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": 10.612013447284216,
+  "best_model_checkpoint": "./checkpoint-17000",
+  "epoch": 97.0873786407767,
+  "eval_steps": 1000,
+  "global_step": 20000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 0.9486,
+      "step": 25
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.6167,
+      "step": 50
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5e-06,
+      "loss": 0.3962,
+      "step": 75
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.3335,
+      "step": 100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.5e-06,
+      "loss": 0.3167,
+      "step": 125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3e-06,
+      "loss": 0.2985,
+      "step": 150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.5e-06,
+      "loss": 0.2635,
+      "step": 175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.2606,
+      "step": 200
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5e-06,
+      "loss": 0.2147,
+      "step": 225
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 5e-06,
+      "loss": 0.2012,
+      "step": 250
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.1905,
+      "step": 275
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 6e-06,
+      "loss": 0.1895,
+      "step": 300
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.1846,
+      "step": 325
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7e-06,
+      "loss": 0.177,
+      "step": 350
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.1854,
+      "step": 375
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1718,
+      "step": 400
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 8.5e-06,
+      "loss": 0.1343,
+      "step": 425
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9e-06,
+      "loss": 0.1114,
+      "step": 450
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 9.5e-06,
+      "loss": 0.1049,
+      "step": 475
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1e-05,
+      "loss": 0.105,
+      "step": 500
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 9.987179487179488e-06,
+      "loss": 0.1093,
+      "step": 525
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.974358974358974e-06,
+      "loss": 0.1059,
+      "step": 550
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 9.961538461538463e-06,
+      "loss": 0.114,
+      "step": 575
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 9.94871794871795e-06,
+      "loss": 0.1051,
+      "step": 600
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 9.935897435897437e-06,
+      "loss": 0.0921,
+      "step": 625
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 9.923076923076923e-06,
+      "loss": 0.0565,
+      "step": 650
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 9.910256410256412e-06,
+      "loss": 0.0554,
+      "step": 675
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 9.897435897435899e-06,
+      "loss": 0.0575,
+      "step": 700
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 9.884615384615386e-06,
+      "loss": 0.0588,
+      "step": 725
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 9.871794871794872e-06,
+      "loss": 0.0558,
+      "step": 750
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 9.858974358974361e-06,
+      "loss": 0.0601,
+      "step": 775
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 9.846153846153848e-06,
+      "loss": 0.0596,
+      "step": 800
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 9.833333333333333e-06,
+      "loss": 0.0569,
+      "step": 825
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 9.820512820512821e-06,
+      "loss": 0.0311,
+      "step": 850
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 9.807692307692308e-06,
+      "loss": 0.0293,
+      "step": 875
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.794871794871795e-06,
+      "loss": 0.035,
+      "step": 900
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 9.782051282051282e-06,
+      "loss": 0.0316,
+      "step": 925
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 9.76923076923077e-06,
+      "loss": 0.0313,
+      "step": 950
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 9.756410256410257e-06,
+      "loss": 0.0346,
+      "step": 975
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 9.743589743589744e-06,
+      "loss": 0.0326,
+      "step": 1000
+    },
+    {
+      "epoch": 4.85,
+      "eval_loss": 0.22997283935546875,
+      "eval_runtime": 1414.2169,
+      "eval_samples_per_second": 4.661,
+      "eval_steps_per_second": 0.291,
+      "eval_wer": 13.327797804690347,
+      "step": 1000
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 9.730769230769231e-06,
+      "loss": 0.0351,
+      "step": 1025
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 9.71794871794872e-06,
+      "loss": 0.0214,
+      "step": 1050
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 9.705128205128206e-06,
+      "loss": 0.0172,
+      "step": 1075
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 9.692307692307693e-06,
+      "loss": 0.0174,
+      "step": 1100
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 9.67948717948718e-06,
+      "loss": 0.0206,
+      "step": 1125
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 0.0194,
+      "step": 1150
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 9.653846153846155e-06,
+      "loss": 0.0215,
+      "step": 1175
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 9.641025641025642e-06,
+      "loss": 0.0209,
+      "step": 1200
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 9.628205128205129e-06,
+      "loss": 0.0211,
+      "step": 1225
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 9.615384615384616e-06,
+      "loss": 0.0164,
+      "step": 1250
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 9.602564102564104e-06,
+      "loss": 0.0117,
+      "step": 1275
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 9.589743589743591e-06,
+      "loss": 0.0133,
+      "step": 1300
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 9.576923076923078e-06,
+      "loss": 0.0113,
+      "step": 1325
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 9.564102564102565e-06,
+      "loss": 0.0109,
+      "step": 1350
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.551282051282053e-06,
+      "loss": 0.0121,
+      "step": 1375
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.53846153846154e-06,
+      "loss": 0.0109,
+      "step": 1400
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.525641025641025e-06,
+      "loss": 0.0112,
+      "step": 1425
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 9.512820512820514e-06,
+      "loss": 0.0115,
+      "step": 1450
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 9.5e-06,
+      "loss": 0.0078,
+      "step": 1475
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 9.487179487179487e-06,
+      "loss": 0.0065,
+      "step": 1500
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 9.474358974358974e-06,
+      "loss": 0.008,
+      "step": 1525
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 9.461538461538463e-06,
+      "loss": 0.0097,
+      "step": 1550
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 9.44871794871795e-06,
+      "loss": 0.0084,
+      "step": 1575
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 9.435897435897436e-06,
+      "loss": 0.0083,
+      "step": 1600
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 9.423076923076923e-06,
+      "loss": 0.0082,
+      "step": 1625
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 9.410256410256412e-06,
+      "loss": 0.0077,
+      "step": 1650
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 9.397435897435899e-06,
+      "loss": 0.0051,
+      "step": 1675
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 9.384615384615385e-06,
+      "loss": 0.0043,
+      "step": 1700
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 9.371794871794872e-06,
+      "loss": 0.0045,
+      "step": 1725
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 9.358974358974359e-06,
+      "loss": 0.0047,
+      "step": 1750
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 9.346153846153847e-06,
+      "loss": 0.007,
+      "step": 1775
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 0.0058,
+      "step": 1800
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 9.320512820512821e-06,
+      "loss": 0.0063,
+      "step": 1825
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 9.307692307692308e-06,
+      "loss": 0.0066,
+      "step": 1850
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 9.294871794871796e-06,
+      "loss": 0.005,
+      "step": 1875
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 9.282051282051283e-06,
+      "loss": 0.0044,
+      "step": 1900
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 9.26923076923077e-06,
+      "loss": 0.0036,
+      "step": 1925
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 9.256410256410257e-06,
+      "loss": 0.0041,
+      "step": 1950
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 9.243589743589745e-06,
+      "loss": 0.0041,
+      "step": 1975
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 0.004,
+      "step": 2000
+    },
+    {
+      "epoch": 9.71,
+      "eval_loss": 0.27230334281921387,
+      "eval_runtime": 1403.6044,
+      "eval_samples_per_second": 4.696,
+      "eval_steps_per_second": 0.294,
+      "eval_wer": 12.203815464376849,
+      "step": 2000
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 9.217948717948717e-06,
+      "loss": 0.0057,
+      "step": 2025
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 9.205128205128206e-06,
+      "loss": 0.0063,
+      "step": 2050
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 9.192307692307693e-06,
+      "loss": 0.0049,
+      "step": 2075
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 9.17948717948718e-06,
+      "loss": 0.0037,
+      "step": 2100
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 0.0026,
+      "step": 2125
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 9.153846153846155e-06,
+      "loss": 0.0027,
+      "step": 2150
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 9.141025641025642e-06,
+      "loss": 0.0036,
+      "step": 2175
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 9.128205128205129e-06,
+      "loss": 0.0037,
+      "step": 2200
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 9.115384615384615e-06,
+      "loss": 0.0043,
+      "step": 2225
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 9.102564102564104e-06,
+      "loss": 0.0031,
+      "step": 2250
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 9.08974358974359e-06,
+      "loss": 0.004,
+      "step": 2275
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 9.076923076923078e-06,
+      "loss": 0.0028,
+      "step": 2300
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 9.064102564102564e-06,
+      "loss": 0.0036,
+      "step": 2325
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 9.051282051282051e-06,
+      "loss": 0.0038,
+      "step": 2350
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 9.03846153846154e-06,
+      "loss": 0.0036,
+      "step": 2375
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 9.025641025641027e-06,
+      "loss": 0.0046,
+      "step": 2400
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 9.012820512820513e-06,
+      "loss": 0.0041,
+      "step": 2425
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 9e-06,
+      "loss": 0.0044,
+      "step": 2450
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 8.987179487179489e-06,
+      "loss": 0.0044,
+      "step": 2475
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 8.974358974358976e-06,
+      "loss": 0.0031,
+      "step": 2500
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 8.961538461538462e-06,
+      "loss": 0.0026,
+      "step": 2525
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 8.94871794871795e-06,
+      "loss": 0.0034,
+      "step": 2550
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 8.935897435897438e-06,
+      "loss": 0.0037,
+      "step": 2575
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 8.923076923076925e-06,
+      "loss": 0.0037,
+      "step": 2600
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 8.910256410256411e-06,
+      "loss": 0.0033,
+      "step": 2625
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 8.897435897435898e-06,
+      "loss": 0.0029,
+      "step": 2650
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 8.884615384615385e-06,
+      "loss": 0.0048,
+      "step": 2675
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 8.871794871794872e-06,
+      "loss": 0.0035,
+      "step": 2700
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 8.858974358974359e-06,
+      "loss": 0.0021,
+      "step": 2725
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.0037,
+      "step": 2750
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 8.833333333333334e-06,
+      "loss": 0.0049,
+      "step": 2775
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 8.820512820512821e-06,
+      "loss": 0.0039,
+      "step": 2800
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 8.807692307692308e-06,
+      "loss": 0.0044,
+      "step": 2825
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 8.794871794871796e-06,
+      "loss": 0.0055,
+      "step": 2850
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 8.782051282051283e-06,
+      "loss": 0.0046,
+      "step": 2875
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 8.76923076923077e-06,
+      "loss": 0.0043,
+      "step": 2900
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 8.756410256410257e-06,
+      "loss": 0.0039,
+      "step": 2925
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.743589743589743e-06,
+      "loss": 0.0042,
+      "step": 2950
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.730769230769232e-06,
+      "loss": 0.0048,
+      "step": 2975
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.717948717948719e-06,
+      "loss": 0.0058,
+      "step": 3000
+    },
+    {
+      "epoch": 14.56,
+      "eval_loss": 0.2770913541316986,
+      "eval_runtime": 1404.2904,
+      "eval_samples_per_second": 4.693,
+      "eval_steps_per_second": 0.293,
+      "eval_wer": 12.424561545627608,
+      "step": 3000
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.705128205128206e-06,
+      "loss": 0.0065,
+      "step": 3025
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.692307692307692e-06,
+      "loss": 0.0071,
+      "step": 3050
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 8.679487179487181e-06,
+      "loss": 0.0058,
+      "step": 3075
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 8.666666666666668e-06,
+      "loss": 0.0048,
+      "step": 3100
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 8.653846153846155e-06,
+      "loss": 0.0042,
+      "step": 3125
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 8.641025641025641e-06,
+      "loss": 0.0061,
+      "step": 3150
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 8.62820512820513e-06,
+      "loss": 0.0041,
+      "step": 3175
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 8.615384615384617e-06,
+      "loss": 0.0038,
+      "step": 3200
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 8.602564102564104e-06,
+      "loss": 0.0051,
+      "step": 3225
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 8.58974358974359e-06,
+      "loss": 0.0047,
+      "step": 3250
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 8.576923076923077e-06,
+      "loss": 0.0037,
+      "step": 3275
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 8.564102564102564e-06,
+      "loss": 0.0034,
+      "step": 3300
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 8.551282051282051e-06,
+      "loss": 0.0035,
+      "step": 3325
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 8.53846153846154e-06,
+      "loss": 0.0028,
+      "step": 3350
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 8.525641025641026e-06,
+      "loss": 0.0024,
+      "step": 3375
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.512820512820513e-06,
+      "loss": 0.002,
+      "step": 3400
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.5e-06,
+      "loss": 0.0035,
+      "step": 3425
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.487179487179488e-06,
+      "loss": 0.0025,
+      "step": 3450
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 8.474358974358975e-06,
+      "loss": 0.0036,
+      "step": 3475
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 8.461538461538462e-06,
+      "loss": 0.0043,
+      "step": 3500
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 8.448717948717949e-06,
+      "loss": 0.0032,
+      "step": 3525
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 8.435897435897436e-06,
+      "loss": 0.0031,
+      "step": 3550
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 8.423076923076924e-06,
+      "loss": 0.003,
+      "step": 3575
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 8.410256410256411e-06,
+      "loss": 0.0026,
+      "step": 3600
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 8.397435897435898e-06,
+      "loss": 0.0027,
+      "step": 3625
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 8.384615384615385e-06,
+      "loss": 0.0022,
+      "step": 3650
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 8.371794871794873e-06,
+      "loss": 0.0024,
+      "step": 3675
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 8.35897435897436e-06,
+      "loss": 0.0017,
+      "step": 3700
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 8.346153846153847e-06,
+      "loss": 0.0017,
+      "step": 3725
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.0019,
+      "step": 3750
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 8.320512820512822e-06,
+      "loss": 0.0016,
+      "step": 3775
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 8.307692307692309e-06,
+      "loss": 0.0019,
+      "step": 3800
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 8.294871794871796e-06,
+      "loss": 0.0015,
+      "step": 3825
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 8.282051282051283e-06,
+      "loss": 0.0027,
+      "step": 3850
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 8.26923076923077e-06,
+      "loss": 0.0034,
+      "step": 3875
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 8.256410256410256e-06,
+      "loss": 0.0034,
+      "step": 3900
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 8.243589743589743e-06,
+      "loss": 0.0047,
+      "step": 3925
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 8.230769230769232e-06,
+      "loss": 0.0034,
+      "step": 3950
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 8.217948717948719e-06,
+      "loss": 0.0027,
+      "step": 3975
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 8.205128205128205e-06,
+      "loss": 0.003,
+      "step": 4000
+    },
+    {
+      "epoch": 19.42,
+      "eval_loss": 0.28377285599708557,
+      "eval_runtime": 1406.4432,
+      "eval_samples_per_second": 4.686,
+      "eval_steps_per_second": 0.293,
+      "eval_wer": 12.211916238000729,
+      "step": 4000
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 8.192307692307692e-06,
+      "loss": 0.0033,
+      "step": 4025
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.17948717948718e-06,
+      "loss": 0.0039,
+      "step": 4050
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 8.166666666666668e-06,
+      "loss": 0.0047,
+      "step": 4075
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 8.153846153846154e-06,
+      "loss": 0.0058,
+      "step": 4100
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 8.141025641025641e-06,
+      "loss": 0.0056,
+      "step": 4125
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 8.12820512820513e-06,
+      "loss": 0.0043,
+      "step": 4150
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 8.115384615384617e-06,
+      "loss": 0.0036,
+      "step": 4175
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 8.102564102564103e-06,
+      "loss": 0.0034,
+      "step": 4200
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 8.08974358974359e-06,
+      "loss": 0.0046,
+      "step": 4225
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 8.076923076923077e-06,
+      "loss": 0.0037,
+      "step": 4250
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 8.064102564102566e-06,
+      "loss": 0.0035,
+      "step": 4275
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 8.051282051282052e-06,
+      "loss": 0.0041,
+      "step": 4300
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 8.03846153846154e-06,
+      "loss": 0.0041,
+      "step": 4325
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 8.025641025641026e-06,
+      "loss": 0.0025,
+      "step": 4350
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 8.012820512820515e-06,
+      "loss": 0.0019,
+      "step": 4375
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.0034,
+      "step": 4400
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 7.987179487179488e-06,
+      "loss": 0.0028,
+      "step": 4425
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 7.974358974358975e-06,
+      "loss": 0.0027,
+      "step": 4450
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 7.961538461538462e-06,
+      "loss": 0.0023,
+      "step": 4475
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 7.948717948717949e-06,
+      "loss": 0.0033,
+      "step": 4500
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 7.935897435897435e-06,
+      "loss": 0.0024,
+      "step": 4525
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 7.923076923076924e-06,
+      "loss": 0.0023,
+      "step": 4550
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 7.91025641025641e-06,
+      "loss": 0.0023,
+      "step": 4575
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 7.897435897435898e-06,
+      "loss": 0.0026,
+      "step": 4600
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 7.884615384615384e-06,
+      "loss": 0.0033,
+      "step": 4625
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 7.871794871794873e-06,
+      "loss": 0.0034,
+      "step": 4650
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 7.85897435897436e-06,
+      "loss": 0.0034,
+      "step": 4675
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 7.846153846153847e-06,
+      "loss": 0.004,
+      "step": 4700
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 7.833333333333333e-06,
+      "loss": 0.0047,
+      "step": 4725
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 7.820512820512822e-06,
+      "loss": 0.0051,
+      "step": 4750
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 7.807692307692309e-06,
+      "loss": 0.0024,
+      "step": 4775
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 7.794871794871796e-06,
+      "loss": 0.0028,
+      "step": 4800
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 7.782051282051282e-06,
+      "loss": 0.0037,
+      "step": 4825
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 7.76923076923077e-06,
+      "loss": 0.0029,
+      "step": 4850
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 7.756410256410258e-06,
+      "loss": 0.0028,
+      "step": 4875
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 7.743589743589745e-06,
+      "loss": 0.0023,
+      "step": 4900
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 7.730769230769231e-06,
+      "loss": 0.0021,
+      "step": 4925
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 7.717948717948718e-06,
+      "loss": 0.0021,
+      "step": 4950
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 7.705128205128207e-06,
+      "loss": 0.0017,
+      "step": 4975
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 7.692307692307694e-06,
+      "loss": 0.003,
+      "step": 5000
+    },
+    {
+      "epoch": 24.27,
+      "eval_loss": 0.27398020029067993,
+      "eval_runtime": 1401.4636,
+      "eval_samples_per_second": 4.703,
+      "eval_steps_per_second": 0.294,
+      "eval_wer": 11.77042407549921,
+      "step": 5000
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 7.67948717948718e-06,
+      "loss": 0.0023,
+      "step": 5025
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 7.666666666666667e-06,
+      "loss": 0.0022,
+      "step": 5050
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 7.653846153846154e-06,
+      "loss": 0.0027,
+      "step": 5075
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 7.641025641025641e-06,
+      "loss": 0.003,
+      "step": 5100
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 7.6282051282051286e-06,
+      "loss": 0.0025,
+      "step": 5125
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 7.615384615384615e-06,
+      "loss": 0.0028,
+      "step": 5150
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 7.602564102564103e-06,
+      "loss": 0.0035,
+      "step": 5175
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 7.58974358974359e-06,
+      "loss": 0.0023,
+      "step": 5200
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 7.5769230769230775e-06,
+      "loss": 0.0019,
+      "step": 5225
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.0018,
+      "step": 5250
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 7.551282051282052e-06,
+      "loss": 0.0024,
+      "step": 5275
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 7.538461538461539e-06,
+      "loss": 0.002,
+      "step": 5300
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 7.5256410256410265e-06,
+      "loss": 0.0019,
+      "step": 5325
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 7.512820512820513e-06,
+      "loss": 0.0017,
+      "step": 5350
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.002,
+      "step": 5375
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 7.487179487179488e-06,
+      "loss": 0.0029,
+      "step": 5400
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 7.474358974358975e-06,
+      "loss": 0.0022,
+      "step": 5425
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 7.461538461538462e-06,
+      "loss": 0.0019,
+      "step": 5450
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 7.448717948717949e-06,
+      "loss": 0.0025,
+      "step": 5475
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 7.435897435897437e-06,
+      "loss": 0.0017,
+      "step": 5500
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 7.423076923076924e-06,
+      "loss": 0.0025,
+      "step": 5525
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 7.410256410256411e-06,
+      "loss": 0.0031,
+      "step": 5550
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 7.397435897435898e-06,
+      "loss": 0.0022,
+      "step": 5575
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 7.384615384615386e-06,
+      "loss": 0.0018,
+      "step": 5600
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 7.371794871794873e-06,
+      "loss": 0.0017,
+      "step": 5625
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 7.35897435897436e-06,
+      "loss": 0.0017,
+      "step": 5650
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 7.346153846153847e-06,
+      "loss": 0.0013,
+      "step": 5675
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 7.333333333333333e-06,
+      "loss": 0.001,
+      "step": 5700
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 7.320512820512821e-06,
+      "loss": 0.0015,
+      "step": 5725
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 7.307692307692308e-06,
+      "loss": 0.0022,
+      "step": 5750
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 7.294871794871795e-06,
+      "loss": 0.0016,
+      "step": 5775
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 7.282051282051282e-06,
+      "loss": 0.0017,
+      "step": 5800
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 7.26923076923077e-06,
+      "loss": 0.0017,
+      "step": 5825
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 7.256410256410257e-06,
+      "loss": 0.0015,
+      "step": 5850
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 7.243589743589744e-06,
+      "loss": 0.0025,
+      "step": 5875
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 7.230769230769231e-06,
+      "loss": 0.002,
+      "step": 5900
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 7.217948717948719e-06,
+      "loss": 0.0023,
+      "step": 5925
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 7.205128205128206e-06,
+      "loss": 0.0021,
+      "step": 5950
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 7.192307692307693e-06,
+      "loss": 0.002,
+      "step": 5975
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 7.17948717948718e-06,
+      "loss": 0.0014,
+      "step": 6000
+    },
+    {
+      "epoch": 29.13,
+      "eval_loss": 0.29363375902175903,
+      "eval_runtime": 1399.0954,
+      "eval_samples_per_second": 4.711,
+      "eval_steps_per_second": 0.294,
+      "eval_wer": 11.543602414030541,
+      "step": 6000
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 7.166666666666667e-06,
+      "loss": 0.0013,
+      "step": 6025
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 7.153846153846155e-06,
+      "loss": 0.001,
+      "step": 6050
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 7.1410256410256414e-06,
+      "loss": 0.0012,
+      "step": 6075
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 7.128205128205129e-06,
+      "loss": 0.0016,
+      "step": 6100
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 7.115384615384616e-06,
+      "loss": 0.0024,
+      "step": 6125
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 7.102564102564104e-06,
+      "loss": 0.0019,
+      "step": 6150
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 7.0897435897435904e-06,
+      "loss": 0.0035,
+      "step": 6175
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 7.076923076923078e-06,
+      "loss": 0.0034,
+      "step": 6200
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 7.0646153846153855e-06,
+      "loss": 0.0027,
+      "step": 6225
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 7.051794871794872e-06,
+      "loss": 0.0022,
+      "step": 6250
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 7.03897435897436e-06,
+      "loss": 0.0029,
+      "step": 6275
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 7.026153846153847e-06,
+      "loss": 0.0023,
+      "step": 6300
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 7.0133333333333345e-06,
+      "loss": 0.0015,
+      "step": 6325
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 7.000512820512821e-06,
+      "loss": 0.0032,
+      "step": 6350
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 6.987692307692309e-06,
+      "loss": 0.0023,
+      "step": 6375
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 6.974871794871796e-06,
+      "loss": 0.0025,
+      "step": 6400
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 6.9620512820512835e-06,
+      "loss": 0.0021,
+      "step": 6425
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 6.94923076923077e-06,
+      "loss": 0.002,
+      "step": 6450
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 6.936410256410256e-06,
+      "loss": 0.0026,
+      "step": 6475
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 6.923589743589744e-06,
+      "loss": 0.0031,
+      "step": 6500
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 6.910769230769231e-06,
+      "loss": 0.0024,
+      "step": 6525
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 6.8979487179487185e-06,
+      "loss": 0.0016,
+      "step": 6550
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 6.885128205128205e-06,
+      "loss": 0.0017,
+      "step": 6575
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 6.872307692307693e-06,
+      "loss": 0.0019,
+      "step": 6600
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 6.85948717948718e-06,
+      "loss": 0.0018,
+      "step": 6625
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 6.846666666666667e-06,
+      "loss": 0.0017,
+      "step": 6650
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 6.833846153846154e-06,
+      "loss": 0.0013,
+      "step": 6675
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 6.821025641025641e-06,
+      "loss": 0.0014,
+      "step": 6700
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 6.808205128205129e-06,
+      "loss": 0.0012,
+      "step": 6725
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 6.795384615384616e-06,
+      "loss": 0.0017,
+      "step": 6750
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 6.782564102564103e-06,
+      "loss": 0.0019,
+      "step": 6775
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 6.76974358974359e-06,
+      "loss": 0.0025,
+      "step": 6800
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 6.756923076923078e-06,
+      "loss": 0.0015,
+      "step": 6825
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 6.744102564102565e-06,
+      "loss": 0.0014,
+      "step": 6850
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 6.731282051282052e-06,
+      "loss": 0.0017,
+      "step": 6875
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 6.718461538461539e-06,
+      "loss": 0.0014,
+      "step": 6900
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 6.705641025641027e-06,
+      "loss": 0.0013,
+      "step": 6925
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 6.692820512820514e-06,
+      "loss": 0.0017,
+      "step": 6950
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 6.680000000000001e-06,
+      "loss": 0.0014,
+      "step": 6975
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 6.667179487179488e-06,
+      "loss": 0.0015,
+      "step": 7000
+    },
+    {
+      "epoch": 33.98,
+      "eval_loss": 0.291135311126709,
+      "eval_runtime": 1399.6671,
+      "eval_samples_per_second": 4.709,
+      "eval_steps_per_second": 0.294,
+      "eval_wer": 11.519300093158897,
+      "step": 7000
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 6.654358974358976e-06,
+      "loss": 0.0024,
+      "step": 7025
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 6.641538461538463e-06,
+      "loss": 0.0022,
+      "step": 7050
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 6.6287179487179486e-06,
+      "loss": 0.0012,
+      "step": 7075
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 6.615897435897436e-06,
+      "loss": 0.0013,
+      "step": 7100
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 6.603076923076923e-06,
+      "loss": 0.0019,
+      "step": 7125
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 6.590256410256411e-06,
+      "loss": 0.0019,
+      "step": 7150
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 6.5774358974358976e-06,
+      "loss": 0.0019,
+      "step": 7175
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 6.564615384615385e-06,
+      "loss": 0.0018,
+      "step": 7200
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 6.551794871794872e-06,
+      "loss": 0.0022,
+      "step": 7225
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 6.538974358974359e-06,
+      "loss": 0.0014,
+      "step": 7250
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 6.5261538461538465e-06,
+      "loss": 0.0011,
+      "step": 7275
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 6.513333333333333e-06,
+      "loss": 0.0009,
+      "step": 7300
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 6.500512820512821e-06,
+      "loss": 0.0009,
+      "step": 7325
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 6.487692307692308e-06,
+      "loss": 0.0011,
+      "step": 7350
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 6.4748717948717955e-06,
+      "loss": 0.001,
+      "step": 7375
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 6.462051282051282e-06,
+      "loss": 0.0006,
+      "step": 7400
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 6.44923076923077e-06,
+      "loss": 0.0008,
+      "step": 7425
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 6.436410256410257e-06,
+      "loss": 0.0004,
+      "step": 7450
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 6.4235897435897445e-06,
+      "loss": 0.0004,
+      "step": 7475
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 6.410769230769231e-06,
+      "loss": 0.0008,
+      "step": 7500
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 6.397948717948719e-06,
+      "loss": 0.0004,
+      "step": 7525
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 6.385128205128206e-06,
+      "loss": 0.0004,
+      "step": 7550
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 6.3723076923076935e-06,
+      "loss": 0.001,
+      "step": 7575
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 6.35948717948718e-06,
+      "loss": 0.0012,
+      "step": 7600
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 6.346666666666668e-06,
+      "loss": 0.0016,
+      "step": 7625
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 6.333846153846155e-06,
+      "loss": 0.0013,
+      "step": 7650
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 6.3210256410256425e-06,
+      "loss": 0.0015,
+      "step": 7675
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 6.3082051282051285e-06,
+      "loss": 0.002,
+      "step": 7700
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 6.295384615384615e-06,
+      "loss": 0.001,
+      "step": 7725
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 6.282564102564103e-06,
+      "loss": 0.0014,
+      "step": 7750
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 6.26974358974359e-06,
+      "loss": 0.0026,
+      "step": 7775
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 6.2569230769230775e-06,
+      "loss": 0.0016,
+      "step": 7800
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 6.244102564102564e-06,
+      "loss": 0.0017,
+      "step": 7825
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 6.231282051282051e-06,
+      "loss": 0.0013,
+      "step": 7850
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 6.218461538461539e-06,
+      "loss": 0.001,
+      "step": 7875
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 6.205641025641026e-06,
+      "loss": 0.0017,
+      "step": 7900
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 6.192820512820513e-06,
+      "loss": 0.0025,
+      "step": 7925
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 6.18e-06,
+      "loss": 0.0023,
+      "step": 7950
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 6.167179487179488e-06,
+      "loss": 0.0018,
+      "step": 7975
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 6.154358974358975e-06,
+      "loss": 0.0012,
+      "step": 8000
+    },
+    {
+      "epoch": 38.83,
+      "eval_loss": 0.29390278458595276,
+      "eval_runtime": 1396.9481,
+      "eval_samples_per_second": 4.718,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 11.367410587711127,
+      "step": 8000
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 6.141538461538462e-06,
+      "loss": 0.0011,
+      "step": 8025
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 6.128717948717949e-06,
+      "loss": 0.0017,
+      "step": 8050
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 6.115897435897437e-06,
+      "loss": 0.0014,
+      "step": 8075
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 6.103076923076924e-06,
+      "loss": 0.0011,
+      "step": 8100
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 6.090256410256411e-06,
+      "loss": 0.0015,
+      "step": 8125
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 6.077435897435898e-06,
+      "loss": 0.0011,
+      "step": 8150
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 6.064615384615386e-06,
+      "loss": 0.0022,
+      "step": 8175
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 6.051794871794873e-06,
+      "loss": 0.0018,
+      "step": 8200
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 6.03897435897436e-06,
+      "loss": 0.0014,
+      "step": 8225
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 6.026153846153847e-06,
+      "loss": 0.0007,
+      "step": 8250
+    },
+    {
+      "epoch": 40.17,
+      "learning_rate": 6.013333333333335e-06,
+      "loss": 0.0008,
+      "step": 8275
+    },
+    {
+      "epoch": 40.29,
+      "learning_rate": 6.000512820512821e-06,
+      "loss": 0.0008,
+      "step": 8300
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 5.9876923076923076e-06,
+      "loss": 0.0009,
+      "step": 8325
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 5.974871794871795e-06,
+      "loss": 0.0012,
+      "step": 8350
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 5.962051282051282e-06,
+      "loss": 0.0013,
+      "step": 8375
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 5.94923076923077e-06,
+      "loss": 0.001,
+      "step": 8400
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 5.9364102564102566e-06,
+      "loss": 0.0008,
+      "step": 8425
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 5.923589743589744e-06,
+      "loss": 0.0006,
+      "step": 8450
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 5.910769230769231e-06,
+      "loss": 0.0002,
+      "step": 8475
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 5.897948717948718e-06,
+      "loss": 0.0005,
+      "step": 8500
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 5.8851282051282056e-06,
+      "loss": 0.0003,
+      "step": 8525
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 5.872307692307692e-06,
+      "loss": 0.0002,
+      "step": 8550
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 5.85948717948718e-06,
+      "loss": 0.0003,
+      "step": 8575
+    },
+    {
+      "epoch": 41.75,
+      "learning_rate": 5.846666666666667e-06,
+      "loss": 0.0009,
+      "step": 8600
+    },
+    {
+      "epoch": 41.87,
+      "learning_rate": 5.8338461538461545e-06,
+      "loss": 0.0006,
+      "step": 8625
+    },
+    {
+      "epoch": 41.99,
+      "learning_rate": 5.821025641025641e-06,
+      "loss": 0.0006,
+      "step": 8650
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 5.808205128205129e-06,
+      "loss": 0.0003,
+      "step": 8675
+    },
+    {
+      "epoch": 42.23,
+      "learning_rate": 5.795384615384616e-06,
+      "loss": 0.0002,
+      "step": 8700
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 5.7825641025641035e-06,
+      "loss": 0.0005,
+      "step": 8725
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 5.76974358974359e-06,
+      "loss": 0.001,
+      "step": 8750
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 5.756923076923078e-06,
+      "loss": 0.001,
+      "step": 8775
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 5.744102564102565e-06,
+      "loss": 0.0009,
+      "step": 8800
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 5.7312820512820525e-06,
+      "loss": 0.0008,
+      "step": 8825
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 5.718461538461539e-06,
+      "loss": 0.0009,
+      "step": 8850
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 5.705641025641027e-06,
+      "loss": 0.0009,
+      "step": 8875
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 5.692820512820513e-06,
+      "loss": 0.0007,
+      "step": 8900
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 5.68e-06,
+      "loss": 0.0003,
+      "step": 8925
+    },
+    {
+      "epoch": 43.45,
+      "learning_rate": 5.6671794871794875e-06,
+      "loss": 0.0003,
+      "step": 8950
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 5.654871794871796e-06,
+      "loss": 0.0005,
+      "step": 8975
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 5.642051282051283e-06,
+      "loss": 0.0009,
+      "step": 9000
+    },
+    {
+      "epoch": 43.69,
+      "eval_loss": 0.303938627243042,
+      "eval_runtime": 1404.6996,
+      "eval_samples_per_second": 4.692,
+      "eval_steps_per_second": 0.293,
+      "eval_wer": 11.413990036048443,
+      "step": 9000
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 5.62923076923077e-06,
+      "loss": 0.0013,
+      "step": 9025
+    },
+    {
+      "epoch": 43.93,
+      "learning_rate": 5.616410256410257e-06,
+      "loss": 0.001,
+      "step": 9050
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 5.603589743589744e-06,
+      "loss": 0.0019,
+      "step": 9075
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 5.590769230769231e-06,
+      "loss": 0.0023,
+      "step": 9100
+    },
+    {
+      "epoch": 44.3,
+      "learning_rate": 5.5779487179487176e-06,
+      "loss": 0.0026,
+      "step": 9125
+    },
+    {
+      "epoch": 44.42,
+      "learning_rate": 5.565128205128205e-06,
+      "loss": 0.0021,
+      "step": 9150
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 5.552307692307692e-06,
+      "loss": 0.002,
+      "step": 9175
+    },
+    {
+      "epoch": 44.66,
+      "learning_rate": 5.53948717948718e-06,
+      "loss": 0.0023,
+      "step": 9200
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 5.5266666666666666e-06,
+      "loss": 0.0021,
+      "step": 9225
+    },
+    {
+      "epoch": 44.9,
+      "learning_rate": 5.513846153846154e-06,
+      "loss": 0.0025,
+      "step": 9250
+    },
+    {
+      "epoch": 45.02,
+      "learning_rate": 5.501025641025641e-06,
+      "loss": 0.002,
+      "step": 9275
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 5.488205128205129e-06,
+      "loss": 0.0013,
+      "step": 9300
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 5.4753846153846155e-06,
+      "loss": 0.001,
+      "step": 9325
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 5.462564102564103e-06,
+      "loss": 0.001,
+      "step": 9350
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 5.44974358974359e-06,
+      "loss": 0.0016,
+      "step": 9375
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 5.436923076923078e-06,
+      "loss": 0.0021,
+      "step": 9400
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 5.4241025641025645e-06,
+      "loss": 0.001,
+      "step": 9425
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 5.411282051282052e-06,
+      "loss": 0.0018,
+      "step": 9450
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 5.398461538461539e-06,
+      "loss": 0.0019,
+      "step": 9475
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 5.385641025641027e-06,
+      "loss": 0.0005,
+      "step": 9500
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 5.3728205128205135e-06,
+      "loss": 0.0006,
+      "step": 9525
+    },
+    {
+      "epoch": 46.36,
+      "learning_rate": 5.36e-06,
+      "loss": 0.0005,
+      "step": 9550
+    },
+    {
+      "epoch": 46.48,
+      "learning_rate": 5.347179487179488e-06,
+      "loss": 0.0003,
+      "step": 9575
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 5.334358974358975e-06,
+      "loss": 0.0003,
+      "step": 9600
+    },
+    {
+      "epoch": 46.72,
+      "learning_rate": 5.3215384615384625e-06,
+      "loss": 0.0003,
+      "step": 9625
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 5.308717948717949e-06,
+      "loss": 0.0012,
+      "step": 9650
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 5.295897435897437e-06,
+      "loss": 0.0004,
+      "step": 9675
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 5.283076923076923e-06,
+      "loss": 0.0005,
+      "step": 9700
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 5.27025641025641e-06,
+      "loss": 0.0003,
+      "step": 9725
+    },
+    {
+      "epoch": 47.33,
+      "learning_rate": 5.2574358974358975e-06,
+      "loss": 0.0004,
+      "step": 9750
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 5.244615384615384e-06,
+      "loss": 0.0009,
+      "step": 9775
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 5.231794871794872e-06,
+      "loss": 0.0003,
+      "step": 9800
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 5.218974358974359e-06,
+      "loss": 0.0002,
+      "step": 9825
+    },
+    {
+      "epoch": 47.82,
+      "learning_rate": 5.2061538461538465e-06,
+      "loss": 0.0001,
+      "step": 9850
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 5.193333333333333e-06,
+      "loss": 0.0003,
+      "step": 9875
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 5.180512820512821e-06,
+      "loss": 0.0004,
+      "step": 9900
+    },
+    {
+      "epoch": 48.18,
+      "learning_rate": 5.167692307692308e-06,
+      "loss": 0.0004,
+      "step": 9925
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 5.1548717948717955e-06,
+      "loss": 0.0002,
+      "step": 9950
+    },
+    {
+      "epoch": 48.42,
+      "learning_rate": 5.142051282051282e-06,
+      "loss": 0.0003,
+      "step": 9975
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 5.12923076923077e-06,
+      "loss": 0.0002,
+      "step": 10000
+    },
+    {
+      "epoch": 48.54,
+      "eval_loss": 0.30629321932792664,
+      "eval_runtime": 1399.0538,
+      "eval_samples_per_second": 4.711,
+      "eval_steps_per_second": 0.294,
+      "eval_wer": 10.962371906517072,
+      "step": 10000
+    },
+    {
+      "epoch": 48.67,
+      "learning_rate": 5.116410256410257e-06,
+      "loss": 0.0003,
+      "step": 10025
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 5.1035897435897445e-06,
+      "loss": 0.0003,
+      "step": 10050
+    },
+    {
+      "epoch": 48.91,
+      "learning_rate": 5.090769230769231e-06,
+      "loss": 0.0003,
+      "step": 10075
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 5.077948717948719e-06,
+      "loss": 0.0008,
+      "step": 10100
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 5.065128205128206e-06,
+      "loss": 0.0007,
+      "step": 10125
+    },
+    {
+      "epoch": 49.27,
+      "learning_rate": 5.052307692307693e-06,
+      "loss": 0.0015,
+      "step": 10150
+    },
+    {
+      "epoch": 49.39,
+      "learning_rate": 5.03948717948718e-06,
+      "loss": 0.0006,
+      "step": 10175
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 5.026666666666667e-06,
+      "loss": 0.0007,
+      "step": 10200
+    },
+    {
+      "epoch": 49.64,
+      "learning_rate": 5.013846153846155e-06,
+      "loss": 0.001,
+      "step": 10225
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 5.001025641025642e-06,
+      "loss": 0.0009,
+      "step": 10250
+    },
+    {
+      "epoch": 49.88,
+      "learning_rate": 4.988205128205128e-06,
+      "loss": 0.0006,
+      "step": 10275
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 4.975384615384616e-06,
+      "loss": 0.0008,
+      "step": 10300
+    },
+    {
+      "epoch": 50.12,
+      "learning_rate": 4.962564102564103e-06,
+      "loss": 0.0008,
+      "step": 10325
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 4.949743589743591e-06,
+      "loss": 0.0005,
+      "step": 10350
+    },
+    {
+      "epoch": 50.36,
+      "learning_rate": 4.936923076923077e-06,
+      "loss": 0.0008,
+      "step": 10375
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 4.924102564102565e-06,
+      "loss": 0.0005,
+      "step": 10400
+    },
+    {
+      "epoch": 50.61,
+      "learning_rate": 4.911282051282052e-06,
+      "loss": 0.0014,
+      "step": 10425
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 4.898461538461539e-06,
+      "loss": 0.0019,
+      "step": 10450
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 4.8856410256410256e-06,
+      "loss": 0.0012,
+      "step": 10475
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 4.872820512820513e-06,
+      "loss": 0.0011,
+      "step": 10500
+    },
+    {
+      "epoch": 51.09,
+      "learning_rate": 4.86e-06,
+      "loss": 0.0007,
+      "step": 10525
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 4.847179487179488e-06,
+      "loss": 0.0006,
+      "step": 10550
+    },
+    {
+      "epoch": 51.33,
+      "learning_rate": 4.8343589743589746e-06,
+      "loss": 0.0006,
+      "step": 10575
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 4.821538461538462e-06,
+      "loss": 0.0005,
+      "step": 10600
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 4.808717948717949e-06,
+      "loss": 0.0013,
+      "step": 10625
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 4.795897435897437e-06,
+      "loss": 0.0007,
+      "step": 10650
+    },
+    {
+      "epoch": 51.82,
+      "learning_rate": 4.7830769230769235e-06,
+      "loss": 0.0006,
+      "step": 10675
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 4.770256410256411e-06,
+      "loss": 0.0009,
+      "step": 10700
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 4.757435897435898e-06,
+      "loss": 0.0005,
+      "step": 10725
+    },
+    {
+      "epoch": 52.18,
+      "learning_rate": 4.744615384615385e-06,
+      "loss": 0.0009,
+      "step": 10750
+    },
+    {
+      "epoch": 52.31,
+      "learning_rate": 4.731794871794872e-06,
+      "loss": 0.0004,
+      "step": 10775
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 4.718974358974359e-06,
+      "loss": 0.0009,
+      "step": 10800
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 4.706153846153846e-06,
+      "loss": 0.0008,
+      "step": 10825
+    },
+    {
+      "epoch": 52.67,
+      "learning_rate": 4.693333333333334e-06,
+      "loss": 0.0007,
+      "step": 10850
+    },
+    {
+      "epoch": 52.79,
+      "learning_rate": 4.680512820512821e-06,
+      "loss": 0.001,
+      "step": 10875
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 4.667692307692308e-06,
+      "loss": 0.0009,
+      "step": 10900
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 4.654871794871795e-06,
+      "loss": 0.0013,
+      "step": 10925
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 4.642051282051283e-06,
+      "loss": 0.0007,
+      "step": 10950
+    },
+    {
+      "epoch": 53.28,
+      "learning_rate": 4.62923076923077e-06,
+      "loss": 0.0011,
+      "step": 10975
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 4.616410256410257e-06,
+      "loss": 0.0009,
+      "step": 11000
+    },
+    {
+      "epoch": 53.4,
+      "eval_loss": 0.30143803358078003,
+      "eval_runtime": 1398.8263,
+      "eval_samples_per_second": 4.712,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 11.3350074932156,
+      "step": 11000
+    },
+    {
+      "epoch": 53.52,
+      "learning_rate": 4.603589743589744e-06,
+      "loss": 0.0011,
+      "step": 11025
+    },
+    {
+      "epoch": 53.64,
+      "learning_rate": 4.590769230769232e-06,
+      "loss": 0.0004,
+      "step": 11050
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 4.577948717948718e-06,
+      "loss": 0.0004,
+      "step": 11075
+    },
+    {
+      "epoch": 53.88,
+      "learning_rate": 4.5651282051282055e-06,
+      "loss": 0.0008,
+      "step": 11100
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 4.552307692307692e-06,
+      "loss": 0.0013,
+      "step": 11125
+    },
+    {
+      "epoch": 54.13,
+      "learning_rate": 4.53948717948718e-06,
+      "loss": 0.0009,
+      "step": 11150
+    },
+    {
+      "epoch": 54.25,
+      "learning_rate": 4.526666666666667e-06,
+      "loss": 0.0003,
+      "step": 11175
+    },
+    {
+      "epoch": 54.37,
+      "learning_rate": 4.5138461538461545e-06,
+      "loss": 0.0003,
+      "step": 11200
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 4.501025641025641e-06,
+      "loss": 0.0004,
+      "step": 11225
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 4.488205128205129e-06,
+      "loss": 0.0004,
+      "step": 11250
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 4.475384615384616e-06,
+      "loss": 0.0002,
+      "step": 11275
+    },
+    {
+      "epoch": 54.85,
+      "learning_rate": 4.4625641025641035e-06,
+      "loss": 0.0006,
+      "step": 11300
+    },
+    {
+      "epoch": 54.98,
+      "learning_rate": 4.44974358974359e-06,
+      "loss": 0.0007,
+      "step": 11325
+    },
+    {
+      "epoch": 55.1,
+      "learning_rate": 4.436923076923078e-06,
+      "loss": 0.0003,
+      "step": 11350
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 4.424102564102564e-06,
+      "loss": 0.0002,
+      "step": 11375
+    },
+    {
+      "epoch": 55.34,
+      "learning_rate": 4.411282051282052e-06,
+      "loss": 0.0001,
+      "step": 11400
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 4.3984615384615384e-06,
+      "loss": 0.0005,
+      "step": 11425
+    },
+    {
+      "epoch": 55.58,
+      "learning_rate": 4.385641025641026e-06,
+      "loss": 0.0003,
+      "step": 11450
+    },
+    {
+      "epoch": 55.7,
+      "learning_rate": 4.372820512820513e-06,
+      "loss": 0.0003,
+      "step": 11475
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 4.360000000000001e-06,
+      "loss": 0.0005,
+      "step": 11500
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 4.3471794871794874e-06,
+      "loss": 0.0006,
+      "step": 11525
+    },
+    {
+      "epoch": 56.07,
+      "learning_rate": 4.334358974358975e-06,
+      "loss": 0.0003,
+      "step": 11550
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 4.321538461538462e-06,
+      "loss": 0.0003,
+      "step": 11575
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 4.30871794871795e-06,
+      "loss": 0.0002,
+      "step": 11600
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 4.2958974358974364e-06,
+      "loss": 0.0001,
+      "step": 11625
+    },
+    {
+      "epoch": 56.55,
+      "learning_rate": 4.283076923076924e-06,
+      "loss": 0.0003,
+      "step": 11650
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 4.27025641025641e-06,
+      "loss": 0.0002,
+      "step": 11675
+    },
+    {
+      "epoch": 56.8,
+      "learning_rate": 4.257435897435898e-06,
+      "loss": 0.0008,
+      "step": 11700
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 4.2446153846153846e-06,
+      "loss": 0.0005,
+      "step": 11725
+    },
+    {
+      "epoch": 57.04,
+      "learning_rate": 4.231794871794872e-06,
+      "loss": 0.0003,
+      "step": 11750
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 4.218974358974359e-06,
+      "loss": 0.0003,
+      "step": 11775
+    },
+    {
+      "epoch": 57.28,
+      "learning_rate": 4.206153846153847e-06,
+      "loss": 0.0002,
+      "step": 11800
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 4.1933333333333336e-06,
+      "loss": 0.0002,
+      "step": 11825
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 4.180512820512821e-06,
+      "loss": 0.0001,
+      "step": 11850
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 4.167692307692308e-06,
+      "loss": 0.0005,
+      "step": 11875
+    },
+    {
+      "epoch": 57.77,
+      "learning_rate": 4.154871794871796e-06,
+      "loss": 0.0001,
+      "step": 11900
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 4.1420512820512826e-06,
+      "loss": 0.0001,
+      "step": 11925
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 4.12974358974359e-06,
+      "loss": 0.0005,
+      "step": 11950
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 4.116923076923078e-06,
+      "loss": 0.0012,
+      "step": 11975
+    },
+    {
+      "epoch": 58.25,
+      "learning_rate": 4.1041025641025645e-06,
+      "loss": 0.0011,
+      "step": 12000
+    },
+    {
+      "epoch": 58.25,
+      "eval_loss": 0.30517110228538513,
+      "eval_runtime": 1398.4457,
+      "eval_samples_per_second": 4.713,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 11.047430029567824,
+      "step": 12000
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 4.091282051282051e-06,
+      "loss": 0.001,
+      "step": 12025
+    },
+    {
+      "epoch": 58.5,
+      "learning_rate": 4.078461538461539e-06,
+      "loss": 0.0004,
+      "step": 12050
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 4.065641025641026e-06,
+      "loss": 0.0002,
+      "step": 12075
+    },
+    {
+      "epoch": 58.74,
+      "learning_rate": 4.0528205128205135e-06,
+      "loss": 0.0013,
+      "step": 12100
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 4.04e-06,
+      "loss": 0.0007,
+      "step": 12125
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 4.027179487179487e-06,
+      "loss": 0.0003,
+      "step": 12150
+    },
+    {
+      "epoch": 59.1,
+      "learning_rate": 4.014358974358975e-06,
+      "loss": 0.0003,
+      "step": 12175
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 4.001538461538462e-06,
+      "loss": 0.0003,
+      "step": 12200
+    },
+    {
+      "epoch": 59.34,
+      "learning_rate": 3.988717948717949e-06,
+      "loss": 0.0002,
+      "step": 12225
+    },
+    {
+      "epoch": 59.47,
+      "learning_rate": 3.975897435897436e-06,
+      "loss": 0.0001,
+      "step": 12250
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 3.963076923076924e-06,
+      "loss": 0.0003,
+      "step": 12275
+    },
+    {
+      "epoch": 59.71,
+      "learning_rate": 3.950256410256411e-06,
+      "loss": 0.0009,
+      "step": 12300
+    },
+    {
+      "epoch": 59.83,
+      "learning_rate": 3.937435897435897e-06,
+      "loss": 0.0012,
+      "step": 12325
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 3.924615384615385e-06,
+      "loss": 0.0005,
+      "step": 12350
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 3.911794871794872e-06,
+      "loss": 0.0009,
+      "step": 12375
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 3.89897435897436e-06,
+      "loss": 0.0007,
+      "step": 12400
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 3.886153846153846e-06,
+      "loss": 0.0011,
+      "step": 12425
+    },
+    {
+      "epoch": 60.44,
+      "learning_rate": 3.873333333333333e-06,
+      "loss": 0.001,
+      "step": 12450
+    },
+    {
+      "epoch": 60.56,
+      "learning_rate": 3.860512820512821e-06,
+      "loss": 0.0007,
+      "step": 12475
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 3.847692307692308e-06,
+      "loss": 0.0006,
+      "step": 12500
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 3.834871794871795e-06,
+      "loss": 0.0007,
+      "step": 12525
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 3.822051282051282e-06,
+      "loss": 0.0004,
+      "step": 12550
+    },
+    {
+      "epoch": 61.04,
+      "learning_rate": 3.8092307692307695e-06,
+      "loss": 0.0002,
+      "step": 12575
+    },
+    {
+      "epoch": 61.17,
+      "learning_rate": 3.7964102564102567e-06,
+      "loss": 0.0002,
+      "step": 12600
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 3.783589743589744e-06,
+      "loss": 0.0002,
+      "step": 12625
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 3.7707692307692312e-06,
+      "loss": 0.0003,
+      "step": 12650
+    },
+    {
+      "epoch": 61.53,
+      "learning_rate": 3.7579487179487185e-06,
+      "loss": 0.0001,
+      "step": 12675
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 3.7451282051282057e-06,
+      "loss": 0.0001,
+      "step": 12700
+    },
+    {
+      "epoch": 61.77,
+      "learning_rate": 3.732307692307693e-06,
+      "loss": 0.0002,
+      "step": 12725
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 3.7194871794871794e-06,
+      "loss": 0.0004,
+      "step": 12750
+    },
+    {
+      "epoch": 62.01,
+      "learning_rate": 3.7066666666666666e-06,
+      "loss": 0.0002,
+      "step": 12775
+    },
+    {
+      "epoch": 62.14,
+      "learning_rate": 3.693846153846154e-06,
+      "loss": 0.0001,
+      "step": 12800
+    },
+    {
+      "epoch": 62.26,
+      "learning_rate": 3.681025641025641e-06,
+      "loss": 0.0001,
+      "step": 12825
+    },
+    {
+      "epoch": 62.38,
+      "learning_rate": 3.6682051282051284e-06,
+      "loss": 0.0001,
+      "step": 12850
+    },
+    {
+      "epoch": 62.5,
+      "learning_rate": 3.6553846153846156e-06,
+      "loss": 0.0001,
+      "step": 12875
+    },
+    {
+      "epoch": 62.62,
+      "learning_rate": 3.642564102564103e-06,
+      "loss": 0.0001,
+      "step": 12900
+    },
+    {
+      "epoch": 62.74,
+      "learning_rate": 3.62974358974359e-06,
+      "loss": 0.0001,
+      "step": 12925
+    },
+    {
+      "epoch": 62.86,
+      "learning_rate": 3.6169230769230773e-06,
+      "loss": 0.0001,
+      "step": 12950
+    },
+    {
+      "epoch": 62.99,
+      "learning_rate": 3.6041025641025646e-06,
+      "loss": 0.0001,
+      "step": 12975
+    },
+    {
+      "epoch": 63.11,
+      "learning_rate": 3.591282051282052e-06,
+      "loss": 0.0001,
+      "step": 13000
+    },
+    {
+      "epoch": 63.11,
+      "eval_loss": 0.3204333782196045,
+      "eval_runtime": 1398.4876,
+      "eval_samples_per_second": 4.713,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.86921300984244,
+      "step": 13000
+    },
+    {
+      "epoch": 63.23,
+      "learning_rate": 3.578461538461539e-06,
+      "loss": 0.0001,
+      "step": 13025
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 3.5656410256410263e-06,
+      "loss": 0.0,
+      "step": 13050
+    },
+    {
+      "epoch": 63.47,
+      "learning_rate": 3.5528205128205127e-06,
+      "loss": 0.0,
+      "step": 13075
+    },
+    {
+      "epoch": 63.59,
+      "learning_rate": 3.54e-06,
+      "loss": 0.0,
+      "step": 13100
+    },
+    {
+      "epoch": 63.71,
+      "learning_rate": 3.5271794871794872e-06,
+      "loss": 0.0001,
+      "step": 13125
+    },
+    {
+      "epoch": 63.83,
+      "learning_rate": 3.5143589743589745e-06,
+      "loss": 0.0001,
+      "step": 13150
+    },
+    {
+      "epoch": 63.96,
+      "learning_rate": 3.5015384615384617e-06,
+      "loss": 0.0001,
+      "step": 13175
+    },
+    {
+      "epoch": 64.08,
+      "learning_rate": 3.488717948717949e-06,
+      "loss": 0.0,
+      "step": 13200
+    },
+    {
+      "epoch": 64.2,
+      "learning_rate": 3.4758974358974362e-06,
+      "loss": 0.0002,
+      "step": 13225
+    },
+    {
+      "epoch": 64.32,
+      "learning_rate": 3.4630769230769235e-06,
+      "loss": 0.0001,
+      "step": 13250
+    },
+    {
+      "epoch": 64.44,
+      "learning_rate": 3.4502564102564107e-06,
+      "loss": 0.0,
+      "step": 13275
+    },
+    {
+      "epoch": 64.56,
+      "learning_rate": 3.437435897435898e-06,
+      "loss": 0.0,
+      "step": 13300
+    },
+    {
+      "epoch": 64.68,
+      "learning_rate": 3.4246153846153852e-06,
+      "loss": 0.0001,
+      "step": 13325
+    },
+    {
+      "epoch": 64.81,
+      "learning_rate": 3.4117948717948725e-06,
+      "loss": 0.0,
+      "step": 13350
+    },
+    {
+      "epoch": 64.93,
+      "learning_rate": 3.398974358974359e-06,
+      "loss": 0.0001,
+      "step": 13375
+    },
+    {
+      "epoch": 65.05,
+      "learning_rate": 3.386153846153846e-06,
+      "loss": 0.0001,
+      "step": 13400
+    },
+    {
+      "epoch": 65.17,
+      "learning_rate": 3.3733333333333334e-06,
+      "loss": 0.0001,
+      "step": 13425
+    },
+    {
+      "epoch": 65.29,
+      "learning_rate": 3.3605128205128206e-06,
+      "loss": 0.0,
+      "step": 13450
+    },
+    {
+      "epoch": 65.41,
+      "learning_rate": 3.347692307692308e-06,
+      "loss": 0.0001,
+      "step": 13475
+    },
+    {
+      "epoch": 65.53,
+      "learning_rate": 3.334871794871795e-06,
+      "loss": 0.0001,
+      "step": 13500
+    },
+    {
+      "epoch": 65.66,
+      "learning_rate": 3.3220512820512824e-06,
+      "loss": 0.0,
+      "step": 13525
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 3.3092307692307696e-06,
+      "loss": 0.0,
+      "step": 13550
+    },
+    {
+      "epoch": 65.9,
+      "learning_rate": 3.296410256410257e-06,
+      "loss": 0.0,
+      "step": 13575
+    },
+    {
+      "epoch": 66.02,
+      "learning_rate": 3.283589743589744e-06,
+      "loss": 0.0,
+      "step": 13600
+    },
+    {
+      "epoch": 66.14,
+      "learning_rate": 3.2707692307692313e-06,
+      "loss": 0.0,
+      "step": 13625
+    },
+    {
+      "epoch": 66.26,
+      "learning_rate": 3.2579487179487186e-06,
+      "loss": 0.0,
+      "step": 13650
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 3.245128205128205e-06,
+      "loss": 0.0,
+      "step": 13675
+    },
+    {
+      "epoch": 66.5,
+      "learning_rate": 3.2323076923076922e-06,
+      "loss": 0.0,
+      "step": 13700
+    },
+    {
+      "epoch": 66.63,
+      "learning_rate": 3.2194871794871795e-06,
+      "loss": 0.0,
+      "step": 13725
+    },
+    {
+      "epoch": 66.75,
+      "learning_rate": 3.2066666666666667e-06,
+      "loss": 0.0,
+      "step": 13750
+    },
+    {
+      "epoch": 66.87,
+      "learning_rate": 3.193846153846154e-06,
+      "loss": 0.0,
+      "step": 13775
+    },
+    {
+      "epoch": 66.99,
+      "learning_rate": 3.1810256410256412e-06,
+      "loss": 0.0,
+      "step": 13800
+    },
+    {
+      "epoch": 67.11,
+      "learning_rate": 3.1682051282051285e-06,
+      "loss": 0.0,
+      "step": 13825
+    },
+    {
+      "epoch": 67.23,
+      "learning_rate": 3.1553846153846157e-06,
+      "loss": 0.0,
+      "step": 13850
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 3.142564102564103e-06,
+      "loss": 0.0,
+      "step": 13875
+    },
+    {
+      "epoch": 67.48,
+      "learning_rate": 3.1297435897435902e-06,
+      "loss": 0.0,
+      "step": 13900
+    },
+    {
+      "epoch": 67.6,
+      "learning_rate": 3.1169230769230775e-06,
+      "loss": 0.0,
+      "step": 13925
+    },
+    {
+      "epoch": 67.72,
+      "learning_rate": 3.1041025641025647e-06,
+      "loss": 0.0,
+      "step": 13950
+    },
+    {
+      "epoch": 67.84,
+      "learning_rate": 3.091282051282051e-06,
+      "loss": 0.0,
+      "step": 13975
+    },
+    {
+      "epoch": 67.96,
+      "learning_rate": 3.0784615384615384e-06,
+      "loss": 0.0,
+      "step": 14000
+    },
+    {
+      "epoch": 67.96,
+      "eval_loss": 0.3413444459438324,
+      "eval_runtime": 1398.7408,
+      "eval_samples_per_second": 4.712,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.709222730770788,
+      "step": 14000
+    },
+    {
+      "epoch": 68.08,
+      "learning_rate": 3.0656410256410256e-06,
+      "loss": 0.0,
+      "step": 14025
+    },
+    {
+      "epoch": 68.2,
+      "learning_rate": 3.052820512820513e-06,
+      "loss": 0.0,
+      "step": 14050
+    },
+    {
+      "epoch": 68.33,
+      "learning_rate": 3.04e-06,
+      "loss": 0.0,
+      "step": 14075
+    },
+    {
+      "epoch": 68.45,
+      "learning_rate": 3.0271794871794874e-06,
+      "loss": 0.0,
+      "step": 14100
+    },
+    {
+      "epoch": 68.57,
+      "learning_rate": 3.0143589743589746e-06,
+      "loss": 0.0,
+      "step": 14125
+    },
+    {
+      "epoch": 68.69,
+      "learning_rate": 3.001538461538462e-06,
+      "loss": 0.0,
+      "step": 14150
+    },
+    {
+      "epoch": 68.81,
+      "learning_rate": 2.988717948717949e-06,
+      "loss": 0.0,
+      "step": 14175
+    },
+    {
+      "epoch": 68.93,
+      "learning_rate": 2.9758974358974364e-06,
+      "loss": 0.0,
+      "step": 14200
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 2.9630769230769236e-06,
+      "loss": 0.0,
+      "step": 14225
+    },
+    {
+      "epoch": 69.17,
+      "learning_rate": 2.950256410256411e-06,
+      "loss": 0.0,
+      "step": 14250
+    },
+    {
+      "epoch": 69.3,
+      "learning_rate": 2.9374358974358973e-06,
+      "loss": 0.0,
+      "step": 14275
+    },
+    {
+      "epoch": 69.42,
+      "learning_rate": 2.9246153846153845e-06,
+      "loss": 0.0,
+      "step": 14300
+    },
+    {
+      "epoch": 69.54,
+      "learning_rate": 2.9117948717948717e-06,
+      "loss": 0.0,
+      "step": 14325
+    },
+    {
+      "epoch": 69.66,
+      "learning_rate": 2.898974358974359e-06,
+      "loss": 0.0,
+      "step": 14350
+    },
+    {
+      "epoch": 69.78,
+      "learning_rate": 2.8861538461538462e-06,
+      "loss": 0.0,
+      "step": 14375
+    },
+    {
+      "epoch": 69.9,
+      "learning_rate": 2.8733333333333335e-06,
+      "loss": 0.0,
+      "step": 14400
+    },
+    {
+      "epoch": 70.02,
+      "learning_rate": 2.8605128205128207e-06,
+      "loss": 0.0,
+      "step": 14425
+    },
+    {
+      "epoch": 70.15,
+      "learning_rate": 2.847692307692308e-06,
+      "loss": 0.0,
+      "step": 14450
+    },
+    {
+      "epoch": 70.27,
+      "learning_rate": 2.8348717948717952e-06,
+      "loss": 0.0,
+      "step": 14475
+    },
+    {
+      "epoch": 70.39,
+      "learning_rate": 2.8220512820512825e-06,
+      "loss": 0.0,
+      "step": 14500
+    },
+    {
+      "epoch": 70.51,
+      "learning_rate": 2.8092307692307697e-06,
+      "loss": 0.0,
+      "step": 14525
+    },
+    {
+      "epoch": 70.63,
+      "learning_rate": 2.796410256410257e-06,
+      "loss": 0.0,
+      "step": 14550
+    },
+    {
+      "epoch": 70.75,
+      "learning_rate": 2.7835897435897434e-06,
+      "loss": 0.0,
+      "step": 14575
+    },
+    {
+      "epoch": 70.87,
+      "learning_rate": 2.7707692307692306e-06,
+      "loss": 0.0,
+      "step": 14600
+    },
+    {
+      "epoch": 71.0,
+      "learning_rate": 2.757948717948718e-06,
+      "loss": 0.0,
+      "step": 14625
+    },
+    {
+      "epoch": 71.12,
+      "learning_rate": 2.745128205128205e-06,
+      "loss": 0.0,
+      "step": 14650
+    },
+    {
+      "epoch": 71.24,
+      "learning_rate": 2.7323076923076924e-06,
+      "loss": 0.0,
+      "step": 14675
+    },
+    {
+      "epoch": 71.36,
+      "learning_rate": 2.7194871794871796e-06,
+      "loss": 0.0,
+      "step": 14700
+    },
+    {
+      "epoch": 71.48,
+      "learning_rate": 2.706666666666667e-06,
+      "loss": 0.0,
+      "step": 14725
+    },
+    {
+      "epoch": 71.6,
+      "learning_rate": 2.693846153846154e-06,
+      "loss": 0.0,
+      "step": 14750
+    },
+    {
+      "epoch": 71.72,
+      "learning_rate": 2.6810256410256414e-06,
+      "loss": 0.0,
+      "step": 14775
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 2.6682051282051286e-06,
+      "loss": 0.0,
+      "step": 14800
+    },
+    {
+      "epoch": 71.97,
+      "learning_rate": 2.655384615384616e-06,
+      "loss": 0.0,
+      "step": 14825
+    },
+    {
+      "epoch": 72.09,
+      "learning_rate": 2.642564102564103e-06,
+      "loss": 0.0,
+      "step": 14850
+    },
+    {
+      "epoch": 72.21,
+      "learning_rate": 2.6297435897435904e-06,
+      "loss": 0.0,
+      "step": 14875
+    },
+    {
+      "epoch": 72.33,
+      "learning_rate": 2.6169230769230768e-06,
+      "loss": 0.0,
+      "step": 14900
+    },
+    {
+      "epoch": 72.45,
+      "learning_rate": 2.604102564102564e-06,
+      "loss": 0.0,
+      "step": 14925
+    },
+    {
+      "epoch": 72.57,
+      "learning_rate": 2.5912820512820513e-06,
+      "loss": 0.0,
+      "step": 14950
+    },
+    {
+      "epoch": 72.69,
+      "learning_rate": 2.5784615384615385e-06,
+      "loss": 0.0,
+      "step": 14975
+    },
+    {
+      "epoch": 72.82,
+      "learning_rate": 2.5656410256410257e-06,
+      "loss": 0.0,
+      "step": 15000
+    },
+    {
+      "epoch": 72.82,
+      "eval_loss": 0.3523952066898346,
+      "eval_runtime": 1397.4432,
+      "eval_samples_per_second": 4.716,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.664668475839441,
+      "step": 15000
+    },
+    {
+      "epoch": 72.94,
+      "learning_rate": 2.552820512820513e-06,
+      "loss": 0.0,
+      "step": 15025
+    },
+    {
+      "epoch": 73.06,
+      "learning_rate": 2.5400000000000002e-06,
+      "loss": 0.0,
+      "step": 15050
+    },
+    {
+      "epoch": 73.18,
+      "learning_rate": 2.5271794871794875e-06,
+      "loss": 0.0,
+      "step": 15075
+    },
+    {
+      "epoch": 73.3,
+      "learning_rate": 2.5143589743589747e-06,
+      "loss": 0.0,
+      "step": 15100
+    },
+    {
+      "epoch": 73.42,
+      "learning_rate": 2.501538461538462e-06,
+      "loss": 0.0,
+      "step": 15125
+    },
+    {
+      "epoch": 73.54,
+      "learning_rate": 2.488717948717949e-06,
+      "loss": 0.0,
+      "step": 15150
+    },
+    {
+      "epoch": 73.67,
+      "learning_rate": 2.475897435897436e-06,
+      "loss": 0.0,
+      "step": 15175
+    },
+    {
+      "epoch": 73.79,
+      "learning_rate": 2.4630769230769233e-06,
+      "loss": 0.0,
+      "step": 15200
+    },
+    {
+      "epoch": 73.91,
+      "learning_rate": 2.4502564102564106e-06,
+      "loss": 0.0,
+      "step": 15225
+    },
+    {
+      "epoch": 74.03,
+      "learning_rate": 2.437435897435898e-06,
+      "loss": 0.0,
+      "step": 15250
+    },
+    {
+      "epoch": 74.15,
+      "learning_rate": 2.4246153846153846e-06,
+      "loss": 0.0,
+      "step": 15275
+    },
+    {
+      "epoch": 74.27,
+      "learning_rate": 2.411794871794872e-06,
+      "loss": 0.0,
+      "step": 15300
+    },
+    {
+      "epoch": 74.39,
+      "learning_rate": 2.398974358974359e-06,
+      "loss": 0.0,
+      "step": 15325
+    },
+    {
+      "epoch": 74.51,
+      "learning_rate": 2.3861538461538464e-06,
+      "loss": 0.0,
+      "step": 15350
+    },
+    {
+      "epoch": 74.64,
+      "learning_rate": 2.3733333333333336e-06,
+      "loss": 0.0,
+      "step": 15375
+    },
+    {
+      "epoch": 74.76,
+      "learning_rate": 2.360512820512821e-06,
+      "loss": 0.0,
+      "step": 15400
+    },
+    {
+      "epoch": 74.88,
+      "learning_rate": 2.3476923076923077e-06,
+      "loss": 0.0,
+      "step": 15425
+    },
+    {
+      "epoch": 75.0,
+      "learning_rate": 2.334871794871795e-06,
+      "loss": 0.0,
+      "step": 15450
+    },
+    {
+      "epoch": 75.12,
+      "learning_rate": 2.322051282051282e-06,
+      "loss": 0.0,
+      "step": 15475
+    },
+    {
+      "epoch": 75.24,
+      "learning_rate": 2.3092307692307694e-06,
+      "loss": 0.0,
+      "step": 15500
+    },
+    {
+      "epoch": 75.36,
+      "learning_rate": 2.2964102564102567e-06,
+      "loss": 0.0,
+      "step": 15525
+    },
+    {
+      "epoch": 75.49,
+      "learning_rate": 2.283589743589744e-06,
+      "loss": 0.0,
+      "step": 15550
+    },
+    {
+      "epoch": 75.61,
+      "learning_rate": 2.2707692307692308e-06,
+      "loss": 0.0,
+      "step": 15575
+    },
+    {
+      "epoch": 75.73,
+      "learning_rate": 2.257948717948718e-06,
+      "loss": 0.0,
+      "step": 15600
+    },
+    {
+      "epoch": 75.85,
+      "learning_rate": 2.2451282051282053e-06,
+      "loss": 0.0,
+      "step": 15625
+    },
+    {
+      "epoch": 75.97,
+      "learning_rate": 2.2323076923076925e-06,
+      "loss": 0.0,
+      "step": 15650
+    },
+    {
+      "epoch": 76.09,
+      "learning_rate": 2.2194871794871797e-06,
+      "loss": 0.0,
+      "step": 15675
+    },
+    {
+      "epoch": 76.21,
+      "learning_rate": 2.206666666666667e-06,
+      "loss": 0.0,
+      "step": 15700
+    },
+    {
+      "epoch": 76.33,
+      "learning_rate": 2.193846153846154e-06,
+      "loss": 0.0,
+      "step": 15725
+    },
+    {
+      "epoch": 76.46,
+      "learning_rate": 2.181025641025641e-06,
+      "loss": 0.0,
+      "step": 15750
+    },
+    {
+      "epoch": 76.58,
+      "learning_rate": 2.1682051282051283e-06,
+      "loss": 0.0,
+      "step": 15775
+    },
+    {
+      "epoch": 76.7,
+      "learning_rate": 2.1553846153846156e-06,
+      "loss": 0.0,
+      "step": 15800
+    },
+    {
+      "epoch": 76.82,
+      "learning_rate": 2.142564102564103e-06,
+      "loss": 0.0,
+      "step": 15825
+    },
+    {
+      "epoch": 76.94,
+      "learning_rate": 2.12974358974359e-06,
+      "loss": 0.0,
+      "step": 15850
+    },
+    {
+      "epoch": 77.06,
+      "learning_rate": 2.116923076923077e-06,
+      "loss": 0.0,
+      "step": 15875
+    },
+    {
+      "epoch": 77.18,
+      "learning_rate": 2.104102564102564e-06,
+      "loss": 0.0,
+      "step": 15900
+    },
+    {
+      "epoch": 77.31,
+      "learning_rate": 2.0912820512820514e-06,
+      "loss": 0.0,
+      "step": 15925
+    },
+    {
+      "epoch": 77.43,
+      "learning_rate": 2.0784615384615386e-06,
+      "loss": 0.0,
+      "step": 15950
+    },
+    {
+      "epoch": 77.55,
+      "learning_rate": 2.065641025641026e-06,
+      "loss": 0.0,
+      "step": 15975
+    },
+    {
+      "epoch": 77.67,
+      "learning_rate": 2.052820512820513e-06,
+      "loss": 0.0,
+      "step": 16000
+    },
+    {
+      "epoch": 77.67,
+      "eval_loss": 0.36066606640815735,
+      "eval_runtime": 1397.551,
+      "eval_samples_per_second": 4.716,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.656567702215563,
+      "step": 16000
+    },
+    {
+      "epoch": 77.79,
+      "learning_rate": 2.04e-06,
+      "loss": 0.0,
+      "step": 16025
+    },
+    {
+      "epoch": 77.91,
+      "learning_rate": 2.027179487179487e-06,
+      "loss": 0.0,
+      "step": 16050
+    },
+    {
+      "epoch": 78.03,
+      "learning_rate": 2.0143589743589744e-06,
+      "loss": 0.0,
+      "step": 16075
+    },
+    {
+      "epoch": 78.16,
+      "learning_rate": 2.0015384615384617e-06,
+      "loss": 0.0,
+      "step": 16100
+    },
+    {
+      "epoch": 78.28,
+      "learning_rate": 1.988717948717949e-06,
+      "loss": 0.0,
+      "step": 16125
+    },
+    {
+      "epoch": 78.4,
+      "learning_rate": 1.975897435897436e-06,
+      "loss": 0.0,
+      "step": 16150
+    },
+    {
+      "epoch": 78.52,
+      "learning_rate": 1.963076923076923e-06,
+      "loss": 0.0,
+      "step": 16175
+    },
+    {
+      "epoch": 78.64,
+      "learning_rate": 1.9502564102564103e-06,
+      "loss": 0.0,
+      "step": 16200
+    },
+    {
+      "epoch": 78.76,
+      "learning_rate": 1.9374358974358975e-06,
+      "loss": 0.0,
+      "step": 16225
+    },
+    {
+      "epoch": 78.88,
+      "learning_rate": 1.9246153846153848e-06,
+      "loss": 0.0,
+      "step": 16250
+    },
+    {
+      "epoch": 79.0,
+      "learning_rate": 1.911794871794872e-06,
+      "loss": 0.0,
+      "step": 16275
+    },
+    {
+      "epoch": 79.13,
+      "learning_rate": 1.8989743589743593e-06,
+      "loss": 0.0,
+      "step": 16300
+    },
+    {
+      "epoch": 79.25,
+      "learning_rate": 1.8861538461538465e-06,
+      "loss": 0.0,
+      "step": 16325
+    },
+    {
+      "epoch": 79.37,
+      "learning_rate": 1.8733333333333333e-06,
+      "loss": 0.0,
+      "step": 16350
+    },
+    {
+      "epoch": 79.49,
+      "learning_rate": 1.8605128205128206e-06,
+      "loss": 0.0,
+      "step": 16375
+    },
+    {
+      "epoch": 79.61,
+      "learning_rate": 1.8476923076923078e-06,
+      "loss": 0.0,
+      "step": 16400
+    },
+    {
+      "epoch": 79.73,
+      "learning_rate": 1.834871794871795e-06,
+      "loss": 0.0,
+      "step": 16425
+    },
+    {
+      "epoch": 79.85,
+      "learning_rate": 1.8220512820512823e-06,
+      "loss": 0.0,
+      "step": 16450
+    },
+    {
+      "epoch": 79.98,
+      "learning_rate": 1.8092307692307696e-06,
+      "loss": 0.0,
+      "step": 16475
+    },
+    {
+      "epoch": 80.1,
+      "learning_rate": 1.7964102564102564e-06,
+      "loss": 0.0,
+      "step": 16500
+    },
+    {
+      "epoch": 80.22,
+      "learning_rate": 1.7835897435897436e-06,
+      "loss": 0.0,
+      "step": 16525
+    },
+    {
+      "epoch": 80.34,
+      "learning_rate": 1.7707692307692309e-06,
+      "loss": 0.0,
+      "step": 16550
+    },
+    {
+      "epoch": 80.46,
+      "learning_rate": 1.7579487179487181e-06,
+      "loss": 0.0,
+      "step": 16575
+    },
+    {
+      "epoch": 80.58,
+      "learning_rate": 1.7451282051282054e-06,
+      "loss": 0.0,
+      "step": 16600
+    },
+    {
+      "epoch": 80.7,
+      "learning_rate": 1.7323076923076926e-06,
+      "loss": 0.0,
+      "step": 16625
+    },
+    {
+      "epoch": 80.83,
+      "learning_rate": 1.7194871794871795e-06,
+      "loss": 0.0,
+      "step": 16650
+    },
+    {
+      "epoch": 80.95,
+      "learning_rate": 1.7066666666666667e-06,
+      "loss": 0.0,
+      "step": 16675
+    },
+    {
+      "epoch": 81.07,
+      "learning_rate": 1.693846153846154e-06,
+      "loss": 0.0,
+      "step": 16700
+    },
+    {
+      "epoch": 81.19,
+      "learning_rate": 1.6810256410256412e-06,
+      "loss": 0.0,
+      "step": 16725
+    },
+    {
+      "epoch": 81.31,
+      "learning_rate": 1.6682051282051284e-06,
+      "loss": 0.0,
+      "step": 16750
+    },
+    {
+      "epoch": 81.43,
+      "learning_rate": 1.6553846153846157e-06,
+      "loss": 0.0,
+      "step": 16775
+    },
+    {
+      "epoch": 81.55,
+      "learning_rate": 1.6425641025641025e-06,
+      "loss": 0.0,
+      "step": 16800
+    },
+    {
+      "epoch": 81.67,
+      "learning_rate": 1.6297435897435898e-06,
+      "loss": 0.0,
+      "step": 16825
+    },
+    {
+      "epoch": 81.8,
+      "learning_rate": 1.616923076923077e-06,
+      "loss": 0.0,
+      "step": 16850
+    },
+    {
+      "epoch": 81.92,
+      "learning_rate": 1.6041025641025643e-06,
+      "loss": 0.0,
+      "step": 16875
+    },
+    {
+      "epoch": 82.04,
+      "learning_rate": 1.5912820512820515e-06,
+      "loss": 0.0,
+      "step": 16900
+    },
+    {
+      "epoch": 82.16,
+      "learning_rate": 1.5784615384615388e-06,
+      "loss": 0.0,
+      "step": 16925
+    },
+    {
+      "epoch": 82.28,
+      "learning_rate": 1.5656410256410256e-06,
+      "loss": 0.0,
+      "step": 16950
+    },
+    {
+      "epoch": 82.4,
+      "learning_rate": 1.5528205128205128e-06,
+      "loss": 0.0,
+      "step": 16975
+    },
+    {
+      "epoch": 82.52,
+      "learning_rate": 1.54e-06,
+      "loss": 0.0,
+      "step": 17000
+    },
+    {
+      "epoch": 82.52,
+      "eval_loss": 0.36754557490348816,
+      "eval_runtime": 1397.4337,
+      "eval_samples_per_second": 4.717,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.612013447284216,
+      "step": 17000
+    },
+    {
+      "epoch": 82.65,
+      "learning_rate": 1.5271794871794873e-06,
+      "loss": 0.0,
+      "step": 17025
+    },
+    {
+      "epoch": 82.77,
+      "learning_rate": 1.5143589743589746e-06,
+      "loss": 0.0,
+      "step": 17050
+    },
+    {
+      "epoch": 82.89,
+      "learning_rate": 1.5015384615384618e-06,
+      "loss": 0.0,
+      "step": 17075
+    },
+    {
+      "epoch": 83.01,
+      "learning_rate": 1.4887179487179486e-06,
+      "loss": 0.0,
+      "step": 17100
+    },
+    {
+      "epoch": 83.13,
+      "learning_rate": 1.475897435897436e-06,
+      "loss": 0.0,
+      "step": 17125
+    },
+    {
+      "epoch": 83.25,
+      "learning_rate": 1.4630769230769231e-06,
+      "loss": 0.0,
+      "step": 17150
+    },
+    {
+      "epoch": 83.37,
+      "learning_rate": 1.4502564102564104e-06,
+      "loss": 0.0,
+      "step": 17175
+    },
+    {
+      "epoch": 83.5,
+      "learning_rate": 1.4374358974358976e-06,
+      "loss": 0.0,
+      "step": 17200
+    },
+    {
+      "epoch": 83.62,
+      "learning_rate": 1.4246153846153849e-06,
+      "loss": 0.0,
+      "step": 17225
+    },
+    {
+      "epoch": 83.74,
+      "learning_rate": 1.4117948717948717e-06,
+      "loss": 0.0,
+      "step": 17250
+    },
+    {
+      "epoch": 83.86,
+      "learning_rate": 1.398974358974359e-06,
+      "loss": 0.0,
+      "step": 17275
+    },
+    {
+      "epoch": 83.98,
+      "learning_rate": 1.3861538461538462e-06,
+      "loss": 0.0,
+      "step": 17300
+    },
+    {
+      "epoch": 84.1,
+      "learning_rate": 1.3733333333333335e-06,
+      "loss": 0.0,
+      "step": 17325
+    },
+    {
+      "epoch": 84.22,
+      "learning_rate": 1.3605128205128207e-06,
+      "loss": 0.0,
+      "step": 17350
+    },
+    {
+      "epoch": 84.34,
+      "learning_rate": 1.347692307692308e-06,
+      "loss": 0.0,
+      "step": 17375
+    },
+    {
+      "epoch": 84.47,
+      "learning_rate": 1.3348717948717948e-06,
+      "loss": 0.0,
+      "step": 17400
+    },
+    {
+      "epoch": 84.59,
+      "learning_rate": 1.322051282051282e-06,
+      "loss": 0.0,
+      "step": 17425
+    },
+    {
+      "epoch": 84.71,
+      "learning_rate": 1.3092307692307693e-06,
+      "loss": 0.0,
+      "step": 17450
+    },
+    {
+      "epoch": 84.83,
+      "learning_rate": 1.2964102564102565e-06,
+      "loss": 0.0,
+      "step": 17475
+    },
+    {
+      "epoch": 84.95,
+      "learning_rate": 1.2835897435897438e-06,
+      "loss": 0.0,
+      "step": 17500
+    },
+    {
+      "epoch": 85.07,
+      "learning_rate": 1.270769230769231e-06,
+      "loss": 0.0,
+      "step": 17525
+    },
+    {
+      "epoch": 85.19,
+      "learning_rate": 1.2579487179487178e-06,
+      "loss": 0.0,
+      "step": 17550
+    },
+    {
+      "epoch": 85.32,
+      "learning_rate": 1.2451282051282053e-06,
+      "loss": 0.0,
+      "step": 17575
+    },
+    {
+      "epoch": 85.44,
+      "learning_rate": 1.2323076923076923e-06,
+      "loss": 0.0,
+      "step": 17600
+    },
+    {
+      "epoch": 85.56,
+      "learning_rate": 1.2194871794871796e-06,
+      "loss": 0.0,
+      "step": 17625
+    },
+    {
+      "epoch": 85.68,
+      "learning_rate": 1.2066666666666668e-06,
+      "loss": 0.0,
+      "step": 17650
+    },
+    {
+      "epoch": 85.8,
+      "learning_rate": 1.1938461538461539e-06,
+      "loss": 0.0,
+      "step": 17675
+    },
+    {
+      "epoch": 85.92,
+      "learning_rate": 1.1810256410256411e-06,
+      "loss": 0.0,
+      "step": 17700
+    },
+    {
+      "epoch": 86.04,
+      "learning_rate": 1.1682051282051284e-06,
+      "loss": 0.0,
+      "step": 17725
+    },
+    {
+      "epoch": 86.17,
+      "learning_rate": 1.1553846153846154e-06,
+      "loss": 0.0,
+      "step": 17750
+    },
+    {
+      "epoch": 86.29,
+      "learning_rate": 1.1425641025641026e-06,
+      "loss": 0.0,
+      "step": 17775
+    },
+    {
+      "epoch": 86.41,
+      "learning_rate": 1.12974358974359e-06,
+      "loss": 0.0,
+      "step": 17800
+    },
+    {
+      "epoch": 86.53,
+      "learning_rate": 1.116923076923077e-06,
+      "loss": 0.0,
+      "step": 17825
+    },
+    {
+      "epoch": 86.65,
+      "learning_rate": 1.1041025641025642e-06,
+      "loss": 0.0,
+      "step": 17850
+    },
+    {
+      "epoch": 86.77,
+      "learning_rate": 1.0912820512820514e-06,
+      "loss": 0.0,
+      "step": 17875
+    },
+    {
+      "epoch": 86.89,
+      "learning_rate": 1.0784615384615385e-06,
+      "loss": 0.0,
+      "step": 17900
+    },
+    {
+      "epoch": 87.01,
+      "learning_rate": 1.0656410256410257e-06,
+      "loss": 0.0,
+      "step": 17925
+    },
+    {
+      "epoch": 87.14,
+      "learning_rate": 1.052820512820513e-06,
+      "loss": 0.0,
+      "step": 17950
+    },
+    {
+      "epoch": 87.26,
+      "learning_rate": 1.04e-06,
+      "loss": 0.0,
+      "step": 17975
+    },
+    {
+      "epoch": 87.38,
+      "learning_rate": 1.0271794871794872e-06,
+      "loss": 0.0,
+      "step": 18000
+    },
+    {
+      "epoch": 87.38,
+      "eval_loss": 0.3736671507358551,
+      "eval_runtime": 1397.7933,
+      "eval_samples_per_second": 4.715,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.614038640690186,
+      "step": 18000
+    },
+    {
+      "epoch": 87.5,
+      "learning_rate": 1.0143589743589745e-06,
+      "loss": 0.0,
+      "step": 18025
+    },
+    {
+      "epoch": 87.62,
+      "learning_rate": 1.0015384615384615e-06,
+      "loss": 0.0,
+      "step": 18050
+    },
+    {
+      "epoch": 87.74,
+      "learning_rate": 9.887179487179488e-07,
+      "loss": 0.0,
+      "step": 18075
+    },
+    {
+      "epoch": 87.86,
+      "learning_rate": 9.75897435897436e-07,
+      "loss": 0.0,
+      "step": 18100
+    },
+    {
+      "epoch": 87.99,
+      "learning_rate": 9.63076923076923e-07,
+      "loss": 0.0,
+      "step": 18125
+    },
+    {
+      "epoch": 88.11,
+      "learning_rate": 9.502564102564103e-07,
+      "loss": 0.0,
+      "step": 18150
+    },
+    {
+      "epoch": 88.23,
+      "learning_rate": 9.374358974358976e-07,
+      "loss": 0.0,
+      "step": 18175
+    },
+    {
+      "epoch": 88.35,
+      "learning_rate": 9.246153846153846e-07,
+      "loss": 0.0,
+      "step": 18200
+    },
+    {
+      "epoch": 88.47,
+      "learning_rate": 9.117948717948718e-07,
+      "loss": 0.0,
+      "step": 18225
+    },
+    {
+      "epoch": 88.59,
+      "learning_rate": 8.989743589743591e-07,
+      "loss": 0.0,
+      "step": 18250
+    },
+    {
+      "epoch": 88.71,
+      "learning_rate": 8.861538461538461e-07,
+      "loss": 0.0,
+      "step": 18275
+    },
+    {
+      "epoch": 88.83,
+      "learning_rate": 8.733333333333334e-07,
+      "loss": 0.0,
+      "step": 18300
+    },
+    {
+      "epoch": 88.96,
+      "learning_rate": 8.605128205128206e-07,
+      "loss": 0.0,
+      "step": 18325
+    },
+    {
+      "epoch": 89.08,
+      "learning_rate": 8.476923076923077e-07,
+      "loss": 0.0,
+      "step": 18350
+    },
+    {
+      "epoch": 89.2,
+      "learning_rate": 8.348717948717949e-07,
+      "loss": 0.0,
+      "step": 18375
+    },
+    {
+      "epoch": 89.32,
+      "learning_rate": 8.220512820512822e-07,
+      "loss": 0.0,
+      "step": 18400
+    },
+    {
+      "epoch": 89.44,
+      "learning_rate": 8.092307692307692e-07,
+      "loss": 0.0,
+      "step": 18425
+    },
+    {
+      "epoch": 89.56,
+      "learning_rate": 7.964102564102564e-07,
+      "loss": 0.0,
+      "step": 18450
+    },
+    {
+      "epoch": 89.68,
+      "learning_rate": 7.835897435897437e-07,
+      "loss": 0.0,
+      "step": 18475
+    },
+    {
+      "epoch": 89.81,
+      "learning_rate": 7.707692307692307e-07,
+      "loss": 0.0,
+      "step": 18500
+    },
+    {
+      "epoch": 89.93,
+      "learning_rate": 7.57948717948718e-07,
+      "loss": 0.0,
+      "step": 18525
+    },
+    {
+      "epoch": 90.05,
+      "learning_rate": 7.451282051282052e-07,
+      "loss": 0.0,
+      "step": 18550
+    },
+    {
+      "epoch": 90.17,
+      "learning_rate": 7.323076923076923e-07,
+      "loss": 0.0,
+      "step": 18575
+    },
+    {
+      "epoch": 90.29,
+      "learning_rate": 7.194871794871795e-07,
+      "loss": 0.0,
+      "step": 18600
+    },
+    {
+      "epoch": 90.41,
+      "learning_rate": 7.066666666666667e-07,
+      "loss": 0.0,
+      "step": 18625
+    },
+    {
+      "epoch": 90.53,
+      "learning_rate": 6.938461538461538e-07,
+      "loss": 0.0,
+      "step": 18650
+    },
+    {
+      "epoch": 90.66,
+      "learning_rate": 6.81025641025641e-07,
+      "loss": 0.0,
+      "step": 18675
+    },
+    {
+      "epoch": 90.78,
+      "learning_rate": 6.682051282051283e-07,
+      "loss": 0.0,
+      "step": 18700
+    },
+    {
+      "epoch": 90.9,
+      "learning_rate": 6.553846153846154e-07,
+      "loss": 0.0,
+      "step": 18725
+    },
+    {
+      "epoch": 91.02,
+      "learning_rate": 6.425641025641026e-07,
+      "loss": 0.0,
+      "step": 18750
+    },
+    {
+      "epoch": 91.14,
+      "learning_rate": 6.297435897435898e-07,
+      "loss": 0.0,
+      "step": 18775
+    },
+    {
+      "epoch": 91.26,
+      "learning_rate": 6.16923076923077e-07,
+      "loss": 0.0,
+      "step": 18800
+    },
+    {
+      "epoch": 91.38,
+      "learning_rate": 6.041025641025641e-07,
+      "loss": 0.0,
+      "step": 18825
+    },
+    {
+      "epoch": 91.5,
+      "learning_rate": 5.912820512820513e-07,
+      "loss": 0.0,
+      "step": 18850
+    },
+    {
+      "epoch": 91.63,
+      "learning_rate": 5.784615384615385e-07,
+      "loss": 0.0,
+      "step": 18875
+    },
+    {
+      "epoch": 91.75,
+      "learning_rate": 5.656410256410256e-07,
+      "loss": 0.0,
+      "step": 18900
+    },
+    {
+      "epoch": 91.87,
+      "learning_rate": 5.528205128205129e-07,
+      "loss": 0.0,
+      "step": 18925
+    },
+    {
+      "epoch": 91.99,
+      "learning_rate": 5.4e-07,
+      "loss": 0.0,
+      "step": 18950
+    },
+    {
+      "epoch": 92.11,
+      "learning_rate": 5.271794871794872e-07,
+      "loss": 0.0,
+      "step": 18975
+    },
+    {
+      "epoch": 92.23,
+      "learning_rate": 5.143589743589744e-07,
+      "loss": 0.0,
+      "step": 19000
+    },
+    {
+      "epoch": 92.23,
+      "eval_loss": 0.37819185853004456,
+      "eval_runtime": 1397.9151,
+      "eval_samples_per_second": 4.715,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.618089027502126,
+      "step": 19000
+    },
+    {
+      "epoch": 92.35,
+      "learning_rate": 5.015384615384615e-07,
+      "loss": 0.0,
+      "step": 19025
+    },
+    {
+      "epoch": 92.48,
+      "learning_rate": 4.887179487179487e-07,
+      "loss": 0.0,
+      "step": 19050
+    },
+    {
+      "epoch": 92.6,
+      "learning_rate": 4.7589743589743594e-07,
+      "loss": 0.0,
+      "step": 19075
+    },
+    {
+      "epoch": 92.72,
+      "learning_rate": 4.630769230769231e-07,
+      "loss": 0.0,
+      "step": 19100
+    },
+    {
+      "epoch": 92.84,
+      "learning_rate": 4.502564102564103e-07,
+      "loss": 0.0,
+      "step": 19125
+    },
+    {
+      "epoch": 92.96,
+      "learning_rate": 4.3743589743589747e-07,
+      "loss": 0.0,
+      "step": 19150
+    },
+    {
+      "epoch": 93.08,
+      "learning_rate": 4.246153846153846e-07,
+      "loss": 0.0,
+      "step": 19175
+    },
+    {
+      "epoch": 93.2,
+      "learning_rate": 4.1179487179487186e-07,
+      "loss": 0.0,
+      "step": 19200
+    },
+    {
+      "epoch": 93.33,
+      "learning_rate": 3.98974358974359e-07,
+      "loss": 0.0,
+      "step": 19225
+    },
+    {
+      "epoch": 93.45,
+      "learning_rate": 3.8615384615384615e-07,
+      "loss": 0.0,
+      "step": 19250
+    },
+    {
+      "epoch": 93.57,
+      "learning_rate": 3.733333333333334e-07,
+      "loss": 0.0,
+      "step": 19275
+    },
+    {
+      "epoch": 93.69,
+      "learning_rate": 3.6051282051282054e-07,
+      "loss": 0.0,
+      "step": 19300
+    },
+    {
+      "epoch": 93.81,
+      "learning_rate": 3.476923076923077e-07,
+      "loss": 0.0,
+      "step": 19325
+    },
+    {
+      "epoch": 93.93,
+      "learning_rate": 3.348717948717949e-07,
+      "loss": 0.0,
+      "step": 19350
+    },
+    {
+      "epoch": 94.05,
+      "learning_rate": 3.2205128205128207e-07,
+      "loss": 0.0,
+      "step": 19375
+    },
+    {
+      "epoch": 94.17,
+      "learning_rate": 3.0923076923076926e-07,
+      "loss": 0.0,
+      "step": 19400
+    },
+    {
+      "epoch": 94.3,
+      "learning_rate": 2.964102564102564e-07,
+      "loss": 0.0,
+      "step": 19425
+    },
+    {
+      "epoch": 94.42,
+      "learning_rate": 2.835897435897436e-07,
+      "loss": 0.0,
+      "step": 19450
+    },
+    {
+      "epoch": 94.54,
+      "learning_rate": 2.707692307692308e-07,
+      "loss": 0.0,
+      "step": 19475
+    },
+    {
+      "epoch": 94.66,
+      "learning_rate": 2.5794871794871794e-07,
+      "loss": 0.0,
+      "step": 19500
+    },
+    {
+      "epoch": 94.78,
+      "learning_rate": 2.4512820512820513e-07,
+      "loss": 0.0,
+      "step": 19525
+    },
+    {
+      "epoch": 94.9,
+      "learning_rate": 2.3230769230769233e-07,
+      "loss": 0.0,
+      "step": 19550
+    },
+    {
+      "epoch": 95.02,
+      "learning_rate": 2.194871794871795e-07,
+      "loss": 0.0,
+      "step": 19575
+    },
+    {
+      "epoch": 95.15,
+      "learning_rate": 2.066666666666667e-07,
+      "loss": 0.0,
+      "step": 19600
+    },
+    {
+      "epoch": 95.27,
+      "learning_rate": 1.9384615384615386e-07,
+      "loss": 0.0,
+      "step": 19625
+    },
+    {
+      "epoch": 95.39,
+      "learning_rate": 1.8102564102564103e-07,
+      "loss": 0.0,
+      "step": 19650
+    },
+    {
+      "epoch": 95.51,
+      "learning_rate": 1.6820512820512822e-07,
+      "loss": 0.0,
+      "step": 19675
+    },
+    {
+      "epoch": 95.63,
+      "learning_rate": 1.553846153846154e-07,
+      "loss": 0.0,
+      "step": 19700
+    },
+    {
+      "epoch": 95.75,
+      "learning_rate": 1.4256410256410259e-07,
+      "loss": 0.0,
+      "step": 19725
+    },
+    {
+      "epoch": 95.87,
+      "learning_rate": 1.2974358974358975e-07,
+      "loss": 0.0,
+      "step": 19750
+    },
+    {
+      "epoch": 96.0,
+      "learning_rate": 1.1692307692307694e-07,
+      "loss": 0.0,
+      "step": 19775
+    },
+    {
+      "epoch": 96.12,
+      "learning_rate": 1.041025641025641e-07,
+      "loss": 0.0,
+      "step": 19800
+    },
+    {
+      "epoch": 96.24,
+      "learning_rate": 9.128205128205129e-08,
+      "loss": 0.0,
+      "step": 19825
+    },
+    {
+      "epoch": 96.36,
+      "learning_rate": 7.846153846153847e-08,
+      "loss": 0.0,
+      "step": 19850
+    },
+    {
+      "epoch": 96.48,
+      "learning_rate": 6.564102564102564e-08,
+      "loss": 0.0,
+      "step": 19875
+    },
+    {
+      "epoch": 96.6,
+      "learning_rate": 5.2820512820512826e-08,
+      "loss": 0.0,
+      "step": 19900
+    },
+    {
+      "epoch": 96.72,
+      "learning_rate": 4e-08,
+      "loss": 0.0,
+      "step": 19925
+    },
+    {
+      "epoch": 96.84,
+      "learning_rate": 2.717948717948718e-08,
+      "loss": 0.0,
+      "step": 19950
+    },
+    {
+      "epoch": 96.97,
+      "learning_rate": 1.4358974358974361e-08,
+      "loss": 0.0,
+      "step": 19975
+    },
+    {
+      "epoch": 97.09,
+      "learning_rate": 1.5384615384615387e-09,
+      "loss": 0.0,
+      "step": 20000
+    },
+    {
+      "epoch": 97.09,
+      "eval_loss": 0.38027650117874146,
+      "eval_runtime": 1398.6239,
+      "eval_samples_per_second": 4.712,
+      "eval_steps_per_second": 0.295,
+      "eval_wer": 10.620114220908098,
+      "step": 20000
+    },
+    {
+      "epoch": 97.09,
+      "step": 20000,
+      "total_flos": 4.3481387177253274e+21,
+      "train_loss": 0.009814149574722979,
+      "train_runtime": 138976.8667,
+      "train_samples_per_second": 9.21,
+      "train_steps_per_second": 0.144
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 20000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 98,
+  "save_steps": 1000,
+  "total_flos": 4.3481387177253274e+21,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}