 0.0,
+      "step": 18975
+    },
+    {
+      "epoch": 67.02,
+      "learning_rate": 5.158974358974359e-07,
+      "loss": 0.0,
+      "step": 19000
+    },
+    {
+      "epoch": 67.02,
+      "eval_loss": 0.40909847617149353,
+      "eval_runtime": 2282.5584,
+      "eval_samples_per_second": 3.974,
+      "eval_steps_per_second": 0.248,
+      "eval_wer": 6.448591988959353,
+      "step": 19000
+    },
+    {
+      "epoch": 67.11,
+      "learning_rate": 5.030769230769231e-07,
+      "loss": 0.0,
+      "step": 19025
+    },
+    {
+      "epoch": 67.2,
+      "learning_rate": 4.902564102564103e-07,
+      "loss": 0.0,
+      "step": 19050
+    },
+    {
+      "epoch": 67.28,
+      "learning_rate": 4.774358974358974e-07,
+      "loss": 0.0,
+      "step": 19075
+    },
+    {
+      "epoch": 67.37,
+      "learning_rate": 4.6461538461538465e-07,
+      "loss": 0.0,
+      "step": 19100
+    },
+    {
+      "epoch": 67.46,
+      "learning_rate": 4.5179487179487185e-07,
+      "loss": 0.0,
+      "step": 19125
+    },
+    {
+      "epoch": 67.55,
+      "learning_rate": 4.38974358974359e-07,
+      "loss": 0.0,
+      "step": 19150
+    },
+    {
+      "epoch": 67.64,
+      "learning_rate": 4.261538461538462e-07,
+      "loss": 0.0,
+      "step": 19175
+    },
+    {
+      "epoch": 67.72,
+      "learning_rate": 4.133333333333334e-07,
+      "loss": 0.0,
+      "step": 19200
+    },
+    {
+      "epoch": 67.81,
+      "learning_rate": 4.005128205128205e-07,
+      "loss": 0.0,
+      "step": 19225
+    },
+    {
+      "epoch": 67.9,
+      "learning_rate": 3.876923076923077e-07,
+      "loss": 0.0,
+      "step": 19250
+    },
+    {
+      "epoch": 67.99,
+      "learning_rate": 3.748717948717949e-07,
+      "loss": 0.0,
+      "step": 19275
+    },
+    {
+      "epoch": 68.08,
+      "learning_rate": 3.6205128205128206e-07,
+      "loss": 0.0,
+      "step": 19300
+    },
+    {
+      "epoch": 68.17,
+      "learning_rate": 3.4923076923076925e-07,
+      "loss": 0.0,
+      "step": 19325
+    },
+    {
+      "epoch": 68.25,
+      "learning_rate": 3.3641025641025645e-07,
+      "loss": 0.0,
+      "step": 19350
+    },
+    {
+      "epoch": 68.34,
+      "learning_rate": 3.235897435897436e-07,
+      "loss": 0.0,
+      "step": 19375
+    },
+    {
+      "epoch": 68.43,
+      "learning_rate": 3.107692307692308e-07,
+      "loss": 0.0,
+      "step": 19400
+    },
+    {
+      "epoch": 68.52,
+      "learning_rate": 2.97948717948718e-07,
+      "loss": 0.0,
+      "step": 19425
+    },
+    {
+      "epoch": 68.61,
+      "learning_rate": 2.8512820512820517e-07,
+      "loss": 0.0,
+      "step": 19450
+    },
+    {
+      "epoch": 68.69,
+      "learning_rate": 2.7230769230769237e-07,
+      "loss": 0.0,
+      "step": 19475
+    },
+    {
+      "epoch": 68.78,
+      "learning_rate": 2.594871794871795e-07,
+      "loss": 0.0,
+      "step": 19500
+    },
+    {
+      "epoch": 68.87,
+      "learning_rate": 2.466666666666667e-07,
+      "loss": 0.0,
+      "step": 19525
+    },
+    {
+      "epoch": 68.96,
+      "learning_rate": 2.3384615384615387e-07,
+      "loss": 0.0,
+      "step": 19550
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 2.2102564102564104e-07,
+      "loss": 0.0,
+      "step": 19575
+    },
+    {
+      "epoch": 69.14,
+      "learning_rate": 2.082051282051282e-07,
+      "loss": 0.0,
+      "step": 19600
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 1.953846153846154e-07,
+      "loss": 0.0,
+      "step": 19625
+    },
+    {
+      "epoch": 69.31,
+      "learning_rate": 1.8256410256410257e-07,
+      "loss": 0.0,
+      "step": 19650
+    },
+    {
+      "epoch": 69.4,
+      "learning_rate": 1.6974358974358974e-07,
+      "loss": 0.0,
+      "step": 19675
+    },
+    {
+      "epoch": 69.49,
+      "learning_rate": 1.5692307692307694e-07,
+      "loss": 0.0,
+      "step": 19700
+    },
+    {
+      "epoch": 69.58,
+      "learning_rate": 1.441025641025641e-07,
+      "loss": 0.0,
+      "step": 19725
+    },
+    {
+      "epoch": 69.66,
+      "learning_rate": 1.3128205128205127e-07,
+      "loss": 0.0,
+      "step": 19750
+    },
+    {
+      "epoch": 69.75,
+      "learning_rate": 1.1846153846153847e-07,
+      "loss": 0.0,
+      "step": 19775
+    },
+    {
+      "epoch": 69.84,
+      "learning_rate": 1.0564102564102565e-07,
+      "loss": 0.0,
+      "step": 19800
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 9.282051282051283e-08,
+      "loss": 0.0,
+      "step": 19825
+    },
+    {
+      "epoch": 70.02,
+      "learning_rate": 8e-08,
+      "loss": 0.0,
+      "step": 19850
+    },
+    {
+      "epoch": 70.11,
+      "learning_rate": 6.717948717948718e-08,
+      "loss": 0.0,
+      "step": 19875
+    },
+    {
+      "epoch": 70.19,
+      "learning_rate": 5.435897435897436e-08,
+      "loss": 0.0,
+      "step": 19900
+    },
+    {
+      "epoch": 70.28,
+      "learning_rate": 4.153846153846154e-08,
+      "loss": 0.0,
+      "step": 19925
+    },
+    {
+      "epoch": 70.37,
+      "learning_rate": 2.8717948717948722e-08,
+      "loss": 0.0,
+      "step": 19950
+    },
+    {
+      "epoch": 70.46,
+      "learning_rate": 1.5897435897435897e-08,
+      "loss": 0.0,
+      "step": 19975
+    },
+    {
+      "epoch": 70.55,
+      "learning_rate": 3.0769230769230774e-09,
+      "loss": 0.0,
+      "step": 20000
+    },
+    {
+      "epoch": 70.55,
+      "eval_loss": 0.4114692807197571,
+      "eval_runtime": 2283.3075,
+      "eval_samples_per_second": 3.973,
+      "eval_steps_per_second": 0.248,
+      "eval_wer": 6.450234942332337,
+      "step": 20000
+    },
+    {
+      "epoch": 70.55,
+      "step": 20000,
+      "total_flos": 2.717658906624e+21,
+      "train_loss": 0.00939158419726882,
+      "train_runtime": 273485.7347,
+      "train_samples_per_second": 4.68,
+      "train_steps_per_second": 0.073
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 20000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 71,
+  "save_steps": 1000,
+  "total_flos": 2.717658906624e+21,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}