diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,5008 @@ +{ + "best_metric": 6.939845474613686, + "best_model_checkpoint": "./checkpoint-20000", + "epoch": 99.01, + "eval_steps": 1000, + "global_step": 20000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4.6000000000000004e-07, + "loss": 1.4118, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 9.600000000000001e-07, + "loss": 0.8278, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 1.46e-06, + "loss": 0.3975, + "step": 75 + }, + { + "epoch": 0.01, + "learning_rate": 1.9600000000000003e-06, + "loss": 0.3074, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 2.46e-06, + "loss": 0.2532, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 2.96e-06, + "loss": 0.2268, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 3.46e-06, + "loss": 0.2064, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 3.96e-06, + "loss": 0.1815, + "step": 200 + }, + { + "epoch": 1.0, + "learning_rate": 4.4600000000000005e-06, + "loss": 0.1705, + "step": 225 + }, + { + "epoch": 1.0, + "learning_rate": 4.960000000000001e-06, + "loss": 0.1821, + "step": 250 + }, + { + "epoch": 1.0, + "learning_rate": 5.460000000000001e-06, + "loss": 0.1507, + "step": 275 + }, + { + "epoch": 1.0, + "learning_rate": 5.9600000000000005e-06, + "loss": 0.1334, + "step": 300 + }, + { + "epoch": 1.01, + "learning_rate": 6.460000000000001e-06, + "loss": 0.1115, + "step": 325 + }, + { + "epoch": 1.01, + "learning_rate": 6.96e-06, + "loss": 0.1023, + "step": 350 + }, + { + "epoch": 1.01, + "learning_rate": 7.4600000000000006e-06, + "loss": 0.0901, + "step": 375 + }, + { + "epoch": 1.01, + "learning_rate": 7.960000000000002e-06, + "loss": 0.0855, + "step": 400 + }, + { + "epoch": 2.0, + "learning_rate": 8.46e-06, + "loss": 0.0718, + "step": 425 + }, + { + "epoch": 2.0, + "learning_rate": 8.96e-06, + "loss": 0.0922, + "step": 450 + }, + { + "epoch": 2.0, + "learning_rate": 9.460000000000001e-06, + "loss": 0.07, + "step": 475 + }, + { + "epoch": 2.0, + "learning_rate": 9.960000000000001e-06, + "loss": 0.0627, + "step": 500 + }, + { + "epoch": 2.01, + "learning_rate": 9.988205128205129e-06, + "loss": 0.0541, + "step": 525 + }, + { + "epoch": 2.01, + "learning_rate": 9.975384615384616e-06, + "loss": 0.0495, + "step": 550 + }, + { + "epoch": 2.01, + "learning_rate": 9.962564102564102e-06, + "loss": 0.0464, + "step": 575 + }, + { + "epoch": 2.01, + "learning_rate": 9.94974358974359e-06, + "loss": 0.0407, + "step": 600 + }, + { + "epoch": 3.0, + "learning_rate": 9.936923076923078e-06, + "loss": 0.0339, + "step": 625 + }, + { + "epoch": 3.0, + "learning_rate": 9.924102564102565e-06, + "loss": 0.0495, + "step": 650 + }, + { + "epoch": 3.0, + "learning_rate": 9.911282051282051e-06, + "loss": 0.0359, + "step": 675 + }, + { + "epoch": 3.0, + "learning_rate": 9.898461538461538e-06, + "loss": 0.0338, + "step": 700 + }, + { + "epoch": 3.01, + "learning_rate": 9.885641025641027e-06, + "loss": 0.0326, + "step": 725 + }, + { + "epoch": 3.01, + "learning_rate": 9.872820512820514e-06, + "loss": 0.0241, + "step": 750 + }, + { + "epoch": 3.01, + "learning_rate": 9.86e-06, + "loss": 0.0231, + "step": 775 + }, + { + "epoch": 3.01, + "learning_rate": 9.847179487179487e-06, + "loss": 0.0215, + "step": 800 + }, + { + "epoch": 4.0, + "learning_rate": 9.834358974358976e-06, + "loss": 0.0187, + "step": 825 + }, + { + "epoch": 4.0, + "learning_rate": 9.821538461538463e-06, + "loss": 0.0281, + "step": 850 + }, + { + "epoch": 4.0, + "learning_rate": 9.80871794871795e-06, + "loss": 0.0209, + "step": 875 + }, + { + "epoch": 4.0, + "learning_rate": 9.795897435897436e-06, + "loss": 0.0203, + "step": 900 + }, + { + "epoch": 4.01, + "learning_rate": 9.783076923076925e-06, + "loss": 0.0155, + "step": 925 + }, + { + "epoch": 4.01, + "learning_rate": 9.770256410256412e-06, + "loss": 0.0146, + "step": 950 + }, + { + "epoch": 4.01, + "learning_rate": 9.757435897435898e-06, + "loss": 0.0161, + "step": 975 + }, + { + "epoch": 4.01, + "learning_rate": 9.744615384615385e-06, + "loss": 0.0126, + "step": 1000 + }, + { + "epoch": 4.01, + "eval_loss": 0.21280750632286072, + "eval_runtime": 2275.147, + "eval_samples_per_second": 2.877, + "eval_steps_per_second": 0.18, + "eval_wer": 8.355753311258278, + "step": 1000 + }, + { + "epoch": 5.0, + "learning_rate": 9.731794871794872e-06, + "loss": 0.0133, + "step": 1025 + }, + { + "epoch": 5.0, + "learning_rate": 9.71897435897436e-06, + "loss": 0.0156, + "step": 1050 + }, + { + "epoch": 5.0, + "learning_rate": 9.706153846153847e-06, + "loss": 0.012, + "step": 1075 + }, + { + "epoch": 5.0, + "learning_rate": 9.693333333333334e-06, + "loss": 0.0117, + "step": 1100 + }, + { + "epoch": 5.01, + "learning_rate": 9.680512820512821e-06, + "loss": 0.0117, + "step": 1125 + }, + { + "epoch": 5.01, + "learning_rate": 9.667692307692308e-06, + "loss": 0.0101, + "step": 1150 + }, + { + "epoch": 5.01, + "learning_rate": 9.654871794871795e-06, + "loss": 0.0085, + "step": 1175 + }, + { + "epoch": 5.01, + "learning_rate": 9.642051282051282e-06, + "loss": 0.0065, + "step": 1200 + }, + { + "epoch": 6.0, + "learning_rate": 9.62923076923077e-06, + "loss": 0.0092, + "step": 1225 + }, + { + "epoch": 6.0, + "learning_rate": 9.616410256410257e-06, + "loss": 0.0111, + "step": 1250 + }, + { + "epoch": 6.0, + "learning_rate": 9.603589743589744e-06, + "loss": 0.009, + "step": 1275 + }, + { + "epoch": 6.0, + "learning_rate": 9.59076923076923e-06, + "loss": 0.0087, + "step": 1300 + }, + { + "epoch": 6.01, + "learning_rate": 9.577948717948719e-06, + "loss": 0.0082, + "step": 1325 + }, + { + "epoch": 6.01, + "learning_rate": 9.565128205128206e-06, + "loss": 0.008, + "step": 1350 + }, + { + "epoch": 6.01, + "learning_rate": 9.552307692307693e-06, + "loss": 0.007, + "step": 1375 + }, + { + "epoch": 6.01, + "learning_rate": 9.53948717948718e-06, + "loss": 0.0064, + "step": 1400 + }, + { + "epoch": 7.0, + "learning_rate": 9.526666666666668e-06, + "loss": 0.0069, + "step": 1425 + }, + { + "epoch": 7.0, + "learning_rate": 9.513846153846155e-06, + "loss": 0.0068, + "step": 1450 + }, + { + "epoch": 7.0, + "learning_rate": 9.501025641025642e-06, + "loss": 0.0061, + "step": 1475 + }, + { + "epoch": 7.0, + "learning_rate": 9.488205128205129e-06, + "loss": 0.0053, + "step": 1500 + }, + { + "epoch": 7.01, + "learning_rate": 9.475384615384617e-06, + "loss": 0.0041, + "step": 1525 + }, + { + "epoch": 7.01, + "learning_rate": 9.462564102564104e-06, + "loss": 0.0051, + "step": 1550 + }, + { + "epoch": 7.01, + "learning_rate": 9.44974358974359e-06, + "loss": 0.0049, + "step": 1575 + }, + { + "epoch": 7.01, + "learning_rate": 9.436923076923078e-06, + "loss": 0.0038, + "step": 1600 + }, + { + "epoch": 8.0, + "learning_rate": 9.424102564102564e-06, + "loss": 0.0039, + "step": 1625 + }, + { + "epoch": 8.0, + "learning_rate": 9.411282051282053e-06, + "loss": 0.0057, + "step": 1650 + }, + { + "epoch": 8.0, + "learning_rate": 9.39846153846154e-06, + "loss": 0.0053, + "step": 1675 + }, + { + "epoch": 8.01, + "learning_rate": 9.385641025641027e-06, + "loss": 0.0056, + "step": 1700 + }, + { + "epoch": 8.01, + "learning_rate": 9.372820512820513e-06, + "loss": 0.0053, + "step": 1725 + }, + { + "epoch": 8.01, + "learning_rate": 9.360000000000002e-06, + "loss": 0.0052, + "step": 1750 + }, + { + "epoch": 8.01, + "learning_rate": 9.347179487179487e-06, + "loss": 0.0042, + "step": 1775 + }, + { + "epoch": 8.01, + "learning_rate": 9.334358974358974e-06, + "loss": 0.0051, + "step": 1800 + }, + { + "epoch": 9.0, + "learning_rate": 9.321538461538462e-06, + "loss": 0.0048, + "step": 1825 + }, + { + "epoch": 9.0, + "learning_rate": 9.30871794871795e-06, + "loss": 0.006, + "step": 1850 + }, + { + "epoch": 9.0, + "learning_rate": 9.295897435897436e-06, + "loss": 0.0052, + "step": 1875 + }, + { + "epoch": 9.01, + "learning_rate": 9.283076923076923e-06, + "loss": 0.0059, + "step": 1900 + }, + { + "epoch": 9.01, + "learning_rate": 9.270256410256411e-06, + "loss": 0.0048, + "step": 1925 + }, + { + "epoch": 9.01, + "learning_rate": 9.257435897435898e-06, + "loss": 0.0065, + "step": 1950 + }, + { + "epoch": 9.01, + "learning_rate": 9.244615384615385e-06, + "loss": 0.0038, + "step": 1975 + }, + { + "epoch": 9.01, + "learning_rate": 9.231794871794872e-06, + "loss": 0.0032, + "step": 2000 + }, + { + "epoch": 9.01, + "eval_loss": 0.22615578770637512, + "eval_runtime": 2247.8561, + "eval_samples_per_second": 2.912, + "eval_steps_per_second": 0.182, + "eval_wer": 6.941570088300221, + "step": 2000 + }, + { + "epoch": 10.0, + "learning_rate": 9.21897435897436e-06, + "loss": 0.0035, + "step": 2025 + }, + { + "epoch": 10.0, + "learning_rate": 9.206153846153847e-06, + "loss": 0.0036, + "step": 2050 + }, + { + "epoch": 10.0, + "learning_rate": 9.193333333333334e-06, + "loss": 0.0028, + "step": 2075 + }, + { + "epoch": 10.01, + "learning_rate": 9.18051282051282e-06, + "loss": 0.0026, + "step": 2100 + }, + { + "epoch": 10.01, + "learning_rate": 9.16769230769231e-06, + "loss": 0.004, + "step": 2125 + }, + { + "epoch": 10.01, + "learning_rate": 9.154871794871796e-06, + "loss": 0.0038, + "step": 2150 + }, + { + "epoch": 10.01, + "learning_rate": 9.142051282051283e-06, + "loss": 0.0031, + "step": 2175 + }, + { + "epoch": 10.01, + "learning_rate": 9.12923076923077e-06, + "loss": 0.0025, + "step": 2200 + }, + { + "epoch": 11.0, + "learning_rate": 9.116410256410257e-06, + "loss": 0.0028, + "step": 2225 + }, + { + "epoch": 11.0, + "learning_rate": 9.103589743589745e-06, + "loss": 0.0043, + "step": 2250 + }, + { + "epoch": 11.0, + "learning_rate": 9.090769230769232e-06, + "loss": 0.0031, + "step": 2275 + }, + { + "epoch": 11.01, + "learning_rate": 9.077948717948719e-06, + "loss": 0.0027, + "step": 2300 + }, + { + "epoch": 11.01, + "learning_rate": 9.065128205128206e-06, + "loss": 0.0029, + "step": 2325 + }, + { + "epoch": 11.01, + "learning_rate": 9.052307692307694e-06, + "loss": 0.0026, + "step": 2350 + }, + { + "epoch": 11.01, + "learning_rate": 9.03948717948718e-06, + "loss": 0.0025, + "step": 2375 + }, + { + "epoch": 11.01, + "learning_rate": 9.026666666666666e-06, + "loss": 0.0025, + "step": 2400 + }, + { + "epoch": 12.0, + "learning_rate": 9.013846153846155e-06, + "loss": 0.0043, + "step": 2425 + }, + { + "epoch": 12.0, + "learning_rate": 9.001025641025641e-06, + "loss": 0.0034, + "step": 2450 + }, + { + "epoch": 12.0, + "learning_rate": 8.988205128205128e-06, + "loss": 0.0016, + "step": 2475 + }, + { + "epoch": 12.01, + "learning_rate": 8.975384615384615e-06, + "loss": 0.003, + "step": 2500 + }, + { + "epoch": 12.01, + "learning_rate": 8.962564102564104e-06, + "loss": 0.0025, + "step": 2525 + }, + { + "epoch": 12.01, + "learning_rate": 8.94974358974359e-06, + "loss": 0.002, + "step": 2550 + }, + { + "epoch": 12.01, + "learning_rate": 8.936923076923077e-06, + "loss": 0.0027, + "step": 2575 + }, + { + "epoch": 12.01, + "learning_rate": 8.924102564102564e-06, + "loss": 0.0024, + "step": 2600 + }, + { + "epoch": 13.0, + "learning_rate": 8.911282051282053e-06, + "loss": 0.0041, + "step": 2625 + }, + { + "epoch": 13.0, + "learning_rate": 8.89846153846154e-06, + "loss": 0.0043, + "step": 2650 + }, + { + "epoch": 13.0, + "learning_rate": 8.885641025641026e-06, + "loss": 0.0027, + "step": 2675 + }, + { + "epoch": 13.01, + "learning_rate": 8.872820512820513e-06, + "loss": 0.0042, + "step": 2700 + }, + { + "epoch": 13.01, + "learning_rate": 8.860000000000002e-06, + "loss": 0.0039, + "step": 2725 + }, + { + "epoch": 13.01, + "learning_rate": 8.847179487179488e-06, + "loss": 0.0048, + "step": 2750 + }, + { + "epoch": 13.01, + "learning_rate": 8.834358974358975e-06, + "loss": 0.0033, + "step": 2775 + }, + { + "epoch": 13.01, + "learning_rate": 8.821538461538462e-06, + "loss": 0.004, + "step": 2800 + }, + { + "epoch": 14.0, + "learning_rate": 8.80871794871795e-06, + "loss": 0.0043, + "step": 2825 + }, + { + "epoch": 14.0, + "learning_rate": 8.795897435897437e-06, + "loss": 0.0035, + "step": 2850 + }, + { + "epoch": 14.0, + "learning_rate": 8.783076923076924e-06, + "loss": 0.0049, + "step": 2875 + }, + { + "epoch": 14.01, + "learning_rate": 8.770256410256411e-06, + "loss": 0.0029, + "step": 2900 + }, + { + "epoch": 14.01, + "learning_rate": 8.757435897435898e-06, + "loss": 0.0032, + "step": 2925 + }, + { + "epoch": 14.01, + "learning_rate": 8.744615384615386e-06, + "loss": 0.0035, + "step": 2950 + }, + { + "epoch": 14.01, + "learning_rate": 8.731794871794873e-06, + "loss": 0.0025, + "step": 2975 + }, + { + "epoch": 14.01, + "learning_rate": 8.718974358974358e-06, + "loss": 0.0022, + "step": 3000 + }, + { + "epoch": 14.01, + "eval_loss": 0.25278234481811523, + "eval_runtime": 2226.9537, + "eval_samples_per_second": 2.939, + "eval_steps_per_second": 0.184, + "eval_wer": 7.112306843267108, + "step": 3000 + }, + { + "epoch": 15.0, + "learning_rate": 8.706153846153847e-06, + "loss": 0.0018, + "step": 3025 + }, + { + "epoch": 15.0, + "learning_rate": 8.693333333333334e-06, + "loss": 0.0029, + "step": 3050 + }, + { + "epoch": 15.0, + "learning_rate": 8.68051282051282e-06, + "loss": 0.004, + "step": 3075 + }, + { + "epoch": 15.01, + "learning_rate": 8.667692307692307e-06, + "loss": 0.0039, + "step": 3100 + }, + { + "epoch": 15.01, + "learning_rate": 8.654871794871796e-06, + "loss": 0.0022, + "step": 3125 + }, + { + "epoch": 15.01, + "learning_rate": 8.642051282051283e-06, + "loss": 0.0031, + "step": 3150 + }, + { + "epoch": 15.01, + "learning_rate": 8.62923076923077e-06, + "loss": 0.0029, + "step": 3175 + }, + { + "epoch": 15.01, + "learning_rate": 8.616410256410256e-06, + "loss": 0.003, + "step": 3200 + }, + { + "epoch": 16.0, + "learning_rate": 8.603589743589745e-06, + "loss": 0.0024, + "step": 3225 + }, + { + "epoch": 16.0, + "learning_rate": 8.590769230769232e-06, + "loss": 0.0036, + "step": 3250 + }, + { + "epoch": 16.0, + "learning_rate": 8.577948717948718e-06, + "loss": 0.0033, + "step": 3275 + }, + { + "epoch": 16.0, + "learning_rate": 8.565128205128205e-06, + "loss": 0.0029, + "step": 3300 + }, + { + "epoch": 16.01, + "learning_rate": 8.552307692307694e-06, + "loss": 0.0025, + "step": 3325 + }, + { + "epoch": 16.01, + "learning_rate": 8.53948717948718e-06, + "loss": 0.0032, + "step": 3350 + }, + { + "epoch": 16.01, + "learning_rate": 8.526666666666667e-06, + "loss": 0.0033, + "step": 3375 + }, + { + "epoch": 16.01, + "learning_rate": 8.513846153846154e-06, + "loss": 0.0026, + "step": 3400 + }, + { + "epoch": 17.0, + "learning_rate": 8.501025641025643e-06, + "loss": 0.0023, + "step": 3425 + }, + { + "epoch": 17.0, + "learning_rate": 8.48820512820513e-06, + "loss": 0.0035, + "step": 3450 + }, + { + "epoch": 17.0, + "learning_rate": 8.475384615384616e-06, + "loss": 0.0032, + "step": 3475 + }, + { + "epoch": 17.0, + "learning_rate": 8.462564102564103e-06, + "loss": 0.0033, + "step": 3500 + }, + { + "epoch": 17.01, + "learning_rate": 8.44974358974359e-06, + "loss": 0.0026, + "step": 3525 + }, + { + "epoch": 17.01, + "learning_rate": 8.436923076923079e-06, + "loss": 0.0028, + "step": 3550 + }, + { + "epoch": 17.01, + "learning_rate": 8.424102564102565e-06, + "loss": 0.0026, + "step": 3575 + }, + { + "epoch": 17.01, + "learning_rate": 8.411282051282052e-06, + "loss": 0.0025, + "step": 3600 + }, + { + "epoch": 18.0, + "learning_rate": 8.398461538461539e-06, + "loss": 0.0033, + "step": 3625 + }, + { + "epoch": 18.0, + "learning_rate": 8.385641025641026e-06, + "loss": 0.0022, + "step": 3650 + }, + { + "epoch": 18.0, + "learning_rate": 8.372820512820513e-06, + "loss": 0.0019, + "step": 3675 + }, + { + "epoch": 18.0, + "learning_rate": 8.36e-06, + "loss": 0.0026, + "step": 3700 + }, + { + "epoch": 18.01, + "learning_rate": 8.347179487179488e-06, + "loss": 0.0027, + "step": 3725 + }, + { + "epoch": 18.01, + "learning_rate": 8.334358974358975e-06, + "loss": 0.0022, + "step": 3750 + }, + { + "epoch": 18.01, + "learning_rate": 8.321538461538462e-06, + "loss": 0.0021, + "step": 3775 + }, + { + "epoch": 18.01, + "learning_rate": 8.308717948717949e-06, + "loss": 0.0024, + "step": 3800 + }, + { + "epoch": 19.0, + "learning_rate": 8.295897435897437e-06, + "loss": 0.0027, + "step": 3825 + }, + { + "epoch": 19.0, + "learning_rate": 8.283076923076924e-06, + "loss": 0.0025, + "step": 3850 + }, + { + "epoch": 19.0, + "learning_rate": 8.27025641025641e-06, + "loss": 0.002, + "step": 3875 + }, + { + "epoch": 19.0, + "learning_rate": 8.257435897435898e-06, + "loss": 0.0022, + "step": 3900 + }, + { + "epoch": 19.01, + "learning_rate": 8.244615384615386e-06, + "loss": 0.0019, + "step": 3925 + }, + { + "epoch": 19.01, + "learning_rate": 8.231794871794873e-06, + "loss": 0.0029, + "step": 3950 + }, + { + "epoch": 19.01, + "learning_rate": 8.21897435897436e-06, + "loss": 0.0032, + "step": 3975 + }, + { + "epoch": 19.01, + "learning_rate": 8.206153846153847e-06, + "loss": 0.0025, + "step": 4000 + }, + { + "epoch": 19.01, + "eval_loss": 0.26427945494651794, + "eval_runtime": 2239.2398, + "eval_samples_per_second": 2.923, + "eval_steps_per_second": 0.183, + "eval_wer": 7.364100441501104, + "step": 4000 + }, + { + "epoch": 20.0, + "learning_rate": 8.193333333333335e-06, + "loss": 0.0016, + "step": 4025 + }, + { + "epoch": 20.0, + "learning_rate": 8.180512820512822e-06, + "loss": 0.002, + "step": 4050 + }, + { + "epoch": 20.0, + "learning_rate": 8.167692307692309e-06, + "loss": 0.0015, + "step": 4075 + }, + { + "epoch": 20.0, + "learning_rate": 8.154871794871796e-06, + "loss": 0.0022, + "step": 4100 + }, + { + "epoch": 20.01, + "learning_rate": 8.142051282051282e-06, + "loss": 0.0029, + "step": 4125 + }, + { + "epoch": 20.01, + "learning_rate": 8.129230769230771e-06, + "loss": 0.0018, + "step": 4150 + }, + { + "epoch": 20.01, + "learning_rate": 8.116410256410258e-06, + "loss": 0.0028, + "step": 4175 + }, + { + "epoch": 20.01, + "learning_rate": 8.103589743589745e-06, + "loss": 0.002, + "step": 4200 + }, + { + "epoch": 21.0, + "learning_rate": 8.090769230769231e-06, + "loss": 0.0022, + "step": 4225 + }, + { + "epoch": 21.0, + "learning_rate": 8.077948717948718e-06, + "loss": 0.0031, + "step": 4250 + }, + { + "epoch": 21.0, + "learning_rate": 8.065128205128205e-06, + "loss": 0.002, + "step": 4275 + }, + { + "epoch": 21.0, + "learning_rate": 8.052307692307692e-06, + "loss": 0.0025, + "step": 4300 + }, + { + "epoch": 21.01, + "learning_rate": 8.03948717948718e-06, + "loss": 0.0013, + "step": 4325 + }, + { + "epoch": 21.01, + "learning_rate": 8.026666666666667e-06, + "loss": 0.0033, + "step": 4350 + }, + { + "epoch": 21.01, + "learning_rate": 8.013846153846154e-06, + "loss": 0.0026, + "step": 4375 + }, + { + "epoch": 21.01, + "learning_rate": 8.00102564102564e-06, + "loss": 0.0016, + "step": 4400 + }, + { + "epoch": 22.0, + "learning_rate": 7.98820512820513e-06, + "loss": 0.0019, + "step": 4425 + }, + { + "epoch": 22.0, + "learning_rate": 7.975384615384616e-06, + "loss": 0.0035, + "step": 4450 + }, + { + "epoch": 22.0, + "learning_rate": 7.962564102564103e-06, + "loss": 0.0022, + "step": 4475 + }, + { + "epoch": 22.0, + "learning_rate": 7.94974358974359e-06, + "loss": 0.0027, + "step": 4500 + }, + { + "epoch": 22.01, + "learning_rate": 7.936923076923078e-06, + "loss": 0.0024, + "step": 4525 + }, + { + "epoch": 22.01, + "learning_rate": 7.924102564102565e-06, + "loss": 0.0027, + "step": 4550 + }, + { + "epoch": 22.01, + "learning_rate": 7.911282051282052e-06, + "loss": 0.0026, + "step": 4575 + }, + { + "epoch": 22.01, + "learning_rate": 7.898461538461539e-06, + "loss": 0.0018, + "step": 4600 + }, + { + "epoch": 23.0, + "learning_rate": 7.885641025641027e-06, + "loss": 0.0014, + "step": 4625 + }, + { + "epoch": 23.0, + "learning_rate": 7.872820512820514e-06, + "loss": 0.0034, + "step": 4650 + }, + { + "epoch": 23.0, + "learning_rate": 7.860000000000001e-06, + "loss": 0.0025, + "step": 4675 + }, + { + "epoch": 23.0, + "learning_rate": 7.847179487179488e-06, + "loss": 0.0021, + "step": 4700 + }, + { + "epoch": 23.01, + "learning_rate": 7.834358974358975e-06, + "loss": 0.0022, + "step": 4725 + }, + { + "epoch": 23.01, + "learning_rate": 7.821538461538463e-06, + "loss": 0.0014, + "step": 4750 + }, + { + "epoch": 23.01, + "learning_rate": 7.80871794871795e-06, + "loss": 0.0018, + "step": 4775 + }, + { + "epoch": 23.01, + "learning_rate": 7.795897435897437e-06, + "loss": 0.0016, + "step": 4800 + }, + { + "epoch": 24.0, + "learning_rate": 7.783076923076924e-06, + "loss": 0.0009, + "step": 4825 + }, + { + "epoch": 24.0, + "learning_rate": 7.77025641025641e-06, + "loss": 0.0016, + "step": 4850 + }, + { + "epoch": 24.0, + "learning_rate": 7.757435897435897e-06, + "loss": 0.0007, + "step": 4875 + }, + { + "epoch": 24.0, + "learning_rate": 7.744615384615384e-06, + "loss": 0.0018, + "step": 4900 + }, + { + "epoch": 24.01, + "learning_rate": 7.731794871794873e-06, + "loss": 0.0014, + "step": 4925 + }, + { + "epoch": 24.01, + "learning_rate": 7.71897435897436e-06, + "loss": 0.0011, + "step": 4950 + }, + { + "epoch": 24.01, + "learning_rate": 7.706153846153846e-06, + "loss": 0.0018, + "step": 4975 + }, + { + "epoch": 24.01, + "learning_rate": 7.693333333333333e-06, + "loss": 0.0015, + "step": 5000 + }, + { + "epoch": 24.01, + "eval_loss": 0.2596248686313629, + "eval_runtime": 2223.9778, + "eval_samples_per_second": 2.943, + "eval_steps_per_second": 0.184, + "eval_wer": 7.336506622516556, + "step": 5000 + }, + { + "epoch": 25.0, + "learning_rate": 7.680512820512822e-06, + "loss": 0.0026, + "step": 5025 + }, + { + "epoch": 25.0, + "learning_rate": 7.667692307692308e-06, + "loss": 0.0018, + "step": 5050 + }, + { + "epoch": 25.0, + "learning_rate": 7.654871794871795e-06, + "loss": 0.0026, + "step": 5075 + }, + { + "epoch": 25.0, + "learning_rate": 7.642051282051282e-06, + "loss": 0.0023, + "step": 5100 + }, + { + "epoch": 25.01, + "learning_rate": 7.62923076923077e-06, + "loss": 0.0018, + "step": 5125 + }, + { + "epoch": 25.01, + "learning_rate": 7.616410256410257e-06, + "loss": 0.0011, + "step": 5150 + }, + { + "epoch": 25.01, + "learning_rate": 7.603589743589744e-06, + "loss": 0.0013, + "step": 5175 + }, + { + "epoch": 25.01, + "learning_rate": 7.590769230769232e-06, + "loss": 0.0018, + "step": 5200 + }, + { + "epoch": 26.0, + "learning_rate": 7.577948717948719e-06, + "loss": 0.0016, + "step": 5225 + }, + { + "epoch": 26.0, + "learning_rate": 7.565128205128206e-06, + "loss": 0.0023, + "step": 5250 + }, + { + "epoch": 26.0, + "learning_rate": 7.552307692307693e-06, + "loss": 0.0023, + "step": 5275 + }, + { + "epoch": 26.0, + "learning_rate": 7.53948717948718e-06, + "loss": 0.0024, + "step": 5300 + }, + { + "epoch": 26.01, + "learning_rate": 7.526666666666668e-06, + "loss": 0.0017, + "step": 5325 + }, + { + "epoch": 26.01, + "learning_rate": 7.5138461538461546e-06, + "loss": 0.0014, + "step": 5350 + }, + { + "epoch": 26.01, + "learning_rate": 7.501025641025642e-06, + "loss": 0.0019, + "step": 5375 + }, + { + "epoch": 26.01, + "learning_rate": 7.488205128205129e-06, + "loss": 0.0019, + "step": 5400 + }, + { + "epoch": 27.0, + "learning_rate": 7.475384615384616e-06, + "loss": 0.0014, + "step": 5425 + }, + { + "epoch": 27.0, + "learning_rate": 7.462564102564103e-06, + "loss": 0.002, + "step": 5450 + }, + { + "epoch": 27.0, + "learning_rate": 7.4497435897435895e-06, + "loss": 0.0012, + "step": 5475 + }, + { + "epoch": 27.0, + "learning_rate": 7.436923076923077e-06, + "loss": 0.001, + "step": 5500 + }, + { + "epoch": 27.01, + "learning_rate": 7.424102564102564e-06, + "loss": 0.0009, + "step": 5525 + }, + { + "epoch": 27.01, + "learning_rate": 7.411282051282052e-06, + "loss": 0.0012, + "step": 5550 + }, + { + "epoch": 27.01, + "learning_rate": 7.3984615384615385e-06, + "loss": 0.0013, + "step": 5575 + }, + { + "epoch": 27.01, + "learning_rate": 7.385641025641026e-06, + "loss": 0.0019, + "step": 5600 + }, + { + "epoch": 28.0, + "learning_rate": 7.372820512820513e-06, + "loss": 0.0015, + "step": 5625 + }, + { + "epoch": 28.0, + "learning_rate": 7.360000000000001e-06, + "loss": 0.002, + "step": 5650 + }, + { + "epoch": 28.0, + "learning_rate": 7.3471794871794875e-06, + "loss": 0.0008, + "step": 5675 + }, + { + "epoch": 28.0, + "learning_rate": 7.334358974358975e-06, + "loss": 0.0014, + "step": 5700 + }, + { + "epoch": 28.01, + "learning_rate": 7.321538461538462e-06, + "loss": 0.0011, + "step": 5725 + }, + { + "epoch": 28.01, + "learning_rate": 7.30871794871795e-06, + "loss": 0.0006, + "step": 5750 + }, + { + "epoch": 28.01, + "learning_rate": 7.2958974358974365e-06, + "loss": 0.0009, + "step": 5775 + }, + { + "epoch": 28.01, + "learning_rate": 7.283076923076924e-06, + "loss": 0.0018, + "step": 5800 + }, + { + "epoch": 29.0, + "learning_rate": 7.270256410256411e-06, + "loss": 0.001, + "step": 5825 + }, + { + "epoch": 29.0, + "learning_rate": 7.257435897435899e-06, + "loss": 0.0013, + "step": 5850 + }, + { + "epoch": 29.0, + "learning_rate": 7.2446153846153855e-06, + "loss": 0.0028, + "step": 5875 + }, + { + "epoch": 29.0, + "learning_rate": 7.231794871794872e-06, + "loss": 0.0024, + "step": 5900 + }, + { + "epoch": 29.01, + "learning_rate": 7.21897435897436e-06, + "loss": 0.0018, + "step": 5925 + }, + { + "epoch": 29.01, + "learning_rate": 7.206153846153847e-06, + "loss": 0.0011, + "step": 5950 + }, + { + "epoch": 29.01, + "learning_rate": 7.1933333333333345e-06, + "loss": 0.0022, + "step": 5975 + }, + { + "epoch": 29.01, + "learning_rate": 7.180512820512821e-06, + "loss": 0.0014, + "step": 6000 + }, + { + "epoch": 29.01, + "eval_loss": 0.2723470628261566, + "eval_runtime": 2244.0399, + "eval_samples_per_second": 2.917, + "eval_steps_per_second": 0.183, + "eval_wer": 7.63658940397351, + "step": 6000 + }, + { + "epoch": 30.0, + "learning_rate": 7.167692307692308e-06, + "loss": 0.0031, + "step": 6025 + }, + { + "epoch": 30.0, + "learning_rate": 7.154871794871795e-06, + "loss": 0.0025, + "step": 6050 + }, + { + "epoch": 30.0, + "learning_rate": 7.142051282051282e-06, + "loss": 0.0025, + "step": 6075 + }, + { + "epoch": 30.0, + "learning_rate": 7.1292307692307695e-06, + "loss": 0.0018, + "step": 6100 + }, + { + "epoch": 30.01, + "learning_rate": 7.116410256410256e-06, + "loss": 0.0016, + "step": 6125 + }, + { + "epoch": 30.01, + "learning_rate": 7.103589743589744e-06, + "loss": 0.0012, + "step": 6150 + }, + { + "epoch": 30.01, + "learning_rate": 7.090769230769231e-06, + "loss": 0.0013, + "step": 6175 + }, + { + "epoch": 30.01, + "learning_rate": 7.0779487179487184e-06, + "loss": 0.0011, + "step": 6200 + }, + { + "epoch": 31.0, + "learning_rate": 7.065128205128205e-06, + "loss": 0.0015, + "step": 6225 + }, + { + "epoch": 31.0, + "learning_rate": 7.052307692307693e-06, + "loss": 0.0028, + "step": 6250 + }, + { + "epoch": 31.0, + "learning_rate": 7.03948717948718e-06, + "loss": 0.0016, + "step": 6275 + }, + { + "epoch": 31.0, + "learning_rate": 7.0266666666666674e-06, + "loss": 0.0019, + "step": 6300 + }, + { + "epoch": 31.01, + "learning_rate": 7.013846153846154e-06, + "loss": 0.002, + "step": 6325 + }, + { + "epoch": 31.01, + "learning_rate": 7.001025641025642e-06, + "loss": 0.0022, + "step": 6350 + }, + { + "epoch": 31.01, + "learning_rate": 6.988205128205129e-06, + "loss": 0.0024, + "step": 6375 + }, + { + "epoch": 31.01, + "learning_rate": 6.9753846153846164e-06, + "loss": 0.0016, + "step": 6400 + }, + { + "epoch": 32.0, + "learning_rate": 6.962564102564103e-06, + "loss": 0.0022, + "step": 6425 + }, + { + "epoch": 32.0, + "learning_rate": 6.949743589743591e-06, + "loss": 0.0023, + "step": 6450 + }, + { + "epoch": 32.0, + "learning_rate": 6.936923076923078e-06, + "loss": 0.0017, + "step": 6475 + }, + { + "epoch": 32.01, + "learning_rate": 6.9241025641025646e-06, + "loss": 0.0018, + "step": 6500 + }, + { + "epoch": 32.01, + "learning_rate": 6.911282051282052e-06, + "loss": 0.0016, + "step": 6525 + }, + { + "epoch": 32.01, + "learning_rate": 6.898461538461539e-06, + "loss": 0.0018, + "step": 6550 + }, + { + "epoch": 32.01, + "learning_rate": 6.885641025641027e-06, + "loss": 0.0015, + "step": 6575 + }, + { + "epoch": 32.01, + "learning_rate": 6.8728205128205136e-06, + "loss": 0.002, + "step": 6600 + }, + { + "epoch": 33.0, + "learning_rate": 6.860000000000001e-06, + "loss": 0.0024, + "step": 6625 + }, + { + "epoch": 33.0, + "learning_rate": 6.847179487179487e-06, + "loss": 0.0017, + "step": 6650 + }, + { + "epoch": 33.0, + "learning_rate": 6.834358974358974e-06, + "loss": 0.0025, + "step": 6675 + }, + { + "epoch": 33.01, + "learning_rate": 6.821538461538462e-06, + "loss": 0.0016, + "step": 6700 + }, + { + "epoch": 33.01, + "learning_rate": 6.8087179487179485e-06, + "loss": 0.0017, + "step": 6725 + }, + { + "epoch": 33.01, + "learning_rate": 6.795897435897436e-06, + "loss": 0.0017, + "step": 6750 + }, + { + "epoch": 33.01, + "learning_rate": 6.783076923076923e-06, + "loss": 0.0019, + "step": 6775 + }, + { + "epoch": 33.01, + "learning_rate": 6.770256410256411e-06, + "loss": 0.0018, + "step": 6800 + }, + { + "epoch": 34.0, + "learning_rate": 6.7574358974358975e-06, + "loss": 0.0013, + "step": 6825 + }, + { + "epoch": 34.0, + "learning_rate": 6.744615384615385e-06, + "loss": 0.0022, + "step": 6850 + }, + { + "epoch": 34.0, + "learning_rate": 6.731794871794872e-06, + "loss": 0.0015, + "step": 6875 + }, + { + "epoch": 34.01, + "learning_rate": 6.71897435897436e-06, + "loss": 0.0023, + "step": 6900 + }, + { + "epoch": 34.01, + "learning_rate": 6.7061538461538465e-06, + "loss": 0.0011, + "step": 6925 + }, + { + "epoch": 34.01, + "learning_rate": 6.693333333333334e-06, + "loss": 0.0028, + "step": 6950 + }, + { + "epoch": 34.01, + "learning_rate": 6.680512820512821e-06, + "loss": 0.0022, + "step": 6975 + }, + { + "epoch": 34.01, + "learning_rate": 6.667692307692309e-06, + "loss": 0.0008, + "step": 7000 + }, + { + "epoch": 34.01, + "eval_loss": 0.2778206467628479, + "eval_runtime": 2237.0863, + "eval_samples_per_second": 2.926, + "eval_steps_per_second": 0.183, + "eval_wer": 7.6089955849889614, + "step": 7000 + }, + { + "epoch": 35.0, + "learning_rate": 6.6548717948717955e-06, + "loss": 0.0015, + "step": 7025 + }, + { + "epoch": 35.0, + "learning_rate": 6.642051282051283e-06, + "loss": 0.0016, + "step": 7050 + }, + { + "epoch": 35.0, + "learning_rate": 6.62923076923077e-06, + "loss": 0.0015, + "step": 7075 + }, + { + "epoch": 35.01, + "learning_rate": 6.616410256410257e-06, + "loss": 0.0019, + "step": 7100 + }, + { + "epoch": 35.01, + "learning_rate": 6.6035897435897445e-06, + "loss": 0.0012, + "step": 7125 + }, + { + "epoch": 35.01, + "learning_rate": 6.590769230769231e-06, + "loss": 0.0011, + "step": 7150 + }, + { + "epoch": 35.01, + "learning_rate": 6.577948717948719e-06, + "loss": 0.0012, + "step": 7175 + }, + { + "epoch": 35.01, + "learning_rate": 6.565128205128206e-06, + "loss": 0.0008, + "step": 7200 + }, + { + "epoch": 36.0, + "learning_rate": 6.5523076923076935e-06, + "loss": 0.0008, + "step": 7225 + }, + { + "epoch": 36.0, + "learning_rate": 6.5394871794871795e-06, + "loss": 0.0023, + "step": 7250 + }, + { + "epoch": 36.0, + "learning_rate": 6.526666666666666e-06, + "loss": 0.0006, + "step": 7275 + }, + { + "epoch": 36.01, + "learning_rate": 6.513846153846154e-06, + "loss": 0.0006, + "step": 7300 + }, + { + "epoch": 36.01, + "learning_rate": 6.501025641025641e-06, + "loss": 0.0015, + "step": 7325 + }, + { + "epoch": 36.01, + "learning_rate": 6.4882051282051285e-06, + "loss": 0.0006, + "step": 7350 + }, + { + "epoch": 36.01, + "learning_rate": 6.475384615384615e-06, + "loss": 0.0007, + "step": 7375 + }, + { + "epoch": 36.01, + "learning_rate": 6.462564102564103e-06, + "loss": 0.0007, + "step": 7400 + }, + { + "epoch": 37.0, + "learning_rate": 6.44974358974359e-06, + "loss": 0.0003, + "step": 7425 + }, + { + "epoch": 37.0, + "learning_rate": 6.4369230769230775e-06, + "loss": 0.0006, + "step": 7450 + }, + { + "epoch": 37.0, + "learning_rate": 6.424102564102564e-06, + "loss": 0.0004, + "step": 7475 + }, + { + "epoch": 37.01, + "learning_rate": 6.411282051282052e-06, + "loss": 0.0005, + "step": 7500 + }, + { + "epoch": 37.01, + "learning_rate": 6.398461538461539e-06, + "loss": 0.0007, + "step": 7525 + }, + { + "epoch": 37.01, + "learning_rate": 6.3856410256410264e-06, + "loss": 0.0005, + "step": 7550 + }, + { + "epoch": 37.01, + "learning_rate": 6.372820512820513e-06, + "loss": 0.0003, + "step": 7575 + }, + { + "epoch": 37.01, + "learning_rate": 6.360000000000001e-06, + "loss": 0.0006, + "step": 7600 + }, + { + "epoch": 38.0, + "learning_rate": 6.347179487179488e-06, + "loss": 0.0003, + "step": 7625 + }, + { + "epoch": 38.0, + "learning_rate": 6.3343589743589754e-06, + "loss": 0.0006, + "step": 7650 + }, + { + "epoch": 38.0, + "learning_rate": 6.321538461538462e-06, + "loss": 0.0002, + "step": 7675 + }, + { + "epoch": 38.01, + "learning_rate": 6.30871794871795e-06, + "loss": 0.0002, + "step": 7700 + }, + { + "epoch": 38.01, + "learning_rate": 6.295897435897437e-06, + "loss": 0.0005, + "step": 7725 + }, + { + "epoch": 38.01, + "learning_rate": 6.283076923076924e-06, + "loss": 0.0002, + "step": 7750 + }, + { + "epoch": 38.01, + "learning_rate": 6.270256410256411e-06, + "loss": 0.0002, + "step": 7775 + }, + { + "epoch": 38.01, + "learning_rate": 6.257435897435898e-06, + "loss": 0.0001, + "step": 7800 + }, + { + "epoch": 39.0, + "learning_rate": 6.244615384615386e-06, + "loss": 0.0001, + "step": 7825 + }, + { + "epoch": 39.0, + "learning_rate": 6.2317948717948726e-06, + "loss": 0.0002, + "step": 7850 + }, + { + "epoch": 39.0, + "learning_rate": 6.2189743589743586e-06, + "loss": 0.0002, + "step": 7875 + }, + { + "epoch": 39.01, + "learning_rate": 6.206153846153846e-06, + "loss": 0.0002, + "step": 7900 + }, + { + "epoch": 39.01, + "learning_rate": 6.193333333333333e-06, + "loss": 0.0001, + "step": 7925 + }, + { + "epoch": 39.01, + "learning_rate": 6.180512820512821e-06, + "loss": 0.0002, + "step": 7950 + }, + { + "epoch": 39.01, + "learning_rate": 6.1676923076923075e-06, + "loss": 0.0001, + "step": 7975 + }, + { + "epoch": 39.01, + "learning_rate": 6.154871794871795e-06, + "loss": 0.0003, + "step": 8000 + }, + { + "epoch": 39.01, + "eval_loss": 0.2879820168018341, + "eval_runtime": 2268.2813, + "eval_samples_per_second": 2.886, + "eval_steps_per_second": 0.181, + "eval_wer": 7.2261313465783665, + "step": 8000 + }, + { + "epoch": 40.0, + "learning_rate": 6.142051282051282e-06, + "loss": 0.0003, + "step": 8025 + }, + { + "epoch": 40.0, + "learning_rate": 6.12923076923077e-06, + "loss": 0.0002, + "step": 8050 + }, + { + "epoch": 40.0, + "learning_rate": 6.1164102564102565e-06, + "loss": 0.001, + "step": 8075 + }, + { + "epoch": 40.01, + "learning_rate": 6.103589743589744e-06, + "loss": 0.0011, + "step": 8100 + }, + { + "epoch": 40.01, + "learning_rate": 6.090769230769231e-06, + "loss": 0.0001, + "step": 8125 + }, + { + "epoch": 40.01, + "learning_rate": 6.077948717948719e-06, + "loss": 0.0001, + "step": 8150 + }, + { + "epoch": 40.01, + "learning_rate": 6.0651282051282055e-06, + "loss": 0.0002, + "step": 8175 + }, + { + "epoch": 40.01, + "learning_rate": 6.052307692307693e-06, + "loss": 0.0006, + "step": 8200 + }, + { + "epoch": 41.0, + "learning_rate": 6.03948717948718e-06, + "loss": 0.0004, + "step": 8225 + }, + { + "epoch": 41.0, + "learning_rate": 6.026666666666668e-06, + "loss": 0.0002, + "step": 8250 + }, + { + "epoch": 41.0, + "learning_rate": 6.0138461538461545e-06, + "loss": 0.0014, + "step": 8275 + }, + { + "epoch": 41.01, + "learning_rate": 6.001025641025642e-06, + "loss": 0.0009, + "step": 8300 + }, + { + "epoch": 41.01, + "learning_rate": 5.988205128205129e-06, + "loss": 0.0002, + "step": 8325 + }, + { + "epoch": 41.01, + "learning_rate": 5.975384615384616e-06, + "loss": 0.0002, + "step": 8350 + }, + { + "epoch": 41.01, + "learning_rate": 5.9625641025641035e-06, + "loss": 0.0003, + "step": 8375 + }, + { + "epoch": 41.01, + "learning_rate": 5.94974358974359e-06, + "loss": 0.0002, + "step": 8400 + }, + { + "epoch": 42.0, + "learning_rate": 5.936923076923078e-06, + "loss": 0.0001, + "step": 8425 + }, + { + "epoch": 42.0, + "learning_rate": 5.924102564102565e-06, + "loss": 0.0001, + "step": 8450 + }, + { + "epoch": 42.0, + "learning_rate": 5.911282051282052e-06, + "loss": 0.0001, + "step": 8475 + }, + { + "epoch": 42.01, + "learning_rate": 5.8984615384615385e-06, + "loss": 0.0001, + "step": 8500 + }, + { + "epoch": 42.01, + "learning_rate": 5.885641025641025e-06, + "loss": 0.0003, + "step": 8525 + }, + { + "epoch": 42.01, + "learning_rate": 5.872820512820513e-06, + "loss": 0.0001, + "step": 8550 + }, + { + "epoch": 42.01, + "learning_rate": 5.86e-06, + "loss": 0.0007, + "step": 8575 + }, + { + "epoch": 42.01, + "learning_rate": 5.8471794871794875e-06, + "loss": 0.0001, + "step": 8600 + }, + { + "epoch": 43.0, + "learning_rate": 5.834358974358974e-06, + "loss": 0.0001, + "step": 8625 + }, + { + "epoch": 43.0, + "learning_rate": 5.821538461538462e-06, + "loss": 0.0001, + "step": 8650 + }, + { + "epoch": 43.0, + "learning_rate": 5.808717948717949e-06, + "loss": 0.0001, + "step": 8675 + }, + { + "epoch": 43.01, + "learning_rate": 5.7958974358974365e-06, + "loss": 0.0001, + "step": 8700 + }, + { + "epoch": 43.01, + "learning_rate": 5.783076923076923e-06, + "loss": 0.0001, + "step": 8725 + }, + { + "epoch": 43.01, + "learning_rate": 5.770256410256411e-06, + "loss": 0.0001, + "step": 8750 + }, + { + "epoch": 43.01, + "learning_rate": 5.757435897435898e-06, + "loss": 0.0001, + "step": 8775 + }, + { + "epoch": 43.01, + "learning_rate": 5.7446153846153855e-06, + "loss": 0.0005, + "step": 8800 + }, + { + "epoch": 44.0, + "learning_rate": 5.731794871794872e-06, + "loss": 0.0001, + "step": 8825 + }, + { + "epoch": 44.0, + "learning_rate": 5.71897435897436e-06, + "loss": 0.0001, + "step": 8850 + }, + { + "epoch": 44.0, + "learning_rate": 5.706153846153847e-06, + "loss": 0.0004, + "step": 8875 + }, + { + "epoch": 44.01, + "learning_rate": 5.6933333333333344e-06, + "loss": 0.0002, + "step": 8900 + }, + { + "epoch": 44.01, + "learning_rate": 5.680512820512821e-06, + "loss": 0.0003, + "step": 8925 + }, + { + "epoch": 44.01, + "learning_rate": 5.667692307692308e-06, + "loss": 0.0002, + "step": 8950 + }, + { + "epoch": 44.01, + "learning_rate": 5.654871794871796e-06, + "loss": 0.0005, + "step": 8975 + }, + { + "epoch": 44.01, + "learning_rate": 5.642051282051283e-06, + "loss": 0.0004, + "step": 9000 + }, + { + "epoch": 44.01, + "eval_loss": 0.29199108481407166, + "eval_runtime": 2223.2313, + "eval_samples_per_second": 2.944, + "eval_steps_per_second": 0.184, + "eval_wer": 7.674530905077264, + "step": 9000 + }, + { + "epoch": 45.0, + "learning_rate": 5.62923076923077e-06, + "loss": 0.0009, + "step": 9025 + }, + { + "epoch": 45.0, + "learning_rate": 5.616410256410257e-06, + "loss": 0.0018, + "step": 9050 + }, + { + "epoch": 45.0, + "learning_rate": 5.603589743589744e-06, + "loss": 0.0027, + "step": 9075 + }, + { + "epoch": 45.01, + "learning_rate": 5.590769230769231e-06, + "loss": 0.0024, + "step": 9100 + }, + { + "epoch": 45.01, + "learning_rate": 5.5779487179487176e-06, + "loss": 0.0031, + "step": 9125 + }, + { + "epoch": 45.01, + "learning_rate": 5.565128205128205e-06, + "loss": 0.0035, + "step": 9150 + }, + { + "epoch": 45.01, + "learning_rate": 5.552307692307692e-06, + "loss": 0.0031, + "step": 9175 + }, + { + "epoch": 45.01, + "learning_rate": 5.53948717948718e-06, + "loss": 0.0033, + "step": 9200 + }, + { + "epoch": 46.0, + "learning_rate": 5.5266666666666666e-06, + "loss": 0.0019, + "step": 9225 + }, + { + "epoch": 46.0, + "learning_rate": 5.513846153846154e-06, + "loss": 0.0038, + "step": 9250 + }, + { + "epoch": 46.0, + "learning_rate": 5.501025641025641e-06, + "loss": 0.0034, + "step": 9275 + }, + { + "epoch": 46.01, + "learning_rate": 5.488205128205129e-06, + "loss": 0.0039, + "step": 9300 + }, + { + "epoch": 46.01, + "learning_rate": 5.4753846153846155e-06, + "loss": 0.003, + "step": 9325 + }, + { + "epoch": 46.01, + "learning_rate": 5.462564102564103e-06, + "loss": 0.0018, + "step": 9350 + }, + { + "epoch": 46.01, + "learning_rate": 5.44974358974359e-06, + "loss": 0.002, + "step": 9375 + }, + { + "epoch": 46.01, + "learning_rate": 5.436923076923078e-06, + "loss": 0.0031, + "step": 9400 + }, + { + "epoch": 47.0, + "learning_rate": 5.4241025641025645e-06, + "loss": 0.0018, + "step": 9425 + }, + { + "epoch": 47.0, + "learning_rate": 5.411282051282052e-06, + "loss": 0.0015, + "step": 9450 + }, + { + "epoch": 47.0, + "learning_rate": 5.398461538461539e-06, + "loss": 0.0015, + "step": 9475 + }, + { + "epoch": 47.01, + "learning_rate": 5.385641025641027e-06, + "loss": 0.0021, + "step": 9500 + }, + { + "epoch": 47.01, + "learning_rate": 5.3728205128205135e-06, + "loss": 0.0019, + "step": 9525 + }, + { + "epoch": 47.01, + "learning_rate": 5.36e-06, + "loss": 0.0017, + "step": 9550 + }, + { + "epoch": 47.01, + "learning_rate": 5.347179487179488e-06, + "loss": 0.0014, + "step": 9575 + }, + { + "epoch": 47.01, + "learning_rate": 5.334358974358975e-06, + "loss": 0.0014, + "step": 9600 + }, + { + "epoch": 48.0, + "learning_rate": 5.3215384615384625e-06, + "loss": 0.0009, + "step": 9625 + }, + { + "epoch": 48.0, + "learning_rate": 5.308717948717949e-06, + "loss": 0.0012, + "step": 9650 + }, + { + "epoch": 48.0, + "learning_rate": 5.295897435897437e-06, + "loss": 0.0014, + "step": 9675 + }, + { + "epoch": 48.01, + "learning_rate": 5.283076923076923e-06, + "loss": 0.0012, + "step": 9700 + }, + { + "epoch": 48.01, + "learning_rate": 5.27025641025641e-06, + "loss": 0.0005, + "step": 9725 + }, + { + "epoch": 48.01, + "learning_rate": 5.2574358974358975e-06, + "loss": 0.0006, + "step": 9750 + }, + { + "epoch": 48.01, + "learning_rate": 5.244615384615384e-06, + "loss": 0.0008, + "step": 9775 + }, + { + "epoch": 48.01, + "learning_rate": 5.231794871794872e-06, + "loss": 0.0008, + "step": 9800 + }, + { + "epoch": 49.0, + "learning_rate": 5.218974358974359e-06, + "loss": 0.0004, + "step": 9825 + }, + { + "epoch": 49.0, + "learning_rate": 5.2061538461538465e-06, + "loss": 0.0005, + "step": 9850 + }, + { + "epoch": 49.0, + "learning_rate": 5.193333333333333e-06, + "loss": 0.0012, + "step": 9875 + }, + { + "epoch": 49.01, + "learning_rate": 5.180512820512821e-06, + "loss": 0.0009, + "step": 9900 + }, + { + "epoch": 49.01, + "learning_rate": 5.167692307692308e-06, + "loss": 0.0002, + "step": 9925 + }, + { + "epoch": 49.01, + "learning_rate": 5.1548717948717955e-06, + "loss": 0.0002, + "step": 9950 + }, + { + "epoch": 49.01, + "learning_rate": 5.142051282051282e-06, + "loss": 0.0007, + "step": 9975 + }, + { + "epoch": 49.01, + "learning_rate": 5.12923076923077e-06, + "loss": 0.0001, + "step": 10000 + }, + { + "epoch": 49.01, + "eval_loss": 0.28544262051582336, + "eval_runtime": 2240.6894, + "eval_samples_per_second": 2.921, + "eval_steps_per_second": 0.183, + "eval_wer": 7.408940397350994, + "step": 10000 + }, + { + "epoch": 50.0, + "learning_rate": 5.116410256410257e-06, + "loss": 0.0003, + "step": 10025 + }, + { + "epoch": 50.0, + "learning_rate": 5.1035897435897445e-06, + "loss": 0.0004, + "step": 10050 + }, + { + "epoch": 50.0, + "learning_rate": 5.090769230769231e-06, + "loss": 0.0002, + "step": 10075 + }, + { + "epoch": 50.01, + "learning_rate": 5.077948717948719e-06, + "loss": 0.0002, + "step": 10100 + }, + { + "epoch": 50.01, + "learning_rate": 5.065128205128206e-06, + "loss": 0.0001, + "step": 10125 + }, + { + "epoch": 50.01, + "learning_rate": 5.052307692307693e-06, + "loss": 0.0001, + "step": 10150 + }, + { + "epoch": 50.01, + "learning_rate": 5.03948717948718e-06, + "loss": 0.0006, + "step": 10175 + }, + { + "epoch": 50.01, + "learning_rate": 5.026666666666667e-06, + "loss": 0.0005, + "step": 10200 + }, + { + "epoch": 51.0, + "learning_rate": 5.013846153846155e-06, + "loss": 0.0007, + "step": 10225 + }, + { + "epoch": 51.0, + "learning_rate": 5.001025641025642e-06, + "loss": 0.0003, + "step": 10250 + }, + { + "epoch": 51.0, + "learning_rate": 4.988205128205128e-06, + "loss": 0.0002, + "step": 10275 + }, + { + "epoch": 51.01, + "learning_rate": 4.975384615384616e-06, + "loss": 0.0001, + "step": 10300 + }, + { + "epoch": 51.01, + "learning_rate": 4.962564102564103e-06, + "loss": 0.0001, + "step": 10325 + }, + { + "epoch": 51.01, + "learning_rate": 4.949743589743591e-06, + "loss": 0.0001, + "step": 10350 + }, + { + "epoch": 51.01, + "learning_rate": 4.936923076923077e-06, + "loss": 0.0002, + "step": 10375 + }, + { + "epoch": 51.01, + "learning_rate": 4.924102564102565e-06, + "loss": 0.0002, + "step": 10400 + }, + { + "epoch": 52.0, + "learning_rate": 4.911282051282052e-06, + "loss": 0.0001, + "step": 10425 + }, + { + "epoch": 52.0, + "learning_rate": 4.898461538461539e-06, + "loss": 0.0001, + "step": 10450 + }, + { + "epoch": 52.0, + "learning_rate": 4.8856410256410256e-06, + "loss": 0.0001, + "step": 10475 + }, + { + "epoch": 52.01, + "learning_rate": 4.872820512820513e-06, + "loss": 0.0001, + "step": 10500 + }, + { + "epoch": 52.01, + "learning_rate": 4.86e-06, + "loss": 0.0001, + "step": 10525 + }, + { + "epoch": 52.01, + "learning_rate": 4.847179487179488e-06, + "loss": 0.0001, + "step": 10550 + }, + { + "epoch": 52.01, + "learning_rate": 4.8343589743589746e-06, + "loss": 0.0001, + "step": 10575 + }, + { + "epoch": 52.01, + "learning_rate": 4.821538461538462e-06, + "loss": 0.0002, + "step": 10600 + }, + { + "epoch": 53.0, + "learning_rate": 4.808717948717949e-06, + "loss": 0.0001, + "step": 10625 + }, + { + "epoch": 53.0, + "learning_rate": 4.795897435897437e-06, + "loss": 0.0001, + "step": 10650 + }, + { + "epoch": 53.0, + "learning_rate": 4.7830769230769235e-06, + "loss": 0.0002, + "step": 10675 + }, + { + "epoch": 53.01, + "learning_rate": 4.770256410256411e-06, + "loss": 0.0001, + "step": 10700 + }, + { + "epoch": 53.01, + "learning_rate": 4.757435897435898e-06, + "loss": 0.0009, + "step": 10725 + }, + { + "epoch": 53.01, + "learning_rate": 4.744615384615385e-06, + "loss": 0.0001, + "step": 10750 + }, + { + "epoch": 53.01, + "learning_rate": 4.731794871794872e-06, + "loss": 0.0001, + "step": 10775 + }, + { + "epoch": 53.01, + "learning_rate": 4.718974358974359e-06, + "loss": 0.0001, + "step": 10800 + }, + { + "epoch": 54.0, + "learning_rate": 4.706153846153846e-06, + "loss": 0.0001, + "step": 10825 + }, + { + "epoch": 54.0, + "learning_rate": 4.693333333333334e-06, + "loss": 0.0001, + "step": 10850 + }, + { + "epoch": 54.0, + "learning_rate": 4.680512820512821e-06, + "loss": 0.0001, + "step": 10875 + }, + { + "epoch": 54.01, + "learning_rate": 4.667692307692308e-06, + "loss": 0.0001, + "step": 10900 + }, + { + "epoch": 54.01, + "learning_rate": 4.654871794871795e-06, + "loss": 0.0001, + "step": 10925 + }, + { + "epoch": 54.01, + "learning_rate": 4.642051282051283e-06, + "loss": 0.0001, + "step": 10950 + }, + { + "epoch": 54.01, + "learning_rate": 4.62923076923077e-06, + "loss": 0.0001, + "step": 10975 + }, + { + "epoch": 54.01, + "learning_rate": 4.616410256410257e-06, + "loss": 0.0, + "step": 11000 + }, + { + "epoch": 54.01, + "eval_loss": 0.3027019202709198, + "eval_runtime": 2246.8011, + "eval_samples_per_second": 2.913, + "eval_steps_per_second": 0.182, + "eval_wer": 7.4365342163355415, + "step": 11000 + }, + { + "epoch": 55.0, + "learning_rate": 4.603589743589744e-06, + "loss": 0.0, + "step": 11025 + }, + { + "epoch": 55.0, + "learning_rate": 4.590769230769232e-06, + "loss": 0.0001, + "step": 11050 + }, + { + "epoch": 55.0, + "learning_rate": 4.577948717948718e-06, + "loss": 0.0, + "step": 11075 + }, + { + "epoch": 55.01, + "learning_rate": 4.5651282051282055e-06, + "loss": 0.0, + "step": 11100 + }, + { + "epoch": 55.01, + "learning_rate": 4.552307692307692e-06, + "loss": 0.0, + "step": 11125 + }, + { + "epoch": 55.01, + "learning_rate": 4.53948717948718e-06, + "loss": 0.0, + "step": 11150 + }, + { + "epoch": 55.01, + "learning_rate": 4.526666666666667e-06, + "loss": 0.0, + "step": 11175 + }, + { + "epoch": 55.01, + "learning_rate": 4.5138461538461545e-06, + "loss": 0.0, + "step": 11200 + }, + { + "epoch": 56.0, + "learning_rate": 4.501025641025641e-06, + "loss": 0.0, + "step": 11225 + }, + { + "epoch": 56.0, + "learning_rate": 4.488205128205129e-06, + "loss": 0.0, + "step": 11250 + }, + { + "epoch": 56.0, + "learning_rate": 4.475384615384616e-06, + "loss": 0.0, + "step": 11275 + }, + { + "epoch": 56.01, + "learning_rate": 4.4625641025641035e-06, + "loss": 0.0, + "step": 11300 + }, + { + "epoch": 56.01, + "learning_rate": 4.44974358974359e-06, + "loss": 0.0, + "step": 11325 + }, + { + "epoch": 56.01, + "learning_rate": 4.436923076923078e-06, + "loss": 0.0, + "step": 11350 + }, + { + "epoch": 56.01, + "learning_rate": 4.424102564102564e-06, + "loss": 0.0, + "step": 11375 + }, + { + "epoch": 56.01, + "learning_rate": 4.411282051282052e-06, + "loss": 0.0, + "step": 11400 + }, + { + "epoch": 57.0, + "learning_rate": 4.3984615384615384e-06, + "loss": 0.0, + "step": 11425 + }, + { + "epoch": 57.0, + "learning_rate": 4.385641025641026e-06, + "loss": 0.0, + "step": 11450 + }, + { + "epoch": 57.0, + "learning_rate": 4.372820512820513e-06, + "loss": 0.0, + "step": 11475 + }, + { + "epoch": 57.01, + "learning_rate": 4.360000000000001e-06, + "loss": 0.0, + "step": 11500 + }, + { + "epoch": 57.01, + "learning_rate": 4.3471794871794874e-06, + "loss": 0.0, + "step": 11525 + }, + { + "epoch": 57.01, + "learning_rate": 4.334358974358975e-06, + "loss": 0.0, + "step": 11550 + }, + { + "epoch": 57.01, + "learning_rate": 4.321538461538462e-06, + "loss": 0.0, + "step": 11575 + }, + { + "epoch": 57.01, + "learning_rate": 4.30871794871795e-06, + "loss": 0.0, + "step": 11600 + }, + { + "epoch": 58.0, + "learning_rate": 4.2958974358974364e-06, + "loss": 0.0, + "step": 11625 + }, + { + "epoch": 58.0, + "learning_rate": 4.283076923076924e-06, + "loss": 0.0, + "step": 11650 + }, + { + "epoch": 58.0, + "learning_rate": 4.27025641025641e-06, + "loss": 0.0, + "step": 11675 + }, + { + "epoch": 58.01, + "learning_rate": 4.257435897435898e-06, + "loss": 0.0, + "step": 11700 + }, + { + "epoch": 58.01, + "learning_rate": 4.2446153846153846e-06, + "loss": 0.0, + "step": 11725 + }, + { + "epoch": 58.01, + "learning_rate": 4.231794871794872e-06, + "loss": 0.0, + "step": 11750 + }, + { + "epoch": 58.01, + "learning_rate": 4.218974358974359e-06, + "loss": 0.0, + "step": 11775 + }, + { + "epoch": 58.01, + "learning_rate": 4.206153846153847e-06, + "loss": 0.0, + "step": 11800 + }, + { + "epoch": 59.0, + "learning_rate": 4.1933333333333336e-06, + "loss": 0.0, + "step": 11825 + }, + { + "epoch": 59.0, + "learning_rate": 4.180512820512821e-06, + "loss": 0.0, + "step": 11850 + }, + { + "epoch": 59.0, + "learning_rate": 4.167692307692308e-06, + "loss": 0.0, + "step": 11875 + }, + { + "epoch": 59.01, + "learning_rate": 4.154871794871796e-06, + "loss": 0.0, + "step": 11900 + }, + { + "epoch": 59.01, + "learning_rate": 4.1420512820512826e-06, + "loss": 0.0, + "step": 11925 + }, + { + "epoch": 59.01, + "learning_rate": 4.12923076923077e-06, + "loss": 0.0, + "step": 11950 + }, + { + "epoch": 59.01, + "learning_rate": 4.116410256410256e-06, + "loss": 0.0, + "step": 11975 + }, + { + "epoch": 59.01, + "learning_rate": 4.103589743589744e-06, + "loss": 0.0, + "step": 12000 + }, + { + "epoch": 59.01, + "eval_loss": 0.31594809889793396, + "eval_runtime": 2252.629, + "eval_samples_per_second": 2.906, + "eval_steps_per_second": 0.182, + "eval_wer": 7.405491169977925, + "step": 12000 + }, + { + "epoch": 60.0, + "learning_rate": 4.090769230769231e-06, + "loss": 0.0, + "step": 12025 + }, + { + "epoch": 60.0, + "learning_rate": 4.077948717948718e-06, + "loss": 0.0, + "step": 12050 + }, + { + "epoch": 60.0, + "learning_rate": 4.065128205128205e-06, + "loss": 0.0, + "step": 12075 + }, + { + "epoch": 60.01, + "learning_rate": 4.052307692307693e-06, + "loss": 0.0, + "step": 12100 + }, + { + "epoch": 60.01, + "learning_rate": 4.03948717948718e-06, + "loss": 0.0, + "step": 12125 + }, + { + "epoch": 60.01, + "learning_rate": 4.026666666666667e-06, + "loss": 0.0, + "step": 12150 + }, + { + "epoch": 60.01, + "learning_rate": 4.013846153846154e-06, + "loss": 0.0, + "step": 12175 + }, + { + "epoch": 60.01, + "learning_rate": 4.001025641025642e-06, + "loss": 0.0, + "step": 12200 + }, + { + "epoch": 61.0, + "learning_rate": 3.988205128205129e-06, + "loss": 0.0, + "step": 12225 + }, + { + "epoch": 61.0, + "learning_rate": 3.975384615384616e-06, + "loss": 0.0, + "step": 12250 + }, + { + "epoch": 61.0, + "learning_rate": 3.962564102564102e-06, + "loss": 0.0, + "step": 12275 + }, + { + "epoch": 61.01, + "learning_rate": 3.94974358974359e-06, + "loss": 0.0, + "step": 12300 + }, + { + "epoch": 61.01, + "learning_rate": 3.936923076923077e-06, + "loss": 0.0, + "step": 12325 + }, + { + "epoch": 61.01, + "learning_rate": 3.9241025641025645e-06, + "loss": 0.0, + "step": 12350 + }, + { + "epoch": 61.01, + "learning_rate": 3.911282051282051e-06, + "loss": 0.0, + "step": 12375 + }, + { + "epoch": 61.01, + "learning_rate": 3.898461538461539e-06, + "loss": 0.0, + "step": 12400 + }, + { + "epoch": 62.0, + "learning_rate": 3.885641025641026e-06, + "loss": 0.0, + "step": 12425 + }, + { + "epoch": 62.0, + "learning_rate": 3.8728205128205135e-06, + "loss": 0.0, + "step": 12450 + }, + { + "epoch": 62.0, + "learning_rate": 3.86e-06, + "loss": 0.0, + "step": 12475 + }, + { + "epoch": 62.01, + "learning_rate": 3.847179487179488e-06, + "loss": 0.0, + "step": 12500 + }, + { + "epoch": 62.01, + "learning_rate": 3.834358974358975e-06, + "loss": 0.0, + "step": 12525 + }, + { + "epoch": 62.01, + "learning_rate": 3.8215384615384625e-06, + "loss": 0.0, + "step": 12550 + }, + { + "epoch": 62.01, + "learning_rate": 3.808717948717949e-06, + "loss": 0.0, + "step": 12575 + }, + { + "epoch": 62.01, + "learning_rate": 3.795897435897436e-06, + "loss": 0.0, + "step": 12600 + }, + { + "epoch": 63.0, + "learning_rate": 3.7830769230769234e-06, + "loss": 0.0, + "step": 12625 + }, + { + "epoch": 63.0, + "learning_rate": 3.7702564102564106e-06, + "loss": 0.0, + "step": 12650 + }, + { + "epoch": 63.0, + "learning_rate": 3.7574358974358974e-06, + "loss": 0.0, + "step": 12675 + }, + { + "epoch": 63.01, + "learning_rate": 3.7446153846153847e-06, + "loss": 0.0, + "step": 12700 + }, + { + "epoch": 63.01, + "learning_rate": 3.731794871794872e-06, + "loss": 0.0, + "step": 12725 + }, + { + "epoch": 63.01, + "learning_rate": 3.718974358974359e-06, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 63.01, + "learning_rate": 3.7061538461538464e-06, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 63.01, + "learning_rate": 3.6933333333333337e-06, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 64.0, + "learning_rate": 3.680512820512821e-06, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 64.0, + "learning_rate": 3.667692307692308e-06, + "loss": 0.0, + "step": 12850 + }, + { + "epoch": 64.0, + "learning_rate": 3.6548717948717954e-06, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 64.0, + "learning_rate": 3.6420512820512823e-06, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 64.01, + "learning_rate": 3.6292307692307695e-06, + "loss": 0.0, + "step": 12925 + }, + { + "epoch": 64.01, + "learning_rate": 3.6164102564102568e-06, + "loss": 0.0, + "step": 12950 + }, + { + "epoch": 64.01, + "learning_rate": 3.6035897435897436e-06, + "loss": 0.0, + "step": 12975 + }, + { + "epoch": 64.01, + "learning_rate": 3.590769230769231e-06, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 64.01, + "eval_loss": 0.3241559565067291, + "eval_runtime": 2238.8608, + "eval_samples_per_second": 2.924, + "eval_steps_per_second": 0.183, + "eval_wer": 7.369274282560706, + "step": 13000 + }, + { + "epoch": 65.0, + "learning_rate": 3.577948717948718e-06, + "loss": 0.0, + "step": 13025 + }, + { + "epoch": 65.0, + "learning_rate": 3.5651282051282053e-06, + "loss": 0.0, + "step": 13050 + }, + { + "epoch": 65.0, + "learning_rate": 3.5523076923076926e-06, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 65.0, + "learning_rate": 3.53948717948718e-06, + "loss": 0.0, + "step": 13100 + }, + { + "epoch": 65.01, + "learning_rate": 3.526666666666667e-06, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 65.01, + "learning_rate": 3.5138461538461543e-06, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 65.01, + "learning_rate": 3.5010256410256416e-06, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 65.01, + "learning_rate": 3.4882051282051284e-06, + "loss": 0.0, + "step": 13200 + }, + { + "epoch": 66.0, + "learning_rate": 3.4753846153846156e-06, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 66.0, + "learning_rate": 3.462564102564103e-06, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 66.0, + "learning_rate": 3.4497435897435897e-06, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 66.0, + "learning_rate": 3.436923076923077e-06, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 66.01, + "learning_rate": 3.424102564102564e-06, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 66.01, + "learning_rate": 3.4112820512820514e-06, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 66.01, + "learning_rate": 3.3984615384615387e-06, + "loss": 0.0, + "step": 13375 + }, + { + "epoch": 66.01, + "learning_rate": 3.385641025641026e-06, + "loss": 0.0, + "step": 13400 + }, + { + "epoch": 67.0, + "learning_rate": 3.372820512820513e-06, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 67.0, + "learning_rate": 3.3600000000000004e-06, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 67.0, + "learning_rate": 3.3471794871794877e-06, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 67.0, + "learning_rate": 3.3343589743589745e-06, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 67.01, + "learning_rate": 3.3215384615384618e-06, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 67.01, + "learning_rate": 3.308717948717949e-06, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 67.01, + "learning_rate": 3.2958974358974363e-06, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 67.01, + "learning_rate": 3.283076923076923e-06, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 68.0, + "learning_rate": 3.2702564102564103e-06, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 68.0, + "learning_rate": 3.2574358974358976e-06, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 68.0, + "learning_rate": 3.244615384615385e-06, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 68.0, + "learning_rate": 3.231794871794872e-06, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 68.01, + "learning_rate": 3.2189743589743593e-06, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 68.01, + "learning_rate": 3.2061538461538466e-06, + "loss": 0.0, + "step": 13750 + }, + { + "epoch": 68.01, + "learning_rate": 3.193333333333334e-06, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 68.01, + "learning_rate": 3.1805128205128206e-06, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 69.0, + "learning_rate": 3.167692307692308e-06, + "loss": 0.0, + "step": 13825 + }, + { + "epoch": 69.0, + "learning_rate": 3.154871794871795e-06, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 69.0, + "learning_rate": 3.1420512820512824e-06, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 69.0, + "learning_rate": 3.129230769230769e-06, + "loss": 0.0, + "step": 13900 + }, + { + "epoch": 69.01, + "learning_rate": 3.1164102564102565e-06, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 69.01, + "learning_rate": 3.1035897435897437e-06, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 69.01, + "learning_rate": 3.090769230769231e-06, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 69.01, + "learning_rate": 3.077948717948718e-06, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 69.01, + "eval_loss": 0.3312433362007141, + "eval_runtime": 2291.6313, + "eval_samples_per_second": 2.856, + "eval_steps_per_second": 0.179, + "eval_wer": 7.3071881898454745, + "step": 14000 + }, + { + "epoch": 70.0, + "learning_rate": 3.0651282051282054e-06, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 70.0, + "learning_rate": 3.0523076923076927e-06, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 70.0, + "learning_rate": 3.03948717948718e-06, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 70.0, + "learning_rate": 3.0266666666666668e-06, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 70.01, + "learning_rate": 3.013846153846154e-06, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 70.01, + "learning_rate": 3.0010256410256413e-06, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 70.01, + "learning_rate": 2.9882051282051285e-06, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 70.01, + "learning_rate": 2.9753846153846153e-06, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 71.0, + "learning_rate": 2.9625641025641026e-06, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 71.0, + "learning_rate": 2.94974358974359e-06, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 71.0, + "learning_rate": 2.936923076923077e-06, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 71.0, + "learning_rate": 2.9241025641025643e-06, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 71.01, + "learning_rate": 2.9112820512820516e-06, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 71.01, + "learning_rate": 2.898461538461539e-06, + "loss": 0.0, + "step": 14350 + }, + { + "epoch": 71.01, + "learning_rate": 2.885641025641026e-06, + "loss": 0.0, + "step": 14375 + }, + { + "epoch": 71.01, + "learning_rate": 2.8728205128205133e-06, + "loss": 0.0, + "step": 14400 + }, + { + "epoch": 72.0, + "learning_rate": 2.86e-06, + "loss": 0.0, + "step": 14425 + }, + { + "epoch": 72.0, + "learning_rate": 2.8471794871794874e-06, + "loss": 0.0, + "step": 14450 + }, + { + "epoch": 72.0, + "learning_rate": 2.8343589743589746e-06, + "loss": 0.0, + "step": 14475 + }, + { + "epoch": 72.0, + "learning_rate": 2.8215384615384615e-06, + "loss": 0.0, + "step": 14500 + }, + { + "epoch": 72.01, + "learning_rate": 2.8087179487179487e-06, + "loss": 0.0, + "step": 14525 + }, + { + "epoch": 72.01, + "learning_rate": 2.795897435897436e-06, + "loss": 0.0, + "step": 14550 + }, + { + "epoch": 72.01, + "learning_rate": 2.783076923076923e-06, + "loss": 0.0, + "step": 14575 + }, + { + "epoch": 72.01, + "learning_rate": 2.7702564102564105e-06, + "loss": 0.0, + "step": 14600 + }, + { + "epoch": 73.0, + "learning_rate": 2.7574358974358977e-06, + "loss": 0.0, + "step": 14625 + }, + { + "epoch": 73.0, + "learning_rate": 2.744615384615385e-06, + "loss": 0.0, + "step": 14650 + }, + { + "epoch": 73.0, + "learning_rate": 2.731794871794872e-06, + "loss": 0.0, + "step": 14675 + }, + { + "epoch": 73.0, + "learning_rate": 2.7189743589743594e-06, + "loss": 0.0, + "step": 14700 + }, + { + "epoch": 73.01, + "learning_rate": 2.7061538461538463e-06, + "loss": 0.0, + "step": 14725 + }, + { + "epoch": 73.01, + "learning_rate": 2.6933333333333335e-06, + "loss": 0.0, + "step": 14750 + }, + { + "epoch": 73.01, + "learning_rate": 2.6805128205128208e-06, + "loss": 0.0, + "step": 14775 + }, + { + "epoch": 73.01, + "learning_rate": 2.6676923076923076e-06, + "loss": 0.0, + "step": 14800 + }, + { + "epoch": 74.0, + "learning_rate": 2.654871794871795e-06, + "loss": 0.0, + "step": 14825 + }, + { + "epoch": 74.0, + "learning_rate": 2.642051282051282e-06, + "loss": 0.0, + "step": 14850 + }, + { + "epoch": 74.0, + "learning_rate": 2.6292307692307693e-06, + "loss": 0.0, + "step": 14875 + }, + { + "epoch": 74.0, + "learning_rate": 2.6164102564102566e-06, + "loss": 0.0, + "step": 14900 + }, + { + "epoch": 74.01, + "learning_rate": 2.603589743589744e-06, + "loss": 0.0, + "step": 14925 + }, + { + "epoch": 74.01, + "learning_rate": 2.590769230769231e-06, + "loss": 0.0, + "step": 14950 + }, + { + "epoch": 74.01, + "learning_rate": 2.5779487179487183e-06, + "loss": 0.0, + "step": 14975 + }, + { + "epoch": 74.01, + "learning_rate": 2.5651282051282056e-06, + "loss": 0.0, + "step": 15000 + }, + { + "epoch": 74.01, + "eval_loss": 0.3378990590572357, + "eval_runtime": 2231.9589, + "eval_samples_per_second": 2.933, + "eval_steps_per_second": 0.184, + "eval_wer": 7.022626931567329, + "step": 15000 + }, + { + "epoch": 75.0, + "learning_rate": 2.5523076923076924e-06, + "loss": 0.0, + "step": 15025 + }, + { + "epoch": 75.0, + "learning_rate": 2.5394871794871797e-06, + "loss": 0.0, + "step": 15050 + }, + { + "epoch": 75.0, + "learning_rate": 2.526666666666667e-06, + "loss": 0.0, + "step": 15075 + }, + { + "epoch": 75.0, + "learning_rate": 2.513846153846154e-06, + "loss": 0.0, + "step": 15100 + }, + { + "epoch": 75.01, + "learning_rate": 2.501025641025641e-06, + "loss": 0.0, + "step": 15125 + }, + { + "epoch": 75.01, + "learning_rate": 2.4882051282051282e-06, + "loss": 0.0, + "step": 15150 + }, + { + "epoch": 75.01, + "learning_rate": 2.4753846153846155e-06, + "loss": 0.0, + "step": 15175 + }, + { + "epoch": 75.01, + "learning_rate": 2.4625641025641027e-06, + "loss": 0.0, + "step": 15200 + }, + { + "epoch": 76.0, + "learning_rate": 2.44974358974359e-06, + "loss": 0.0, + "step": 15225 + }, + { + "epoch": 76.0, + "learning_rate": 2.436923076923077e-06, + "loss": 0.0, + "step": 15250 + }, + { + "epoch": 76.0, + "learning_rate": 2.424102564102564e-06, + "loss": 0.0, + "step": 15275 + }, + { + "epoch": 76.0, + "learning_rate": 2.4112820512820513e-06, + "loss": 0.0, + "step": 15300 + }, + { + "epoch": 76.01, + "learning_rate": 2.3984615384615385e-06, + "loss": 0.0, + "step": 15325 + }, + { + "epoch": 76.01, + "learning_rate": 2.3856410256410258e-06, + "loss": 0.0, + "step": 15350 + }, + { + "epoch": 76.01, + "learning_rate": 2.372820512820513e-06, + "loss": 0.0, + "step": 15375 + }, + { + "epoch": 76.01, + "learning_rate": 2.3600000000000003e-06, + "loss": 0.0, + "step": 15400 + }, + { + "epoch": 77.0, + "learning_rate": 2.347179487179487e-06, + "loss": 0.0, + "step": 15425 + }, + { + "epoch": 77.0, + "learning_rate": 2.3343589743589743e-06, + "loss": 0.0, + "step": 15450 + }, + { + "epoch": 77.0, + "learning_rate": 2.3215384615384616e-06, + "loss": 0.0, + "step": 15475 + }, + { + "epoch": 77.0, + "learning_rate": 2.308717948717949e-06, + "loss": 0.0, + "step": 15500 + }, + { + "epoch": 77.01, + "learning_rate": 2.295897435897436e-06, + "loss": 0.0, + "step": 15525 + }, + { + "epoch": 77.01, + "learning_rate": 2.2830769230769233e-06, + "loss": 0.0, + "step": 15550 + }, + { + "epoch": 77.01, + "learning_rate": 2.27025641025641e-06, + "loss": 0.0, + "step": 15575 + }, + { + "epoch": 77.01, + "learning_rate": 2.2574358974358974e-06, + "loss": 0.0, + "step": 15600 + }, + { + "epoch": 78.0, + "learning_rate": 2.2446153846153847e-06, + "loss": 0.0, + "step": 15625 + }, + { + "epoch": 78.0, + "learning_rate": 2.231794871794872e-06, + "loss": 0.0, + "step": 15650 + }, + { + "epoch": 78.0, + "learning_rate": 2.218974358974359e-06, + "loss": 0.0, + "step": 15675 + }, + { + "epoch": 78.0, + "learning_rate": 2.2061538461538464e-06, + "loss": 0.0, + "step": 15700 + }, + { + "epoch": 78.01, + "learning_rate": 2.1933333333333332e-06, + "loss": 0.0, + "step": 15725 + }, + { + "epoch": 78.01, + "learning_rate": 2.1805128205128205e-06, + "loss": 0.0, + "step": 15750 + }, + { + "epoch": 78.01, + "learning_rate": 2.1676923076923077e-06, + "loss": 0.0, + "step": 15775 + }, + { + "epoch": 78.01, + "learning_rate": 2.154871794871795e-06, + "loss": 0.0, + "step": 15800 + }, + { + "epoch": 79.0, + "learning_rate": 2.1420512820512822e-06, + "loss": 0.0, + "step": 15825 + }, + { + "epoch": 79.0, + "learning_rate": 2.1292307692307695e-06, + "loss": 0.0, + "step": 15850 + }, + { + "epoch": 79.0, + "learning_rate": 2.1164102564102563e-06, + "loss": 0.0, + "step": 15875 + }, + { + "epoch": 79.0, + "learning_rate": 2.1035897435897435e-06, + "loss": 0.0, + "step": 15900 + }, + { + "epoch": 79.01, + "learning_rate": 2.0907692307692308e-06, + "loss": 0.0, + "step": 15925 + }, + { + "epoch": 79.01, + "learning_rate": 2.077948717948718e-06, + "loss": 0.0, + "step": 15950 + }, + { + "epoch": 79.01, + "learning_rate": 2.0651282051282053e-06, + "loss": 0.0, + "step": 15975 + }, + { + "epoch": 79.01, + "learning_rate": 2.0523076923076925e-06, + "loss": 0.0, + "step": 16000 + }, + { + "epoch": 79.01, + "eval_loss": 0.3442104458808899, + "eval_runtime": 2218.5267, + "eval_samples_per_second": 2.951, + "eval_steps_per_second": 0.185, + "eval_wer": 7.001931567328919, + "step": 16000 + }, + { + "epoch": 80.0, + "learning_rate": 2.0394871794871794e-06, + "loss": 0.0, + "step": 16025 + }, + { + "epoch": 80.0, + "learning_rate": 2.0266666666666666e-06, + "loss": 0.0, + "step": 16050 + }, + { + "epoch": 80.0, + "learning_rate": 2.013846153846154e-06, + "loss": 0.0, + "step": 16075 + }, + { + "epoch": 80.0, + "learning_rate": 2.001025641025641e-06, + "loss": 0.0, + "step": 16100 + }, + { + "epoch": 80.01, + "learning_rate": 1.9882051282051283e-06, + "loss": 0.0, + "step": 16125 + }, + { + "epoch": 80.01, + "learning_rate": 1.9753846153846156e-06, + "loss": 0.0, + "step": 16150 + }, + { + "epoch": 80.01, + "learning_rate": 1.9625641025641024e-06, + "loss": 0.0, + "step": 16175 + }, + { + "epoch": 80.01, + "learning_rate": 1.9497435897435897e-06, + "loss": 0.0, + "step": 16200 + }, + { + "epoch": 81.0, + "learning_rate": 1.936923076923077e-06, + "loss": 0.0, + "step": 16225 + }, + { + "epoch": 81.0, + "learning_rate": 1.924102564102564e-06, + "loss": 0.0, + "step": 16250 + }, + { + "epoch": 81.0, + "learning_rate": 1.9112820512820514e-06, + "loss": 0.0, + "step": 16275 + }, + { + "epoch": 81.0, + "learning_rate": 1.8984615384615387e-06, + "loss": 0.0, + "step": 16300 + }, + { + "epoch": 81.01, + "learning_rate": 1.8856410256410257e-06, + "loss": 0.0, + "step": 16325 + }, + { + "epoch": 81.01, + "learning_rate": 1.872820512820513e-06, + "loss": 0.0, + "step": 16350 + }, + { + "epoch": 81.01, + "learning_rate": 1.8600000000000002e-06, + "loss": 0.0, + "step": 16375 + }, + { + "epoch": 81.01, + "learning_rate": 1.8471794871794874e-06, + "loss": 0.0, + "step": 16400 + }, + { + "epoch": 82.0, + "learning_rate": 1.8343589743589747e-06, + "loss": 0.0, + "step": 16425 + }, + { + "epoch": 82.0, + "learning_rate": 1.8215384615384617e-06, + "loss": 0.0, + "step": 16450 + }, + { + "epoch": 82.0, + "learning_rate": 1.8087179487179488e-06, + "loss": 0.0, + "step": 16475 + }, + { + "epoch": 82.0, + "learning_rate": 1.795897435897436e-06, + "loss": 0.0, + "step": 16500 + }, + { + "epoch": 82.01, + "learning_rate": 1.7830769230769233e-06, + "loss": 0.0, + "step": 16525 + }, + { + "epoch": 82.01, + "learning_rate": 1.7702564102564105e-06, + "loss": 0.0, + "step": 16550 + }, + { + "epoch": 82.01, + "learning_rate": 1.7574358974358978e-06, + "loss": 0.0, + "step": 16575 + }, + { + "epoch": 82.01, + "learning_rate": 1.7446153846153848e-06, + "loss": 0.0, + "step": 16600 + }, + { + "epoch": 83.0, + "learning_rate": 1.7317948717948718e-06, + "loss": 0.0, + "step": 16625 + }, + { + "epoch": 83.0, + "learning_rate": 1.718974358974359e-06, + "loss": 0.0, + "step": 16650 + }, + { + "epoch": 83.0, + "learning_rate": 1.7061538461538463e-06, + "loss": 0.0, + "step": 16675 + }, + { + "epoch": 83.0, + "learning_rate": 1.6933333333333336e-06, + "loss": 0.0, + "step": 16700 + }, + { + "epoch": 83.01, + "learning_rate": 1.6805128205128208e-06, + "loss": 0.0, + "step": 16725 + }, + { + "epoch": 83.01, + "learning_rate": 1.6676923076923079e-06, + "loss": 0.0, + "step": 16750 + }, + { + "epoch": 83.01, + "learning_rate": 1.6548717948717949e-06, + "loss": 0.0, + "step": 16775 + }, + { + "epoch": 83.01, + "learning_rate": 1.6420512820512821e-06, + "loss": 0.0, + "step": 16800 + }, + { + "epoch": 84.0, + "learning_rate": 1.6292307692307694e-06, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 84.0, + "learning_rate": 1.6164102564102566e-06, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 84.0, + "learning_rate": 1.6035897435897439e-06, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 84.0, + "learning_rate": 1.590769230769231e-06, + "loss": 0.0, + "step": 16900 + }, + { + "epoch": 84.01, + "learning_rate": 1.577948717948718e-06, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 84.01, + "learning_rate": 1.5651282051282052e-06, + "loss": 0.0, + "step": 16950 + }, + { + "epoch": 84.01, + "learning_rate": 1.5523076923076924e-06, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 84.01, + "learning_rate": 1.5394871794871797e-06, + "loss": 0.0, + "step": 17000 + }, + { + "epoch": 84.01, + "eval_loss": 0.3499981462955475, + "eval_runtime": 2228.2995, + "eval_samples_per_second": 2.938, + "eval_steps_per_second": 0.184, + "eval_wer": 6.993308498896248, + "step": 17000 + }, + { + "epoch": 85.0, + "learning_rate": 1.526666666666667e-06, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 85.0, + "learning_rate": 1.513846153846154e-06, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 85.0, + "learning_rate": 1.501025641025641e-06, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 85.0, + "learning_rate": 1.4882051282051283e-06, + "loss": 0.0, + "step": 17100 + }, + { + "epoch": 85.01, + "learning_rate": 1.4753846153846155e-06, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 85.01, + "learning_rate": 1.4625641025641028e-06, + "loss": 0.0, + "step": 17150 + }, + { + "epoch": 85.01, + "learning_rate": 1.44974358974359e-06, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 85.01, + "learning_rate": 1.436923076923077e-06, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 86.0, + "learning_rate": 1.424102564102564e-06, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 86.0, + "learning_rate": 1.4112820512820513e-06, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 86.0, + "learning_rate": 1.3984615384615386e-06, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 86.0, + "learning_rate": 1.3856410256410258e-06, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 86.01, + "learning_rate": 1.372820512820513e-06, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 86.01, + "learning_rate": 1.3600000000000001e-06, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 86.01, + "learning_rate": 1.3471794871794871e-06, + "loss": 0.0, + "step": 17375 + }, + { + "epoch": 86.01, + "learning_rate": 1.3343589743589744e-06, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 87.0, + "learning_rate": 1.3215384615384616e-06, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 87.0, + "learning_rate": 1.3087179487179489e-06, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 87.0, + "learning_rate": 1.2958974358974361e-06, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 87.0, + "learning_rate": 1.2830769230769232e-06, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 87.01, + "learning_rate": 1.2702564102564102e-06, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 87.01, + "learning_rate": 1.2574358974358975e-06, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 87.01, + "learning_rate": 1.2446153846153847e-06, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 87.01, + "learning_rate": 1.231794871794872e-06, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 88.0, + "learning_rate": 1.2189743589743592e-06, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 88.0, + "learning_rate": 1.2061538461538462e-06, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 88.0, + "learning_rate": 1.1933333333333335e-06, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 88.0, + "learning_rate": 1.1805128205128207e-06, + "loss": 0.0, + "step": 17700 + }, + { + "epoch": 88.01, + "learning_rate": 1.1676923076923078e-06, + "loss": 0.0, + "step": 17725 + }, + { + "epoch": 88.01, + "learning_rate": 1.154871794871795e-06, + "loss": 0.0, + "step": 17750 + }, + { + "epoch": 88.01, + "learning_rate": 1.1420512820512823e-06, + "loss": 0.0, + "step": 17775 + }, + { + "epoch": 88.01, + "learning_rate": 1.1292307692307693e-06, + "loss": 0.0, + "step": 17800 + }, + { + "epoch": 89.0, + "learning_rate": 1.1164102564102565e-06, + "loss": 0.0, + "step": 17825 + }, + { + "epoch": 89.0, + "learning_rate": 1.1035897435897438e-06, + "loss": 0.0, + "step": 17850 + }, + { + "epoch": 89.0, + "learning_rate": 1.0907692307692308e-06, + "loss": 0.0, + "step": 17875 + }, + { + "epoch": 89.0, + "learning_rate": 1.077948717948718e-06, + "loss": 0.0, + "step": 17900 + }, + { + "epoch": 89.01, + "learning_rate": 1.0651282051282053e-06, + "loss": 0.0, + "step": 17925 + }, + { + "epoch": 89.01, + "learning_rate": 1.0523076923076924e-06, + "loss": 0.0, + "step": 17950 + }, + { + "epoch": 89.01, + "learning_rate": 1.0394871794871796e-06, + "loss": 0.0, + "step": 17975 + }, + { + "epoch": 89.01, + "learning_rate": 1.0266666666666669e-06, + "loss": 0.0, + "step": 18000 + }, + { + "epoch": 89.01, + "eval_loss": 0.35504841804504395, + "eval_runtime": 2203.3661, + "eval_samples_per_second": 2.971, + "eval_steps_per_second": 0.186, + "eval_wer": 6.960540838852097, + "step": 18000 + }, + { + "epoch": 90.0, + "learning_rate": 1.013846153846154e-06, + "loss": 0.0, + "step": 18025 + }, + { + "epoch": 90.0, + "learning_rate": 1.0010256410256411e-06, + "loss": 0.0, + "step": 18050 + }, + { + "epoch": 90.0, + "learning_rate": 9.882051282051284e-07, + "loss": 0.0, + "step": 18075 + }, + { + "epoch": 90.0, + "learning_rate": 9.753846153846154e-07, + "loss": 0.0, + "step": 18100 + }, + { + "epoch": 90.01, + "learning_rate": 9.625641025641027e-07, + "loss": 0.0, + "step": 18125 + }, + { + "epoch": 90.01, + "learning_rate": 9.497435897435898e-07, + "loss": 0.0, + "step": 18150 + }, + { + "epoch": 90.01, + "learning_rate": 9.36923076923077e-07, + "loss": 0.0, + "step": 18175 + }, + { + "epoch": 90.01, + "learning_rate": 9.241025641025642e-07, + "loss": 0.0, + "step": 18200 + }, + { + "epoch": 91.0, + "learning_rate": 9.112820512820514e-07, + "loss": 0.0, + "step": 18225 + }, + { + "epoch": 91.0, + "learning_rate": 8.984615384615385e-07, + "loss": 0.0, + "step": 18250 + }, + { + "epoch": 91.0, + "learning_rate": 8.856410256410257e-07, + "loss": 0.0, + "step": 18275 + }, + { + "epoch": 91.0, + "learning_rate": 8.728205128205129e-07, + "loss": 0.0, + "step": 18300 + }, + { + "epoch": 91.01, + "learning_rate": 8.6e-07, + "loss": 0.0, + "step": 18325 + }, + { + "epoch": 91.01, + "learning_rate": 8.471794871794873e-07, + "loss": 0.0, + "step": 18350 + }, + { + "epoch": 91.01, + "learning_rate": 8.343589743589744e-07, + "loss": 0.0, + "step": 18375 + }, + { + "epoch": 91.01, + "learning_rate": 8.215384615384617e-07, + "loss": 0.0, + "step": 18400 + }, + { + "epoch": 92.0, + "learning_rate": 8.087179487179488e-07, + "loss": 0.0, + "step": 18425 + }, + { + "epoch": 92.0, + "learning_rate": 7.958974358974359e-07, + "loss": 0.0, + "step": 18450 + }, + { + "epoch": 92.0, + "learning_rate": 7.830769230769232e-07, + "loss": 0.0, + "step": 18475 + }, + { + "epoch": 92.0, + "learning_rate": 7.702564102564103e-07, + "loss": 0.0, + "step": 18500 + }, + { + "epoch": 92.01, + "learning_rate": 7.574358974358975e-07, + "loss": 0.0, + "step": 18525 + }, + { + "epoch": 92.01, + "learning_rate": 7.446153846153847e-07, + "loss": 0.0, + "step": 18550 + }, + { + "epoch": 92.01, + "learning_rate": 7.317948717948719e-07, + "loss": 0.0, + "step": 18575 + }, + { + "epoch": 92.01, + "learning_rate": 7.18974358974359e-07, + "loss": 0.0, + "step": 18600 + }, + { + "epoch": 93.0, + "learning_rate": 7.061538461538463e-07, + "loss": 0.0, + "step": 18625 + }, + { + "epoch": 93.0, + "learning_rate": 6.933333333333334e-07, + "loss": 0.0, + "step": 18650 + }, + { + "epoch": 93.0, + "learning_rate": 6.805128205128205e-07, + "loss": 0.0, + "step": 18675 + }, + { + "epoch": 93.0, + "learning_rate": 6.676923076923078e-07, + "loss": 0.0, + "step": 18700 + }, + { + "epoch": 93.01, + "learning_rate": 6.548717948717949e-07, + "loss": 0.0, + "step": 18725 + }, + { + "epoch": 93.01, + "learning_rate": 6.420512820512821e-07, + "loss": 0.0, + "step": 18750 + }, + { + "epoch": 93.01, + "learning_rate": 6.292307692307693e-07, + "loss": 0.0, + "step": 18775 + }, + { + "epoch": 93.01, + "learning_rate": 6.164102564102565e-07, + "loss": 0.0, + "step": 18800 + }, + { + "epoch": 94.0, + "learning_rate": 6.035897435897436e-07, + "loss": 0.0, + "step": 18825 + }, + { + "epoch": 94.0, + "learning_rate": 5.907692307692309e-07, + "loss": 0.0, + "step": 18850 + }, + { + "epoch": 94.0, + "learning_rate": 5.77948717948718e-07, + "loss": 0.0, + "step": 18875 + }, + { + "epoch": 94.0, + "learning_rate": 5.651282051282051e-07, + "loss": 0.0, + "step": 18900 + }, + { + "epoch": 94.01, + "learning_rate": 5.523076923076924e-07, + "loss": 0.0, + "step": 18925 + }, + { + "epoch": 94.01, + "learning_rate": 5.394871794871795e-07, + "loss": 0.0, + "step": 18950 + }, + { + "epoch": 94.01, + "learning_rate": 5.266666666666667e-07, + "loss": 0.0, + "step": 18975 + }, + { + "epoch": 94.01, + "learning_rate": 5.138461538461539e-07, + "loss": 0.0, + "step": 19000 + }, + { + "epoch": 94.01, + "eval_loss": 0.35885128378868103, + "eval_runtime": 2263.3771, + "eval_samples_per_second": 2.892, + "eval_steps_per_second": 0.181, + "eval_wer": 6.946743929359823, + "step": 19000 + }, + { + "epoch": 95.0, + "learning_rate": 5.010256410256411e-07, + "loss": 0.0, + "step": 19025 + }, + { + "epoch": 95.0, + "learning_rate": 4.882051282051282e-07, + "loss": 0.0, + "step": 19050 + }, + { + "epoch": 95.0, + "learning_rate": 4.753846153846154e-07, + "loss": 0.0, + "step": 19075 + }, + { + "epoch": 95.0, + "learning_rate": 4.6256410256410265e-07, + "loss": 0.0, + "step": 19100 + }, + { + "epoch": 95.01, + "learning_rate": 4.497435897435898e-07, + "loss": 0.0, + "step": 19125 + }, + { + "epoch": 95.01, + "learning_rate": 4.3692307692307693e-07, + "loss": 0.0, + "step": 19150 + }, + { + "epoch": 95.01, + "learning_rate": 4.241025641025642e-07, + "loss": 0.0, + "step": 19175 + }, + { + "epoch": 95.01, + "learning_rate": 4.112820512820513e-07, + "loss": 0.0, + "step": 19200 + }, + { + "epoch": 96.0, + "learning_rate": 3.9846153846153846e-07, + "loss": 0.0, + "step": 19225 + }, + { + "epoch": 96.0, + "learning_rate": 3.856410256410257e-07, + "loss": 0.0, + "step": 19250 + }, + { + "epoch": 96.0, + "learning_rate": 3.7282051282051285e-07, + "loss": 0.0, + "step": 19275 + }, + { + "epoch": 96.0, + "learning_rate": 3.6e-07, + "loss": 0.0, + "step": 19300 + }, + { + "epoch": 96.01, + "learning_rate": 3.4717948717948724e-07, + "loss": 0.0, + "step": 19325 + }, + { + "epoch": 96.01, + "learning_rate": 3.343589743589744e-07, + "loss": 0.0, + "step": 19350 + }, + { + "epoch": 96.01, + "learning_rate": 3.215384615384615e-07, + "loss": 0.0, + "step": 19375 + }, + { + "epoch": 96.01, + "learning_rate": 3.087179487179488e-07, + "loss": 0.0, + "step": 19400 + }, + { + "epoch": 97.0, + "learning_rate": 2.958974358974359e-07, + "loss": 0.0, + "step": 19425 + }, + { + "epoch": 97.0, + "learning_rate": 2.830769230769231e-07, + "loss": 0.0, + "step": 19450 + }, + { + "epoch": 97.0, + "learning_rate": 2.702564102564103e-07, + "loss": 0.0, + "step": 19475 + }, + { + "epoch": 97.0, + "learning_rate": 2.5743589743589745e-07, + "loss": 0.0, + "step": 19500 + }, + { + "epoch": 97.01, + "learning_rate": 2.4461538461538464e-07, + "loss": 0.0, + "step": 19525 + }, + { + "epoch": 97.01, + "learning_rate": 2.3179487179487184e-07, + "loss": 0.0, + "step": 19550 + }, + { + "epoch": 97.01, + "learning_rate": 2.1897435897435898e-07, + "loss": 0.0, + "step": 19575 + }, + { + "epoch": 97.01, + "learning_rate": 2.0615384615384618e-07, + "loss": 0.0, + "step": 19600 + }, + { + "epoch": 98.0, + "learning_rate": 1.9333333333333337e-07, + "loss": 0.0, + "step": 19625 + }, + { + "epoch": 98.0, + "learning_rate": 1.805128205128205e-07, + "loss": 0.0, + "step": 19650 + }, + { + "epoch": 98.0, + "learning_rate": 1.676923076923077e-07, + "loss": 0.0, + "step": 19675 + }, + { + "epoch": 98.0, + "learning_rate": 1.548717948717949e-07, + "loss": 0.0, + "step": 19700 + }, + { + "epoch": 98.01, + "learning_rate": 1.4205128205128207e-07, + "loss": 0.0, + "step": 19725 + }, + { + "epoch": 98.01, + "learning_rate": 1.2923076923076924e-07, + "loss": 0.0, + "step": 19750 + }, + { + "epoch": 98.01, + "learning_rate": 1.1641025641025642e-07, + "loss": 0.0, + "step": 19775 + }, + { + "epoch": 98.01, + "learning_rate": 1.035897435897436e-07, + "loss": 0.0, + "step": 19800 + }, + { + "epoch": 99.0, + "learning_rate": 9.076923076923077e-08, + "loss": 0.0, + "step": 19825 + }, + { + "epoch": 99.0, + "learning_rate": 7.794871794871795e-08, + "loss": 0.0, + "step": 19850 + }, + { + "epoch": 99.0, + "learning_rate": 6.512820512820514e-08, + "loss": 0.0, + "step": 19875 + }, + { + "epoch": 99.0, + "learning_rate": 5.230769230769231e-08, + "loss": 0.0, + "step": 19900 + }, + { + "epoch": 99.01, + "learning_rate": 3.948717948717949e-08, + "loss": 0.0, + "step": 19925 + }, + { + "epoch": 99.01, + "learning_rate": 2.6666666666666667e-08, + "loss": 0.0, + "step": 19950 + }, + { + "epoch": 99.01, + "learning_rate": 1.3846153846153846e-08, + "loss": 0.0, + "step": 19975 + }, + { + "epoch": 99.01, + "learning_rate": 1.0256410256410256e-09, + "loss": 0.0, + "step": 20000 + }, + { + "epoch": 99.01, + "eval_loss": 0.3605120778083801, + "eval_runtime": 2205.8575, + "eval_samples_per_second": 2.968, + "eval_steps_per_second": 0.186, + "eval_wer": 6.939845474613686, + "step": 20000 + }, + { + "epoch": 99.01, + "step": 20000, + "total_flos": 2.716385004011523e+21, + "train_loss": 0.008237474963504065, + "train_runtime": 377050.5996, + "train_samples_per_second": 3.395, + "train_steps_per_second": 0.053 + } + ], + "logging_steps": 25, + "max_steps": 20000, + "num_train_epochs": 9223372036854775807, + "save_steps": 1000, + "total_flos": 2.716385004011523e+21, + "trial_name": null, + "trial_params": null +}