{ "best_metric": 12.981489732269432, "best_model_checkpoint": "./checkpoint-20000", "epoch": 97.0009, "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": 2.3743, "step": 25 }, { "epoch": 0.0, "learning_rate": 9.600000000000001e-07, "loss": 1.2975, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.46e-06, "loss": 0.8902, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.9600000000000003e-06, "loss": 0.6796, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.46e-06, "loss": 0.5704, "step": 125 }, { "epoch": 0.01, "learning_rate": 2.96e-06, "loss": 0.4795, "step": 150 }, { "epoch": 0.01, "learning_rate": 3.46e-06, "loss": 0.4317, "step": 175 }, { "epoch": 0.01, "learning_rate": 3.96e-06, "loss": 0.3784, "step": 200 }, { "epoch": 1.0, "learning_rate": 4.4600000000000005e-06, "loss": 0.3844, "step": 225 }, { "epoch": 1.0, "learning_rate": 4.960000000000001e-06, "loss": 0.3283, "step": 250 }, { "epoch": 1.0, "learning_rate": 5.460000000000001e-06, "loss": 0.3188, "step": 275 }, { "epoch": 1.0, "learning_rate": 5.9600000000000005e-06, "loss": 0.2675, "step": 300 }, { "epoch": 1.01, "learning_rate": 6.460000000000001e-06, "loss": 0.2506, "step": 325 }, { "epoch": 1.01, "learning_rate": 6.96e-06, "loss": 0.2087, "step": 350 }, { "epoch": 1.01, "learning_rate": 7.4600000000000006e-06, "loss": 0.1919, "step": 375 }, { "epoch": 1.01, "learning_rate": 7.960000000000002e-06, "loss": 0.1639, "step": 400 }, { "epoch": 2.0, "learning_rate": 8.46e-06, "loss": 0.1738, "step": 425 }, { "epoch": 2.0, "learning_rate": 8.96e-06, "loss": 0.1518, "step": 450 }, { "epoch": 2.0, "learning_rate": 9.460000000000001e-06, "loss": 0.1398, "step": 475 }, { "epoch": 2.0, "learning_rate": 9.960000000000001e-06, "loss": 0.1255, "step": 500 }, { "epoch": 2.01, "learning_rate": 9.988205128205129e-06, "loss": 0.1064, "step": 525 }, { "epoch": 2.01, "learning_rate": 9.975384615384616e-06, "loss": 0.0861, "step": 550 }, { "epoch": 2.01, "learning_rate": 9.962564102564102e-06, "loss": 0.0778, "step": 575 }, { "epoch": 2.01, "learning_rate": 9.94974358974359e-06, "loss": 0.0646, "step": 600 }, { "epoch": 3.0, "learning_rate": 9.936923076923078e-06, "loss": 0.0699, "step": 625 }, { "epoch": 3.0, "learning_rate": 9.924102564102565e-06, "loss": 0.061, "step": 650 }, { "epoch": 3.0, "learning_rate": 9.911282051282051e-06, "loss": 0.0543, "step": 675 }, { "epoch": 3.0, "learning_rate": 9.898461538461538e-06, "loss": 0.0506, "step": 700 }, { "epoch": 3.01, "learning_rate": 9.885641025641027e-06, "loss": 0.0456, "step": 725 }, { "epoch": 3.01, "learning_rate": 9.872820512820514e-06, "loss": 0.0351, "step": 750 }, { "epoch": 3.01, "learning_rate": 9.86e-06, "loss": 0.0347, "step": 775 }, { "epoch": 3.01, "learning_rate": 9.847179487179487e-06, "loss": 0.0295, "step": 800 }, { "epoch": 4.0, "learning_rate": 9.834358974358976e-06, "loss": 0.0303, "step": 825 }, { "epoch": 4.0, "learning_rate": 9.821538461538463e-06, "loss": 0.0295, "step": 850 }, { "epoch": 4.0, "learning_rate": 9.80871794871795e-06, "loss": 0.0263, "step": 875 }, { "epoch": 4.0, "learning_rate": 9.795897435897436e-06, "loss": 0.0284, "step": 900 }, { "epoch": 4.01, "learning_rate": 9.783076923076925e-06, "loss": 0.0255, "step": 925 }, { "epoch": 4.01, "learning_rate": 9.770256410256412e-06, "loss": 0.0203, "step": 950 }, { "epoch": 4.01, "learning_rate": 9.757435897435898e-06, "loss": 0.0198, "step": 975 }, { "epoch": 4.01, "learning_rate": 9.744615384615385e-06, "loss": 0.0162, "step": 1000 }, { "epoch": 4.01, "eval_loss": 0.31586089730262756, "eval_runtime": 2362.2793, "eval_samples_per_second": 2.79, "eval_steps_per_second": 0.174, "eval_wer": 16.857709911296528, "step": 1000 }, { "epoch": 4.01, "learning_rate": 9.731794871794872e-06, "loss": 0.0172, "step": 1025 }, { "epoch": 5.0, "learning_rate": 9.71897435897436e-06, "loss": 0.017, "step": 1050 }, { "epoch": 5.0, "learning_rate": 9.706153846153847e-06, "loss": 0.0144, "step": 1075 }, { "epoch": 5.0, "learning_rate": 9.693333333333334e-06, "loss": 0.0172, "step": 1100 }, { "epoch": 5.0, "learning_rate": 9.680512820512821e-06, "loss": 0.0154, "step": 1125 }, { "epoch": 5.01, "learning_rate": 9.667692307692308e-06, "loss": 0.014, "step": 1150 }, { "epoch": 5.01, "learning_rate": 9.654871794871795e-06, "loss": 0.0134, "step": 1175 }, { "epoch": 5.01, "learning_rate": 9.642051282051282e-06, "loss": 0.0132, "step": 1200 }, { "epoch": 5.01, "learning_rate": 9.62923076923077e-06, "loss": 0.0123, "step": 1225 }, { "epoch": 6.0, "learning_rate": 9.616410256410257e-06, "loss": 0.0133, "step": 1250 }, { "epoch": 6.0, "learning_rate": 9.603589743589744e-06, "loss": 0.0112, "step": 1275 }, { "epoch": 6.0, "learning_rate": 9.59076923076923e-06, "loss": 0.009, "step": 1300 }, { "epoch": 6.0, "learning_rate": 9.577948717948719e-06, "loss": 0.0108, "step": 1325 }, { "epoch": 6.01, "learning_rate": 9.565128205128206e-06, "loss": 0.0222, "step": 1350 }, { "epoch": 6.01, "learning_rate": 9.552307692307693e-06, "loss": 0.0095, "step": 1375 }, { "epoch": 6.01, "learning_rate": 9.53948717948718e-06, "loss": 0.0098, "step": 1400 }, { "epoch": 6.01, "learning_rate": 9.526666666666668e-06, "loss": 0.0084, "step": 1425 }, { "epoch": 7.0, "learning_rate": 9.513846153846155e-06, "loss": 0.0086, "step": 1450 }, { "epoch": 7.0, "learning_rate": 9.501025641025642e-06, "loss": 0.0069, "step": 1475 }, { "epoch": 7.0, "learning_rate": 9.488205128205129e-06, "loss": 0.0086, "step": 1500 }, { "epoch": 7.0, "learning_rate": 9.475384615384617e-06, "loss": 0.0104, "step": 1525 }, { "epoch": 7.01, "learning_rate": 9.462564102564104e-06, "loss": 0.0083, "step": 1550 }, { "epoch": 7.01, "learning_rate": 9.44974358974359e-06, "loss": 0.0064, "step": 1575 }, { "epoch": 7.01, "learning_rate": 9.436923076923078e-06, "loss": 0.0061, "step": 1600 }, { "epoch": 7.01, "learning_rate": 9.424102564102564e-06, "loss": 0.0063, "step": 1625 }, { "epoch": 8.0, "learning_rate": 9.411282051282053e-06, "loss": 0.0072, "step": 1650 }, { "epoch": 8.0, "learning_rate": 9.39846153846154e-06, "loss": 0.0063, "step": 1675 }, { "epoch": 8.0, "learning_rate": 9.385641025641027e-06, "loss": 0.0061, "step": 1700 }, { "epoch": 8.0, "learning_rate": 9.372820512820513e-06, "loss": 0.0064, "step": 1725 }, { "epoch": 8.01, "learning_rate": 9.360000000000002e-06, "loss": 0.0054, "step": 1750 }, { "epoch": 8.01, "learning_rate": 9.347179487179487e-06, "loss": 0.0045, "step": 1775 }, { "epoch": 8.01, "learning_rate": 9.334358974358974e-06, "loss": 0.0051, "step": 1800 }, { "epoch": 8.01, "learning_rate": 9.321538461538462e-06, "loss": 0.0068, "step": 1825 }, { "epoch": 8.01, "learning_rate": 9.30871794871795e-06, "loss": 0.0046, "step": 1850 }, { "epoch": 9.0, "learning_rate": 9.295897435897436e-06, "loss": 0.0046, "step": 1875 }, { "epoch": 9.0, "learning_rate": 9.283076923076923e-06, "loss": 0.0051, "step": 1900 }, { "epoch": 9.0, "learning_rate": 9.270256410256411e-06, "loss": 0.0054, "step": 1925 }, { "epoch": 9.0, "learning_rate": 9.257435897435898e-06, "loss": 0.0062, "step": 1950 }, { "epoch": 9.01, "learning_rate": 9.244615384615385e-06, "loss": 0.0037, "step": 1975 }, { "epoch": 9.01, "learning_rate": 9.231794871794872e-06, "loss": 0.0042, "step": 2000 }, { "epoch": 9.01, "eval_loss": 0.3181460499763489, "eval_runtime": 2362.6731, "eval_samples_per_second": 2.79, "eval_steps_per_second": 0.174, "eval_wer": 15.160597837093443, "step": 2000 }, { "epoch": 9.01, "learning_rate": 9.21897435897436e-06, "loss": 0.0049, "step": 2025 }, { "epoch": 9.01, "learning_rate": 9.206153846153847e-06, "loss": 0.0039, "step": 2050 }, { "epoch": 10.0, "learning_rate": 9.193333333333334e-06, "loss": 0.0052, "step": 2075 }, { "epoch": 10.0, "learning_rate": 9.18051282051282e-06, "loss": 0.0033, "step": 2100 }, { "epoch": 10.0, "learning_rate": 9.16769230769231e-06, "loss": 0.0031, "step": 2125 }, { "epoch": 10.0, "learning_rate": 9.154871794871796e-06, "loss": 0.0039, "step": 2150 }, { "epoch": 10.01, "learning_rate": 9.142051282051283e-06, "loss": 0.0024, "step": 2175 }, { "epoch": 10.01, "learning_rate": 9.12923076923077e-06, "loss": 0.0031, "step": 2200 }, { "epoch": 10.01, "learning_rate": 9.116410256410257e-06, "loss": 0.0039, "step": 2225 }, { "epoch": 10.01, "learning_rate": 9.103589743589745e-06, "loss": 0.0038, "step": 2250 }, { "epoch": 11.0, "learning_rate": 9.090769230769232e-06, "loss": 0.0032, "step": 2275 }, { "epoch": 11.0, "learning_rate": 9.077948717948719e-06, "loss": 0.004, "step": 2300 }, { "epoch": 11.0, "learning_rate": 9.065128205128206e-06, "loss": 0.0027, "step": 2325 }, { "epoch": 11.0, "learning_rate": 9.052307692307694e-06, "loss": 0.0036, "step": 2350 }, { "epoch": 11.01, "learning_rate": 9.03948717948718e-06, "loss": 0.0037, "step": 2375 }, { "epoch": 11.01, "learning_rate": 9.026666666666666e-06, "loss": 0.0039, "step": 2400 }, { "epoch": 11.01, "learning_rate": 9.013846153846155e-06, "loss": 0.0032, "step": 2425 }, { "epoch": 11.01, "learning_rate": 9.001025641025641e-06, "loss": 0.0027, "step": 2450 }, { "epoch": 12.0, "learning_rate": 8.988205128205128e-06, "loss": 0.0029, "step": 2475 }, { "epoch": 12.0, "learning_rate": 8.975384615384615e-06, "loss": 0.0032, "step": 2500 }, { "epoch": 12.0, "learning_rate": 8.962564102564104e-06, "loss": 0.0036, "step": 2525 }, { "epoch": 12.0, "learning_rate": 8.94974358974359e-06, "loss": 0.0037, "step": 2550 }, { "epoch": 12.01, "learning_rate": 8.936923076923077e-06, "loss": 0.0031, "step": 2575 }, { "epoch": 12.01, "learning_rate": 8.924102564102564e-06, "loss": 0.0028, "step": 2600 }, { "epoch": 12.01, "learning_rate": 8.911282051282053e-06, "loss": 0.0036, "step": 2625 }, { "epoch": 12.01, "learning_rate": 8.89846153846154e-06, "loss": 0.0046, "step": 2650 }, { "epoch": 12.01, "learning_rate": 8.885641025641026e-06, "loss": 0.0033, "step": 2675 }, { "epoch": 13.0, "learning_rate": 8.872820512820513e-06, "loss": 0.0036, "step": 2700 }, { "epoch": 13.0, "learning_rate": 8.860000000000002e-06, "loss": 0.0033, "step": 2725 }, { "epoch": 13.0, "learning_rate": 8.847179487179488e-06, "loss": 0.0026, "step": 2750 }, { "epoch": 13.0, "learning_rate": 8.834358974358975e-06, "loss": 0.0043, "step": 2775 }, { "epoch": 13.01, "learning_rate": 8.821538461538462e-06, "loss": 0.0042, "step": 2800 }, { "epoch": 13.01, "learning_rate": 8.80871794871795e-06, "loss": 0.0047, "step": 2825 }, { "epoch": 13.01, "learning_rate": 8.795897435897437e-06, "loss": 0.0035, "step": 2850 }, { "epoch": 13.01, "learning_rate": 8.783076923076924e-06, "loss": 0.0029, "step": 2875 }, { "epoch": 14.0, "learning_rate": 8.770256410256411e-06, "loss": 0.0027, "step": 2900 }, { "epoch": 14.0, "learning_rate": 8.757435897435898e-06, "loss": 0.0025, "step": 2925 }, { "epoch": 14.0, "learning_rate": 8.744615384615386e-06, "loss": 0.0054, "step": 2950 }, { "epoch": 14.0, "learning_rate": 8.731794871794873e-06, "loss": 0.005, "step": 2975 }, { "epoch": 14.01, "learning_rate": 8.718974358974358e-06, "loss": 0.0038, "step": 3000 }, { "epoch": 14.01, "eval_loss": 0.3366771936416626, "eval_runtime": 2360.8374, "eval_samples_per_second": 2.792, "eval_steps_per_second": 0.175, "eval_wer": 14.721130867997894, "step": 3000 }, { "epoch": 14.01, "learning_rate": 8.706153846153847e-06, "loss": 0.0036, "step": 3025 }, { "epoch": 14.01, "learning_rate": 8.693333333333334e-06, "loss": 0.0036, "step": 3050 }, { "epoch": 14.01, "learning_rate": 8.68051282051282e-06, "loss": 0.0043, "step": 3075 }, { "epoch": 15.0, "learning_rate": 8.667692307692307e-06, "loss": 0.0031, "step": 3100 }, { "epoch": 15.0, "learning_rate": 8.654871794871796e-06, "loss": 0.0034, "step": 3125 }, { "epoch": 15.0, "learning_rate": 8.642051282051283e-06, "loss": 0.0027, "step": 3150 }, { "epoch": 15.0, "learning_rate": 8.62923076923077e-06, "loss": 0.0037, "step": 3175 }, { "epoch": 15.01, "learning_rate": 8.616410256410256e-06, "loss": 0.0043, "step": 3200 }, { "epoch": 15.01, "learning_rate": 8.603589743589745e-06, "loss": 0.0023, "step": 3225 }, { "epoch": 15.01, "learning_rate": 8.590769230769232e-06, "loss": 0.0024, "step": 3250 }, { "epoch": 15.01, "learning_rate": 8.577948717948718e-06, "loss": 0.0033, "step": 3275 }, { "epoch": 16.0, "learning_rate": 8.565128205128205e-06, "loss": 0.0024, "step": 3300 }, { "epoch": 16.0, "learning_rate": 8.552307692307694e-06, "loss": 0.0039, "step": 3325 }, { "epoch": 16.0, "learning_rate": 8.53948717948718e-06, "loss": 0.0029, "step": 3350 }, { "epoch": 16.0, "learning_rate": 8.526666666666667e-06, "loss": 0.0027, "step": 3375 }, { "epoch": 16.01, "learning_rate": 8.513846153846154e-06, "loss": 0.0024, "step": 3400 }, { "epoch": 16.01, "learning_rate": 8.501025641025643e-06, "loss": 0.0022, "step": 3425 }, { "epoch": 16.01, "learning_rate": 8.48820512820513e-06, "loss": 0.0019, "step": 3450 }, { "epoch": 16.01, "learning_rate": 8.475384615384616e-06, "loss": 0.0025, "step": 3475 }, { "epoch": 16.01, "learning_rate": 8.462564102564103e-06, "loss": 0.0019, "step": 3500 }, { "epoch": 17.0, "learning_rate": 8.44974358974359e-06, "loss": 0.0033, "step": 3525 }, { "epoch": 17.0, "learning_rate": 8.436923076923079e-06, "loss": 0.004, "step": 3550 }, { "epoch": 17.0, "learning_rate": 8.424102564102565e-06, "loss": 0.0033, "step": 3575 }, { "epoch": 17.0, "learning_rate": 8.411282051282052e-06, "loss": 0.0031, "step": 3600 }, { "epoch": 17.01, "learning_rate": 8.398461538461539e-06, "loss": 0.004, "step": 3625 }, { "epoch": 17.01, "learning_rate": 8.385641025641026e-06, "loss": 0.0044, "step": 3650 }, { "epoch": 17.01, "learning_rate": 8.372820512820513e-06, "loss": 0.0044, "step": 3675 }, { "epoch": 17.01, "learning_rate": 8.36e-06, "loss": 0.0027, "step": 3700 }, { "epoch": 18.0, "learning_rate": 8.347179487179488e-06, "loss": 0.0032, "step": 3725 }, { "epoch": 18.0, "learning_rate": 8.334358974358975e-06, "loss": 0.0037, "step": 3750 }, { "epoch": 18.0, "learning_rate": 8.321538461538462e-06, "loss": 0.0041, "step": 3775 }, { "epoch": 18.0, "learning_rate": 8.308717948717949e-06, "loss": 0.0034, "step": 3800 }, { "epoch": 18.01, "learning_rate": 8.295897435897437e-06, "loss": 0.0036, "step": 3825 }, { "epoch": 18.01, "learning_rate": 8.283076923076924e-06, "loss": 0.0023, "step": 3850 }, { "epoch": 18.01, "learning_rate": 8.27025641025641e-06, "loss": 0.0035, "step": 3875 }, { "epoch": 18.01, "learning_rate": 8.257435897435898e-06, "loss": 0.0018, "step": 3900 }, { "epoch": 19.0, "learning_rate": 8.244615384615386e-06, "loss": 0.0026, "step": 3925 }, { "epoch": 19.0, "learning_rate": 8.231794871794873e-06, "loss": 0.0021, "step": 3950 }, { "epoch": 19.0, "learning_rate": 8.21897435897436e-06, "loss": 0.0032, "step": 3975 }, { "epoch": 19.0, "learning_rate": 8.206153846153847e-06, "loss": 0.0035, "step": 4000 }, { "epoch": 19.0, "eval_loss": 0.341941773891449, "eval_runtime": 2359.8566, "eval_samples_per_second": 2.793, "eval_steps_per_second": 0.175, "eval_wer": 14.591518490015796, "step": 4000 }, { "epoch": 19.01, "learning_rate": 8.193333333333335e-06, "loss": 0.0031, "step": 4025 }, { "epoch": 19.01, "learning_rate": 8.180512820512822e-06, "loss": 0.0026, "step": 4050 }, { "epoch": 19.01, "learning_rate": 8.167692307692309e-06, "loss": 0.0036, "step": 4075 }, { "epoch": 19.01, "learning_rate": 8.154871794871796e-06, "loss": 0.0037, "step": 4100 }, { "epoch": 20.0, "learning_rate": 8.142051282051282e-06, "loss": 0.0042, "step": 4125 }, { "epoch": 20.0, "learning_rate": 8.129230769230771e-06, "loss": 0.0041, "step": 4150 }, { "epoch": 20.0, "learning_rate": 8.116410256410258e-06, "loss": 0.0024, "step": 4175 }, { "epoch": 20.0, "learning_rate": 8.103589743589745e-06, "loss": 0.0029, "step": 4200 }, { "epoch": 20.01, "learning_rate": 8.090769230769231e-06, "loss": 0.0021, "step": 4225 }, { "epoch": 20.01, "learning_rate": 8.077948717948718e-06, "loss": 0.0024, "step": 4250 }, { "epoch": 20.01, "learning_rate": 8.065128205128205e-06, "loss": 0.0025, "step": 4275 }, { "epoch": 20.01, "learning_rate": 8.052307692307692e-06, "loss": 0.0021, "step": 4300 }, { "epoch": 20.01, "learning_rate": 8.03948717948718e-06, "loss": 0.0029, "step": 4325 }, { "epoch": 21.0, "learning_rate": 8.026666666666667e-06, "loss": 0.0023, "step": 4350 }, { "epoch": 21.0, "learning_rate": 8.013846153846154e-06, "loss": 0.002, "step": 4375 }, { "epoch": 21.0, "learning_rate": 8.00102564102564e-06, "loss": 0.0026, "step": 4400 }, { "epoch": 21.0, "learning_rate": 7.98820512820513e-06, "loss": 0.002, "step": 4425 }, { "epoch": 21.01, "learning_rate": 7.975384615384616e-06, "loss": 0.0021, "step": 4450 }, { "epoch": 21.01, "learning_rate": 7.962564102564103e-06, "loss": 0.0023, "step": 4475 }, { "epoch": 21.01, "learning_rate": 7.94974358974359e-06, "loss": 0.0025, "step": 4500 }, { "epoch": 21.01, "learning_rate": 7.936923076923078e-06, "loss": 0.0026, "step": 4525 }, { "epoch": 22.0, "learning_rate": 7.924615384615386e-06, "loss": 0.0021, "step": 4550 }, { "epoch": 22.0, "learning_rate": 7.911794871794873e-06, "loss": 0.0017, "step": 4575 }, { "epoch": 22.0, "learning_rate": 7.89897435897436e-06, "loss": 0.0018, "step": 4600 }, { "epoch": 22.0, "learning_rate": 7.886153846153846e-06, "loss": 0.0022, "step": 4625 }, { "epoch": 22.01, "learning_rate": 7.873333333333335e-06, "loss": 0.0019, "step": 4650 }, { "epoch": 22.01, "learning_rate": 7.860512820512822e-06, "loss": 0.0016, "step": 4675 }, { "epoch": 22.01, "learning_rate": 7.847692307692308e-06, "loss": 0.0028, "step": 4700 }, { "epoch": 22.01, "learning_rate": 7.834871794871795e-06, "loss": 0.002, "step": 4725 }, { "epoch": 23.0, "learning_rate": 7.822051282051282e-06, "loss": 0.002, "step": 4750 }, { "epoch": 23.0, "learning_rate": 7.80923076923077e-06, "loss": 0.0027, "step": 4775 }, { "epoch": 23.0, "learning_rate": 7.796410256410257e-06, "loss": 0.0012, "step": 4800 }, { "epoch": 23.0, "learning_rate": 7.783589743589744e-06, "loss": 0.001, "step": 4825 }, { "epoch": 23.01, "learning_rate": 7.770769230769231e-06, "loss": 0.0018, "step": 4850 }, { "epoch": 23.01, "learning_rate": 7.75794871794872e-06, "loss": 0.0012, "step": 4875 }, { "epoch": 23.01, "learning_rate": 7.745128205128206e-06, "loss": 0.0012, "step": 4900 }, { "epoch": 23.01, "learning_rate": 7.732307692307693e-06, "loss": 0.0009, "step": 4925 }, { "epoch": 24.0, "learning_rate": 7.71948717948718e-06, "loss": 0.0018, "step": 4950 }, { "epoch": 24.0, "learning_rate": 7.706666666666669e-06, "loss": 0.0012, "step": 4975 }, { "epoch": 24.0, "learning_rate": 7.693846153846154e-06, "loss": 0.0012, "step": 5000 }, { "epoch": 24.0, "eval_loss": 0.34889736771583557, "eval_runtime": 2367.2506, "eval_samples_per_second": 2.784, "eval_steps_per_second": 0.174, "eval_wer": 14.358621248329214, "step": 5000 }, { "epoch": 24.0, "learning_rate": 7.68102564102564e-06, "loss": 0.001, "step": 5025 }, { "epoch": 24.01, "learning_rate": 7.668205128205129e-06, "loss": 0.0007, "step": 5050 }, { "epoch": 24.01, "learning_rate": 7.655384615384616e-06, "loss": 0.0006, "step": 5075 }, { "epoch": 24.01, "learning_rate": 7.642564102564103e-06, "loss": 0.0008, "step": 5100 }, { "epoch": 24.01, "learning_rate": 7.62974358974359e-06, "loss": 0.0006, "step": 5125 }, { "epoch": 24.01, "learning_rate": 7.616923076923077e-06, "loss": 0.0016, "step": 5150 }, { "epoch": 25.0, "learning_rate": 7.604102564102565e-06, "loss": 0.001, "step": 5175 }, { "epoch": 25.0, "learning_rate": 7.591282051282052e-06, "loss": 0.001, "step": 5200 }, { "epoch": 25.0, "learning_rate": 7.578461538461539e-06, "loss": 0.0013, "step": 5225 }, { "epoch": 25.0, "learning_rate": 7.565641025641026e-06, "loss": 0.0009, "step": 5250 }, { "epoch": 25.01, "learning_rate": 7.552820512820514e-06, "loss": 0.0006, "step": 5275 }, { "epoch": 25.01, "learning_rate": 7.540000000000001e-06, "loss": 0.0009, "step": 5300 }, { "epoch": 25.01, "learning_rate": 7.5271794871794875e-06, "loss": 0.0009, "step": 5325 }, { "epoch": 25.01, "learning_rate": 7.514358974358975e-06, "loss": 0.001, "step": 5350 }, { "epoch": 26.0, "learning_rate": 7.501538461538462e-06, "loss": 0.0012, "step": 5375 }, { "epoch": 26.0, "learning_rate": 7.48871794871795e-06, "loss": 0.0016, "step": 5400 }, { "epoch": 26.0, "learning_rate": 7.4758974358974365e-06, "loss": 0.0018, "step": 5425 }, { "epoch": 26.0, "learning_rate": 7.463076923076924e-06, "loss": 0.0015, "step": 5450 }, { "epoch": 26.01, "learning_rate": 7.450256410256411e-06, "loss": 0.0016, "step": 5475 }, { "epoch": 26.01, "learning_rate": 7.437435897435899e-06, "loss": 0.0015, "step": 5500 }, { "epoch": 26.01, "learning_rate": 7.4246153846153855e-06, "loss": 0.002, "step": 5525 }, { "epoch": 26.01, "learning_rate": 7.411794871794873e-06, "loss": 0.0015, "step": 5550 }, { "epoch": 27.0, "learning_rate": 7.39897435897436e-06, "loss": 0.002, "step": 5575 }, { "epoch": 27.0, "learning_rate": 7.386153846153846e-06, "loss": 0.0019, "step": 5600 }, { "epoch": 27.0, "learning_rate": 7.373333333333334e-06, "loss": 0.0014, "step": 5625 }, { "epoch": 27.0, "learning_rate": 7.3605128205128204e-06, "loss": 0.0023, "step": 5650 }, { "epoch": 27.01, "learning_rate": 7.347692307692308e-06, "loss": 0.0025, "step": 5675 }, { "epoch": 27.01, "learning_rate": 7.334871794871795e-06, "loss": 0.0028, "step": 5700 }, { "epoch": 27.01, "learning_rate": 7.322051282051283e-06, "loss": 0.0019, "step": 5725 }, { "epoch": 27.01, "learning_rate": 7.309230769230769e-06, "loss": 0.0024, "step": 5750 }, { "epoch": 28.0, "learning_rate": 7.296410256410257e-06, "loss": 0.0035, "step": 5775 }, { "epoch": 28.0, "learning_rate": 7.283589743589744e-06, "loss": 0.0026, "step": 5800 }, { "epoch": 28.0, "learning_rate": 7.270769230769232e-06, "loss": 0.0033, "step": 5825 }, { "epoch": 28.0, "learning_rate": 7.257948717948718e-06, "loss": 0.0057, "step": 5850 }, { "epoch": 28.01, "learning_rate": 7.245128205128206e-06, "loss": 0.0048, "step": 5875 }, { "epoch": 28.01, "learning_rate": 7.232307692307693e-06, "loss": 0.0038, "step": 5900 }, { "epoch": 28.01, "learning_rate": 7.21948717948718e-06, "loss": 0.0032, "step": 5925 }, { "epoch": 28.01, "learning_rate": 7.206666666666667e-06, "loss": 0.0041, "step": 5950 }, { "epoch": 29.0, "learning_rate": 7.193846153846154e-06, "loss": 0.0043, "step": 5975 }, { "epoch": 29.0, "learning_rate": 7.181025641025642e-06, "loss": 0.0029, "step": 6000 }, { "epoch": 29.0, "eval_loss": 0.3650364875793457, "eval_runtime": 2367.1423, "eval_samples_per_second": 2.784, "eval_steps_per_second": 0.174, "eval_wer": 14.674551419660578, "step": 6000 }, { "epoch": 29.0, "learning_rate": 7.168205128205129e-06, "loss": 0.0045, "step": 6025 }, { "epoch": 29.0, "learning_rate": 7.155384615384616e-06, "loss": 0.0034, "step": 6050 }, { "epoch": 29.01, "learning_rate": 7.142564102564103e-06, "loss": 0.0036, "step": 6075 }, { "epoch": 29.01, "learning_rate": 7.129743589743591e-06, "loss": 0.0031, "step": 6100 }, { "epoch": 29.01, "learning_rate": 7.116923076923078e-06, "loss": 0.0029, "step": 6125 }, { "epoch": 29.01, "learning_rate": 7.104102564102565e-06, "loss": 0.0024, "step": 6150 }, { "epoch": 29.01, "learning_rate": 7.091282051282052e-06, "loss": 0.0023, "step": 6175 }, { "epoch": 30.0, "learning_rate": 7.078461538461538e-06, "loss": 0.0022, "step": 6200 }, { "epoch": 30.0, "learning_rate": 7.065641025641026e-06, "loss": 0.002, "step": 6225 }, { "epoch": 30.0, "learning_rate": 7.052820512820513e-06, "loss": 0.0015, "step": 6250 }, { "epoch": 30.0, "learning_rate": 7.04e-06, "loss": 0.0016, "step": 6275 }, { "epoch": 30.01, "learning_rate": 7.027179487179487e-06, "loss": 0.0018, "step": 6300 }, { "epoch": 30.01, "learning_rate": 7.014358974358975e-06, "loss": 0.0015, "step": 6325 }, { "epoch": 30.01, "learning_rate": 7.001538461538462e-06, "loss": 0.001, "step": 6350 }, { "epoch": 30.01, "learning_rate": 6.988717948717949e-06, "loss": 0.0018, "step": 6375 }, { "epoch": 31.0, "learning_rate": 6.975897435897436e-06, "loss": 0.0015, "step": 6400 }, { "epoch": 31.0, "learning_rate": 6.963076923076924e-06, "loss": 0.0016, "step": 6425 }, { "epoch": 31.0, "learning_rate": 6.950256410256411e-06, "loss": 0.0016, "step": 6450 }, { "epoch": 31.0, "learning_rate": 6.937435897435898e-06, "loss": 0.0017, "step": 6475 }, { "epoch": 31.01, "learning_rate": 6.924615384615385e-06, "loss": 0.0027, "step": 6500 }, { "epoch": 31.01, "learning_rate": 6.911794871794872e-06, "loss": 0.0012, "step": 6525 }, { "epoch": 31.01, "learning_rate": 6.89897435897436e-06, "loss": 0.0016, "step": 6550 }, { "epoch": 31.01, "learning_rate": 6.8861538461538465e-06, "loss": 0.0027, "step": 6575 }, { "epoch": 32.0, "learning_rate": 6.873333333333334e-06, "loss": 0.0023, "step": 6600 }, { "epoch": 32.0, "learning_rate": 6.860512820512821e-06, "loss": 0.0015, "step": 6625 }, { "epoch": 32.0, "learning_rate": 6.847692307692309e-06, "loss": 0.0026, "step": 6650 }, { "epoch": 32.0, "learning_rate": 6.8348717948717955e-06, "loss": 0.0015, "step": 6675 }, { "epoch": 32.01, "learning_rate": 6.822051282051283e-06, "loss": 0.0024, "step": 6700 }, { "epoch": 32.01, "learning_rate": 6.80923076923077e-06, "loss": 0.002, "step": 6725 }, { "epoch": 32.01, "learning_rate": 6.796410256410258e-06, "loss": 0.0017, "step": 6750 }, { "epoch": 32.01, "learning_rate": 6.7835897435897445e-06, "loss": 0.0021, "step": 6775 }, { "epoch": 33.0, "learning_rate": 6.770769230769232e-06, "loss": 0.0015, "step": 6800 }, { "epoch": 33.0, "learning_rate": 6.757948717948718e-06, "loss": 0.0012, "step": 6825 }, { "epoch": 33.0, "learning_rate": 6.745128205128205e-06, "loss": 0.0011, "step": 6850 }, { "epoch": 33.0, "learning_rate": 6.732307692307693e-06, "loss": 0.0019, "step": 6875 }, { "epoch": 33.01, "learning_rate": 6.7194871794871794e-06, "loss": 0.0017, "step": 6900 }, { "epoch": 33.01, "learning_rate": 6.706666666666667e-06, "loss": 0.0015, "step": 6925 }, { "epoch": 33.01, "learning_rate": 6.693846153846154e-06, "loss": 0.0009, "step": 6950 }, { "epoch": 33.01, "learning_rate": 6.681025641025642e-06, "loss": 0.0012, "step": 6975 }, { "epoch": 33.01, "learning_rate": 6.6682051282051284e-06, "loss": 0.0011, "step": 7000 }, { "epoch": 33.01, "eval_loss": 0.36434951424598694, "eval_runtime": 2329.4025, "eval_samples_per_second": 2.829, "eval_steps_per_second": 0.177, "eval_wer": 13.813844222123212, "step": 7000 }, { "epoch": 34.0, "learning_rate": 6.655384615384616e-06, "loss": 0.0017, "step": 7025 }, { "epoch": 34.0, "learning_rate": 6.642564102564103e-06, "loss": 0.0018, "step": 7050 }, { "epoch": 34.0, "learning_rate": 6.629743589743591e-06, "loss": 0.0013, "step": 7075 }, { "epoch": 34.0, "learning_rate": 6.616923076923077e-06, "loss": 0.0022, "step": 7100 }, { "epoch": 34.01, "learning_rate": 6.604102564102565e-06, "loss": 0.0012, "step": 7125 }, { "epoch": 34.01, "learning_rate": 6.591282051282052e-06, "loss": 0.0015, "step": 7150 }, { "epoch": 34.01, "learning_rate": 6.578461538461539e-06, "loss": 0.0011, "step": 7175 }, { "epoch": 34.01, "learning_rate": 6.565641025641026e-06, "loss": 0.0015, "step": 7200 }, { "epoch": 35.0, "learning_rate": 6.552820512820513e-06, "loss": 0.0009, "step": 7225 }, { "epoch": 35.0, "learning_rate": 6.540000000000001e-06, "loss": 0.0003, "step": 7250 }, { "epoch": 35.0, "learning_rate": 6.527179487179488e-06, "loss": 0.0008, "step": 7275 }, { "epoch": 35.0, "learning_rate": 6.514358974358975e-06, "loss": 0.0006, "step": 7300 }, { "epoch": 35.01, "learning_rate": 6.501538461538462e-06, "loss": 0.0013, "step": 7325 }, { "epoch": 35.01, "learning_rate": 6.48871794871795e-06, "loss": 0.0011, "step": 7350 }, { "epoch": 35.01, "learning_rate": 6.475897435897437e-06, "loss": 0.0009, "step": 7375 }, { "epoch": 35.01, "learning_rate": 6.463076923076924e-06, "loss": 0.001, "step": 7400 }, { "epoch": 36.0, "learning_rate": 6.45025641025641e-06, "loss": 0.0008, "step": 7425 }, { "epoch": 36.0, "learning_rate": 6.437435897435897e-06, "loss": 0.0004, "step": 7450 }, { "epoch": 36.0, "learning_rate": 6.424615384615385e-06, "loss": 0.0009, "step": 7475 }, { "epoch": 36.0, "learning_rate": 6.411794871794872e-06, "loss": 0.0004, "step": 7500 }, { "epoch": 36.01, "learning_rate": 6.398974358974359e-06, "loss": 0.0011, "step": 7525 }, { "epoch": 36.01, "learning_rate": 6.386153846153846e-06, "loss": 0.0008, "step": 7550 }, { "epoch": 36.01, "learning_rate": 6.373333333333334e-06, "loss": 0.0003, "step": 7575 }, { "epoch": 36.01, "learning_rate": 6.360512820512821e-06, "loss": 0.0006, "step": 7600 }, { "epoch": 37.0, "learning_rate": 6.347692307692308e-06, "loss": 0.0007, "step": 7625 }, { "epoch": 37.0, "learning_rate": 6.334871794871795e-06, "loss": 0.0011, "step": 7650 }, { "epoch": 37.0, "learning_rate": 6.322051282051283e-06, "loss": 0.0007, "step": 7675 }, { "epoch": 37.0, "learning_rate": 6.30923076923077e-06, "loss": 0.0011, "step": 7700 }, { "epoch": 37.01, "learning_rate": 6.296410256410257e-06, "loss": 0.0004, "step": 7725 }, { "epoch": 37.01, "learning_rate": 6.283589743589744e-06, "loss": 0.0007, "step": 7750 }, { "epoch": 37.01, "learning_rate": 6.270769230769231e-06, "loss": 0.0002, "step": 7775 }, { "epoch": 37.01, "learning_rate": 6.257948717948719e-06, "loss": 0.0004, "step": 7800 }, { "epoch": 37.01, "learning_rate": 6.2451282051282055e-06, "loss": 0.0005, "step": 7825 }, { "epoch": 38.0, "learning_rate": 6.232307692307693e-06, "loss": 0.0015, "step": 7850 }, { "epoch": 38.0, "learning_rate": 6.21948717948718e-06, "loss": 0.0012, "step": 7875 }, { "epoch": 38.0, "learning_rate": 6.206666666666668e-06, "loss": 0.0022, "step": 7900 }, { "epoch": 38.0, "learning_rate": 6.1938461538461545e-06, "loss": 0.0011, "step": 7925 }, { "epoch": 38.01, "learning_rate": 6.181025641025642e-06, "loss": 0.0016, "step": 7950 }, { "epoch": 38.01, "learning_rate": 6.168205128205129e-06, "loss": 0.0011, "step": 7975 }, { "epoch": 38.01, "learning_rate": 6.155384615384617e-06, "loss": 0.0006, "step": 8000 }, { "epoch": 38.01, "eval_loss": 0.3627721071243286, "eval_runtime": 2357.9498, "eval_samples_per_second": 2.795, "eval_steps_per_second": 0.175, "eval_wer": 14.004212402284418, "step": 8000 }, { "epoch": 38.01, "learning_rate": 6.142564102564103e-06, "loss": 0.0007, "step": 8025 }, { "epoch": 39.0, "learning_rate": 6.1297435897435895e-06, "loss": 0.0014, "step": 8050 }, { "epoch": 39.0, "learning_rate": 6.116923076923077e-06, "loss": 0.0008, "step": 8075 }, { "epoch": 39.0, "learning_rate": 6.104102564102564e-06, "loss": 0.0017, "step": 8100 }, { "epoch": 39.0, "learning_rate": 6.091282051282052e-06, "loss": 0.0017, "step": 8125 }, { "epoch": 39.01, "learning_rate": 6.0784615384615384e-06, "loss": 0.0021, "step": 8150 }, { "epoch": 39.01, "learning_rate": 6.065641025641026e-06, "loss": 0.0017, "step": 8175 }, { "epoch": 39.01, "learning_rate": 6.052820512820513e-06, "loss": 0.0015, "step": 8200 }, { "epoch": 39.01, "learning_rate": 6.040000000000001e-06, "loss": 0.0012, "step": 8225 }, { "epoch": 40.0, "learning_rate": 6.0271794871794874e-06, "loss": 0.0006, "step": 8250 }, { "epoch": 40.0, "learning_rate": 6.014358974358975e-06, "loss": 0.001, "step": 8275 }, { "epoch": 40.0, "learning_rate": 6.001538461538462e-06, "loss": 0.0015, "step": 8300 }, { "epoch": 40.0, "learning_rate": 5.98871794871795e-06, "loss": 0.0025, "step": 8325 }, { "epoch": 40.01, "learning_rate": 5.9758974358974364e-06, "loss": 0.0032, "step": 8350 }, { "epoch": 40.01, "learning_rate": 5.963076923076923e-06, "loss": 0.0031, "step": 8375 }, { "epoch": 40.01, "learning_rate": 5.950256410256411e-06, "loss": 0.0022, "step": 8400 }, { "epoch": 40.01, "learning_rate": 5.937435897435898e-06, "loss": 0.0016, "step": 8425 }, { "epoch": 41.0, "learning_rate": 5.9246153846153854e-06, "loss": 0.0014, "step": 8450 }, { "epoch": 41.0, "learning_rate": 5.911794871794872e-06, "loss": 0.0011, "step": 8475 }, { "epoch": 41.0, "learning_rate": 5.89897435897436e-06, "loss": 0.0016, "step": 8500 }, { "epoch": 41.0, "learning_rate": 5.886153846153847e-06, "loss": 0.0013, "step": 8525 }, { "epoch": 41.01, "learning_rate": 5.873333333333334e-06, "loss": 0.002, "step": 8550 }, { "epoch": 41.01, "learning_rate": 5.860512820512821e-06, "loss": 0.0024, "step": 8575 }, { "epoch": 41.01, "learning_rate": 5.847692307692309e-06, "loss": 0.0015, "step": 8600 }, { "epoch": 41.01, "learning_rate": 5.834871794871796e-06, "loss": 0.0009, "step": 8625 }, { "epoch": 41.01, "learning_rate": 5.822051282051282e-06, "loss": 0.0013, "step": 8650 }, { "epoch": 42.0, "learning_rate": 5.809230769230769e-06, "loss": 0.0007, "step": 8675 }, { "epoch": 42.0, "learning_rate": 5.796410256410256e-06, "loss": 0.001, "step": 8700 }, { "epoch": 42.0, "learning_rate": 5.783589743589744e-06, "loss": 0.001, "step": 8725 }, { "epoch": 42.0, "learning_rate": 5.770769230769231e-06, "loss": 0.0008, "step": 8750 }, { "epoch": 42.01, "learning_rate": 5.757948717948718e-06, "loss": 0.0013, "step": 8775 }, { "epoch": 42.01, "learning_rate": 5.745128205128205e-06, "loss": 0.0009, "step": 8800 }, { "epoch": 42.01, "learning_rate": 5.732307692307693e-06, "loss": 0.0006, "step": 8825 }, { "epoch": 42.01, "learning_rate": 5.71948717948718e-06, "loss": 0.0006, "step": 8850 }, { "epoch": 43.0, "learning_rate": 5.706666666666667e-06, "loss": 0.0014, "step": 8875 }, { "epoch": 43.0, "learning_rate": 5.693846153846154e-06, "loss": 0.0005, "step": 8900 }, { "epoch": 43.0, "learning_rate": 5.681025641025642e-06, "loss": 0.0004, "step": 8925 }, { "epoch": 43.0, "learning_rate": 5.668205128205129e-06, "loss": 0.0008, "step": 8950 }, { "epoch": 43.01, "learning_rate": 5.6553846153846155e-06, "loss": 0.0005, "step": 8975 }, { "epoch": 43.01, "learning_rate": 5.642564102564103e-06, "loss": 0.0009, "step": 9000 }, { "epoch": 43.01, "eval_loss": 0.366083562374115, "eval_runtime": 2341.4448, "eval_samples_per_second": 2.815, "eval_steps_per_second": 0.176, "eval_wer": 14.004212402284418, "step": 9000 }, { "epoch": 43.01, "learning_rate": 5.62974358974359e-06, "loss": 0.0003, "step": 9025 }, { "epoch": 43.01, "learning_rate": 5.616923076923078e-06, "loss": 0.0002, "step": 9050 }, { "epoch": 44.0, "learning_rate": 5.6041025641025645e-06, "loss": 0.0004, "step": 9075 }, { "epoch": 44.0, "learning_rate": 5.591282051282052e-06, "loss": 0.0002, "step": 9100 }, { "epoch": 44.0, "learning_rate": 5.578461538461539e-06, "loss": 0.0009, "step": 9125 }, { "epoch": 44.0, "learning_rate": 5.565641025641027e-06, "loss": 0.0004, "step": 9150 }, { "epoch": 44.01, "learning_rate": 5.5528205128205135e-06, "loss": 0.0009, "step": 9175 }, { "epoch": 44.01, "learning_rate": 5.540000000000001e-06, "loss": 0.0005, "step": 9200 }, { "epoch": 44.01, "learning_rate": 5.527179487179488e-06, "loss": 0.0003, "step": 9225 }, { "epoch": 44.01, "learning_rate": 5.514358974358974e-06, "loss": 0.0004, "step": 9250 }, { "epoch": 45.0, "learning_rate": 5.501538461538462e-06, "loss": 0.0004, "step": 9275 }, { "epoch": 45.0, "learning_rate": 5.4887179487179485e-06, "loss": 0.0005, "step": 9300 }, { "epoch": 45.0, "learning_rate": 5.475897435897436e-06, "loss": 0.0002, "step": 9325 }, { "epoch": 45.0, "learning_rate": 5.463076923076923e-06, "loss": 0.0003, "step": 9350 }, { "epoch": 45.01, "learning_rate": 5.450256410256411e-06, "loss": 0.0012, "step": 9375 }, { "epoch": 45.01, "learning_rate": 5.4374358974358975e-06, "loss": 0.0012, "step": 9400 }, { "epoch": 45.01, "learning_rate": 5.424615384615385e-06, "loss": 0.001, "step": 9425 }, { "epoch": 45.01, "learning_rate": 5.411794871794872e-06, "loss": 0.0004, "step": 9450 }, { "epoch": 45.01, "learning_rate": 5.39897435897436e-06, "loss": 0.0011, "step": 9475 }, { "epoch": 46.0, "learning_rate": 5.3861538461538464e-06, "loss": 0.0006, "step": 9500 }, { "epoch": 46.0, "learning_rate": 5.373333333333334e-06, "loss": 0.0021, "step": 9525 }, { "epoch": 46.0, "learning_rate": 5.360512820512821e-06, "loss": 0.0007, "step": 9550 }, { "epoch": 46.0, "learning_rate": 5.347692307692308e-06, "loss": 0.0004, "step": 9575 }, { "epoch": 46.01, "learning_rate": 5.3348717948717954e-06, "loss": 0.0006, "step": 9600 }, { "epoch": 46.01, "learning_rate": 5.322051282051282e-06, "loss": 0.0003, "step": 9625 }, { "epoch": 46.01, "learning_rate": 5.30923076923077e-06, "loss": 0.0003, "step": 9650 }, { "epoch": 46.01, "learning_rate": 5.296410256410257e-06, "loss": 0.0002, "step": 9675 }, { "epoch": 47.0, "learning_rate": 5.2835897435897444e-06, "loss": 0.0005, "step": 9700 }, { "epoch": 47.0, "learning_rate": 5.270769230769231e-06, "loss": 0.0003, "step": 9725 }, { "epoch": 47.0, "learning_rate": 5.257948717948719e-06, "loss": 0.0001, "step": 9750 }, { "epoch": 47.0, "learning_rate": 5.245128205128206e-06, "loss": 0.0001, "step": 9775 }, { "epoch": 47.01, "learning_rate": 5.2323076923076934e-06, "loss": 0.0001, "step": 9800 }, { "epoch": 47.01, "learning_rate": 5.21948717948718e-06, "loss": 0.0003, "step": 9825 }, { "epoch": 47.01, "learning_rate": 5.206666666666668e-06, "loss": 0.0002, "step": 9850 }, { "epoch": 47.01, "learning_rate": 5.193846153846154e-06, "loss": 0.0006, "step": 9875 }, { "epoch": 48.0, "learning_rate": 5.181025641025641e-06, "loss": 0.0008, "step": 9900 }, { "epoch": 48.0, "learning_rate": 5.168205128205128e-06, "loss": 0.0002, "step": 9925 }, { "epoch": 48.0, "learning_rate": 5.155384615384615e-06, "loss": 0.0003, "step": 9950 }, { "epoch": 48.0, "learning_rate": 5.142564102564103e-06, "loss": 0.0007, "step": 9975 }, { "epoch": 48.01, "learning_rate": 5.12974358974359e-06, "loss": 0.0003, "step": 10000 }, { "epoch": 48.01, "eval_loss": 0.3793827295303345, "eval_runtime": 2359.3516, "eval_samples_per_second": 2.794, "eval_steps_per_second": 0.175, "eval_wer": 13.716634938636641, "step": 10000 }, { "epoch": 48.01, "learning_rate": 5.116923076923077e-06, "loss": 0.0005, "step": 10025 }, { "epoch": 48.01, "learning_rate": 5.104102564102564e-06, "loss": 0.0004, "step": 10050 }, { "epoch": 48.01, "learning_rate": 5.091282051282052e-06, "loss": 0.0008, "step": 10075 }, { "epoch": 49.0, "learning_rate": 5.078461538461539e-06, "loss": 0.0004, "step": 10100 }, { "epoch": 49.0, "learning_rate": 5.065641025641026e-06, "loss": 0.0003, "step": 10125 }, { "epoch": 49.0, "learning_rate": 5.052820512820513e-06, "loss": 0.0007, "step": 10150 }, { "epoch": 49.0, "learning_rate": 5.04e-06, "loss": 0.0018, "step": 10175 }, { "epoch": 49.01, "learning_rate": 5.027179487179488e-06, "loss": 0.0008, "step": 10200 }, { "epoch": 49.01, "learning_rate": 5.0143589743589745e-06, "loss": 0.0008, "step": 10225 }, { "epoch": 49.01, "learning_rate": 5.001538461538462e-06, "loss": 0.0008, "step": 10250 }, { "epoch": 49.01, "learning_rate": 4.988717948717949e-06, "loss": 0.0007, "step": 10275 }, { "epoch": 49.01, "learning_rate": 4.975897435897437e-06, "loss": 0.0011, "step": 10300 }, { "epoch": 50.0, "learning_rate": 4.9630769230769235e-06, "loss": 0.0009, "step": 10325 }, { "epoch": 50.0, "learning_rate": 4.95025641025641e-06, "loss": 0.0013, "step": 10350 }, { "epoch": 50.0, "learning_rate": 4.937435897435898e-06, "loss": 0.0013, "step": 10375 }, { "epoch": 50.01, "learning_rate": 4.924615384615385e-06, "loss": 0.0008, "step": 10400 }, { "epoch": 50.01, "learning_rate": 4.9117948717948725e-06, "loss": 0.0017, "step": 10425 }, { "epoch": 50.01, "learning_rate": 4.898974358974359e-06, "loss": 0.0023, "step": 10450 }, { "epoch": 50.01, "learning_rate": 4.886153846153847e-06, "loss": 0.0021, "step": 10475 }, { "epoch": 50.01, "learning_rate": 4.873333333333334e-06, "loss": 0.0009, "step": 10500 }, { "epoch": 51.0, "learning_rate": 4.860512820512821e-06, "loss": 0.0008, "step": 10525 }, { "epoch": 51.0, "learning_rate": 4.847692307692308e-06, "loss": 0.0013, "step": 10550 }, { "epoch": 51.0, "learning_rate": 4.834871794871795e-06, "loss": 0.0008, "step": 10575 }, { "epoch": 51.0, "learning_rate": 4.822051282051283e-06, "loss": 0.0009, "step": 10600 }, { "epoch": 51.01, "learning_rate": 4.80923076923077e-06, "loss": 0.0008, "step": 10625 }, { "epoch": 51.01, "learning_rate": 4.7964102564102565e-06, "loss": 0.0017, "step": 10650 }, { "epoch": 51.01, "learning_rate": 4.783589743589744e-06, "loss": 0.0011, "step": 10675 }, { "epoch": 51.01, "learning_rate": 4.770769230769231e-06, "loss": 0.0008, "step": 10700 }, { "epoch": 52.0, "learning_rate": 4.757948717948719e-06, "loss": 0.0011, "step": 10725 }, { "epoch": 52.0, "learning_rate": 4.7451282051282055e-06, "loss": 0.0005, "step": 10750 }, { "epoch": 52.0, "learning_rate": 4.732307692307693e-06, "loss": 0.0011, "step": 10775 }, { "epoch": 52.0, "learning_rate": 4.71948717948718e-06, "loss": 0.0006, "step": 10800 }, { "epoch": 52.01, "learning_rate": 4.706666666666667e-06, "loss": 0.0002, "step": 10825 }, { "epoch": 52.01, "learning_rate": 4.6938461538461544e-06, "loss": 0.0004, "step": 10850 }, { "epoch": 52.01, "learning_rate": 4.681025641025641e-06, "loss": 0.0009, "step": 10875 }, { "epoch": 52.01, "learning_rate": 4.668205128205129e-06, "loss": 0.0005, "step": 10900 }, { "epoch": 53.0, "learning_rate": 4.655384615384616e-06, "loss": 0.0008, "step": 10925 }, { "epoch": 53.0, "learning_rate": 4.642564102564103e-06, "loss": 0.0003, "step": 10950 }, { "epoch": 53.0, "learning_rate": 4.62974358974359e-06, "loss": 0.0004, "step": 10975 }, { "epoch": 53.0, "learning_rate": 4.616923076923077e-06, "loss": 0.0003, "step": 11000 }, { "epoch": 53.0, "eval_loss": 0.379313200712204, "eval_runtime": 2348.6859, "eval_samples_per_second": 2.806, "eval_steps_per_second": 0.175, "eval_wer": 13.692332617764997, "step": 11000 }, { "epoch": 53.01, "learning_rate": 4.604102564102565e-06, "loss": 0.0005, "step": 11025 }, { "epoch": 53.01, "learning_rate": 4.591282051282052e-06, "loss": 0.0002, "step": 11050 }, { "epoch": 53.01, "learning_rate": 4.578461538461539e-06, "loss": 0.0002, "step": 11075 }, { "epoch": 53.01, "learning_rate": 4.565641025641026e-06, "loss": 0.0007, "step": 11100 }, { "epoch": 54.0, "learning_rate": 4.552820512820513e-06, "loss": 0.0002, "step": 11125 }, { "epoch": 54.0, "learning_rate": 4.540000000000001e-06, "loss": 0.0002, "step": 11150 }, { "epoch": 54.0, "learning_rate": 4.527179487179487e-06, "loss": 0.0003, "step": 11175 }, { "epoch": 54.0, "learning_rate": 4.514358974358975e-06, "loss": 0.0003, "step": 11200 }, { "epoch": 54.01, "learning_rate": 4.501538461538462e-06, "loss": 0.0003, "step": 11225 }, { "epoch": 54.01, "learning_rate": 4.488717948717949e-06, "loss": 0.0002, "step": 11250 }, { "epoch": 54.01, "learning_rate": 4.475897435897436e-06, "loss": 0.0003, "step": 11275 }, { "epoch": 54.01, "learning_rate": 4.463076923076923e-06, "loss": 0.0001, "step": 11300 }, { "epoch": 54.01, "learning_rate": 4.450256410256411e-06, "loss": 0.0003, "step": 11325 }, { "epoch": 55.0, "learning_rate": 4.437435897435898e-06, "loss": 0.0001, "step": 11350 }, { "epoch": 55.0, "learning_rate": 4.424615384615385e-06, "loss": 0.0002, "step": 11375 }, { "epoch": 55.0, "learning_rate": 4.411794871794872e-06, "loss": 0.0003, "step": 11400 }, { "epoch": 55.0, "learning_rate": 4.398974358974359e-06, "loss": 0.0002, "step": 11425 }, { "epoch": 55.01, "learning_rate": 4.386153846153847e-06, "loss": 0.0001, "step": 11450 }, { "epoch": 55.01, "learning_rate": 4.3733333333333335e-06, "loss": 0.0001, "step": 11475 }, { "epoch": 55.01, "learning_rate": 4.360512820512821e-06, "loss": 0.0001, "step": 11500 }, { "epoch": 55.01, "learning_rate": 4.347692307692308e-06, "loss": 0.0001, "step": 11525 }, { "epoch": 56.0, "learning_rate": 4.334871794871795e-06, "loss": 0.0001, "step": 11550 }, { "epoch": 56.0, "learning_rate": 4.3220512820512825e-06, "loss": 0.0001, "step": 11575 }, { "epoch": 56.0, "learning_rate": 4.309230769230769e-06, "loss": 0.0001, "step": 11600 }, { "epoch": 56.0, "learning_rate": 4.296410256410257e-06, "loss": 0.0001, "step": 11625 }, { "epoch": 56.01, "learning_rate": 4.283589743589744e-06, "loss": 0.0001, "step": 11650 }, { "epoch": 56.01, "learning_rate": 4.2707692307692315e-06, "loss": 0.0, "step": 11675 }, { "epoch": 56.01, "learning_rate": 4.257948717948718e-06, "loss": 0.0, "step": 11700 }, { "epoch": 56.01, "learning_rate": 4.245128205128205e-06, "loss": 0.0, "step": 11725 }, { "epoch": 57.0, "learning_rate": 4.232307692307693e-06, "loss": 0.0, "step": 11750 }, { "epoch": 57.0, "learning_rate": 4.21948717948718e-06, "loss": 0.0, "step": 11775 }, { "epoch": 57.0, "learning_rate": 4.206666666666667e-06, "loss": 0.0, "step": 11800 }, { "epoch": 57.0, "learning_rate": 4.193846153846154e-06, "loss": 0.0, "step": 11825 }, { "epoch": 57.01, "learning_rate": 4.181025641025641e-06, "loss": 0.0, "step": 11850 }, { "epoch": 57.01, "learning_rate": 4.168205128205129e-06, "loss": 0.0, "step": 11875 }, { "epoch": 57.01, "learning_rate": 4.1553846153846155e-06, "loss": 0.0, "step": 11900 }, { "epoch": 57.01, "learning_rate": 4.142564102564103e-06, "loss": 0.0, "step": 11925 }, { "epoch": 58.0, "learning_rate": 4.12974358974359e-06, "loss": 0.0, "step": 11950 }, { "epoch": 58.0, "learning_rate": 4.116923076923078e-06, "loss": 0.0, "step": 11975 }, { "epoch": 58.0, "learning_rate": 4.1041025641025645e-06, "loss": 0.0, "step": 12000 }, { "epoch": 58.0, "eval_loss": 0.3991302251815796, "eval_runtime": 2351.4062, "eval_samples_per_second": 2.803, "eval_steps_per_second": 0.175, "eval_wer": 13.402729960711248, "step": 12000 }, { "epoch": 58.0, "learning_rate": 4.091282051282051e-06, "loss": 0.0, "step": 12025 }, { "epoch": 58.01, "learning_rate": 4.078461538461539e-06, "loss": 0.0, "step": 12050 }, { "epoch": 58.01, "learning_rate": 4.065641025641026e-06, "loss": 0.0, "step": 12075 }, { "epoch": 58.01, "learning_rate": 4.0528205128205135e-06, "loss": 0.0, "step": 12100 }, { "epoch": 58.01, "learning_rate": 4.04e-06, "loss": 0.0, "step": 12125 }, { "epoch": 58.01, "learning_rate": 4.027179487179487e-06, "loss": 0.0, "step": 12150 }, { "epoch": 59.0, "learning_rate": 4.014358974358975e-06, "loss": 0.0, "step": 12175 }, { "epoch": 59.0, "learning_rate": 4.001538461538462e-06, "loss": 0.0, "step": 12200 }, { "epoch": 59.0, "learning_rate": 3.988717948717949e-06, "loss": 0.0, "step": 12225 }, { "epoch": 59.0, "learning_rate": 3.975897435897436e-06, "loss": 0.0, "step": 12250 }, { "epoch": 59.01, "learning_rate": 3.963076923076924e-06, "loss": 0.0, "step": 12275 }, { "epoch": 59.01, "learning_rate": 3.950256410256411e-06, "loss": 0.0, "step": 12300 }, { "epoch": 59.01, "learning_rate": 3.937435897435897e-06, "loss": 0.0, "step": 12325 }, { "epoch": 59.01, "learning_rate": 3.924615384615385e-06, "loss": 0.0, "step": 12350 }, { "epoch": 60.0, "learning_rate": 3.911794871794872e-06, "loss": 0.0, "step": 12375 }, { "epoch": 60.0, "learning_rate": 3.89897435897436e-06, "loss": 0.0, "step": 12400 }, { "epoch": 60.0, "learning_rate": 3.886153846153846e-06, "loss": 0.0, "step": 12425 }, { "epoch": 60.0, "learning_rate": 3.873333333333333e-06, "loss": 0.0, "step": 12450 }, { "epoch": 60.01, "learning_rate": 3.860512820512821e-06, "loss": 0.0, "step": 12475 }, { "epoch": 60.01, "learning_rate": 3.847692307692308e-06, "loss": 0.0, "step": 12500 }, { "epoch": 60.01, "learning_rate": 3.834871794871795e-06, "loss": 0.0, "step": 12525 }, { "epoch": 60.01, "learning_rate": 3.822051282051282e-06, "loss": 0.0, "step": 12550 }, { "epoch": 61.0, "learning_rate": 3.8092307692307695e-06, "loss": 0.0, "step": 12575 }, { "epoch": 61.0, "learning_rate": 3.7964102564102567e-06, "loss": 0.0, "step": 12600 }, { "epoch": 61.0, "learning_rate": 3.783589743589744e-06, "loss": 0.0, "step": 12625 }, { "epoch": 61.0, "learning_rate": 3.7707692307692312e-06, "loss": 0.0, "step": 12650 }, { "epoch": 61.01, "learning_rate": 3.7579487179487185e-06, "loss": 0.0, "step": 12675 }, { "epoch": 61.01, "learning_rate": 3.7451282051282057e-06, "loss": 0.0, "step": 12700 }, { "epoch": 61.01, "learning_rate": 3.732307692307693e-06, "loss": 0.0, "step": 12725 }, { "epoch": 61.01, "learning_rate": 3.7194871794871794e-06, "loss": 0.0, "step": 12750 }, { "epoch": 62.0, "learning_rate": 3.7066666666666666e-06, "loss": 0.0, "step": 12775 }, { "epoch": 62.0, "learning_rate": 3.693846153846154e-06, "loss": 0.0, "step": 12800 }, { "epoch": 62.0, "learning_rate": 3.681025641025641e-06, "loss": 0.0, "step": 12825 }, { "epoch": 62.0, "learning_rate": 3.6682051282051284e-06, "loss": 0.0, "step": 12850 }, { "epoch": 62.01, "learning_rate": 3.6553846153846156e-06, "loss": 0.0, "step": 12875 }, { "epoch": 62.01, "learning_rate": 3.642564102564103e-06, "loss": 0.0, "step": 12900 }, { "epoch": 62.01, "learning_rate": 3.62974358974359e-06, "loss": 0.0, "step": 12925 }, { "epoch": 62.01, "learning_rate": 3.6169230769230773e-06, "loss": 0.0, "step": 12950 }, { "epoch": 62.01, "learning_rate": 3.6041025641025646e-06, "loss": 0.0, "step": 12975 }, { "epoch": 63.0, "learning_rate": 3.591282051282052e-06, "loss": 0.0, "step": 13000 }, { "epoch": 63.0, "eval_loss": 0.41185224056243896, "eval_runtime": 2328.4103, "eval_samples_per_second": 2.831, "eval_steps_per_second": 0.177, "eval_wer": 13.356150512373933, "step": 13000 }, { "epoch": 63.0, "learning_rate": 3.578461538461539e-06, "loss": 0.0, "step": 13025 }, { "epoch": 63.0, "learning_rate": 3.5656410256410263e-06, "loss": 0.0, "step": 13050 }, { "epoch": 63.0, "learning_rate": 3.5528205128205127e-06, "loss": 0.0, "step": 13075 }, { "epoch": 63.01, "learning_rate": 3.54e-06, "loss": 0.0, "step": 13100 }, { "epoch": 63.01, "learning_rate": 3.5271794871794872e-06, "loss": 0.0, "step": 13125 }, { "epoch": 63.01, "learning_rate": 3.5143589743589745e-06, "loss": 0.0, "step": 13150 }, { "epoch": 63.01, "learning_rate": 3.5015384615384617e-06, "loss": 0.0, "step": 13175 }, { "epoch": 64.0, "learning_rate": 3.488717948717949e-06, "loss": 0.0, "step": 13200 }, { "epoch": 64.0, "learning_rate": 3.4758974358974362e-06, "loss": 0.0, "step": 13225 }, { "epoch": 64.0, "learning_rate": 3.4630769230769235e-06, "loss": 0.0, "step": 13250 }, { "epoch": 64.0, "learning_rate": 3.4502564102564107e-06, "loss": 0.0, "step": 13275 }, { "epoch": 64.01, "learning_rate": 3.437435897435898e-06, "loss": 0.0, "step": 13300 }, { "epoch": 64.01, "learning_rate": 3.4246153846153852e-06, "loss": 0.0, "step": 13325 }, { "epoch": 64.01, "learning_rate": 3.4117948717948725e-06, "loss": 0.0, "step": 13350 }, { "epoch": 64.01, "learning_rate": 3.398974358974359e-06, "loss": 0.0, "step": 13375 }, { "epoch": 65.0, "learning_rate": 3.386153846153846e-06, "loss": 0.0, "step": 13400 }, { "epoch": 65.0, "learning_rate": 3.3733333333333334e-06, "loss": 0.0, "step": 13425 }, { "epoch": 65.0, "learning_rate": 3.3605128205128206e-06, "loss": 0.0, "step": 13450 }, { "epoch": 65.0, "learning_rate": 3.347692307692308e-06, "loss": 0.0, "step": 13475 }, { "epoch": 65.01, "learning_rate": 3.334871794871795e-06, "loss": 0.0, "step": 13500 }, { "epoch": 65.01, "learning_rate": 3.3220512820512824e-06, "loss": 0.0, "step": 13525 }, { "epoch": 65.01, "learning_rate": 3.3092307692307696e-06, "loss": 0.0, "step": 13550 }, { "epoch": 65.01, "learning_rate": 3.296410256410257e-06, "loss": 0.0, "step": 13575 }, { "epoch": 66.0, "learning_rate": 3.283589743589744e-06, "loss": 0.0, "step": 13600 }, { "epoch": 66.0, "learning_rate": 3.2707692307692313e-06, "loss": 0.0, "step": 13625 }, { "epoch": 66.0, "learning_rate": 3.2579487179487186e-06, "loss": 0.0, "step": 13650 }, { "epoch": 66.0, "learning_rate": 3.245128205128205e-06, "loss": 0.0, "step": 13675 }, { "epoch": 66.01, "learning_rate": 3.2323076923076922e-06, "loss": 0.0, "step": 13700 }, { "epoch": 66.01, "learning_rate": 3.2194871794871795e-06, "loss": 0.0, "step": 13725 }, { "epoch": 66.01, "learning_rate": 3.2066666666666667e-06, "loss": 0.0, "step": 13750 }, { "epoch": 66.01, "learning_rate": 3.193846153846154e-06, "loss": 0.0, "step": 13775 }, { "epoch": 66.01, "learning_rate": 3.1810256410256412e-06, "loss": 0.0, "step": 13800 }, { "epoch": 67.0, "learning_rate": 3.1682051282051285e-06, "loss": 0.0, "step": 13825 }, { "epoch": 67.0, "learning_rate": 3.1553846153846157e-06, "loss": 0.0, "step": 13850 }, { "epoch": 67.0, "learning_rate": 3.142564102564103e-06, "loss": 0.0, "step": 13875 }, { "epoch": 67.0, "learning_rate": 3.1297435897435902e-06, "loss": 0.0, "step": 13900 }, { "epoch": 67.01, "learning_rate": 3.1169230769230775e-06, "loss": 0.0, "step": 13925 }, { "epoch": 67.01, "learning_rate": 3.1041025641025647e-06, "loss": 0.0, "step": 13950 }, { "epoch": 67.01, "learning_rate": 3.091282051282051e-06, "loss": 0.0, "step": 13975 }, { "epoch": 67.01, "learning_rate": 3.0784615384615384e-06, "loss": 0.0, "step": 14000 }, { "epoch": 67.01, "eval_loss": 0.42092934250831604, "eval_runtime": 2298.1193, "eval_samples_per_second": 2.868, "eval_steps_per_second": 0.179, "eval_wer": 13.218437360767954, "step": 14000 }, { "epoch": 68.0, "learning_rate": 3.0656410256410256e-06, "loss": 0.0, "step": 14025 }, { "epoch": 68.0, "learning_rate": 3.052820512820513e-06, "loss": 0.0, "step": 14050 }, { "epoch": 68.0, "learning_rate": 3.04e-06, "loss": 0.0, "step": 14075 }, { "epoch": 68.0, "learning_rate": 3.0271794871794874e-06, "loss": 0.0, "step": 14100 }, { "epoch": 68.01, "learning_rate": 3.0143589743589746e-06, "loss": 0.0, "step": 14125 }, { "epoch": 68.01, "learning_rate": 3.001538461538462e-06, "loss": 0.0, "step": 14150 }, { "epoch": 68.01, "learning_rate": 2.988717948717949e-06, "loss": 0.0, "step": 14175 }, { "epoch": 68.01, "learning_rate": 2.9758974358974364e-06, "loss": 0.0, "step": 14200 }, { "epoch": 69.0, "learning_rate": 2.9630769230769236e-06, "loss": 0.0, "step": 14225 }, { "epoch": 69.0, "learning_rate": 2.950256410256411e-06, "loss": 0.0, "step": 14250 }, { "epoch": 69.0, "learning_rate": 2.9374358974358973e-06, "loss": 0.0, "step": 14275 }, { "epoch": 69.0, "learning_rate": 2.9246153846153845e-06, "loss": 0.0, "step": 14300 }, { "epoch": 69.01, "learning_rate": 2.9117948717948717e-06, "loss": 0.0, "step": 14325 }, { "epoch": 69.01, "learning_rate": 2.898974358974359e-06, "loss": 0.0, "step": 14350 }, { "epoch": 69.01, "learning_rate": 2.8861538461538462e-06, "loss": 0.0, "step": 14375 }, { "epoch": 69.01, "learning_rate": 2.8733333333333335e-06, "loss": 0.0, "step": 14400 }, { "epoch": 70.0, "learning_rate": 2.8605128205128207e-06, "loss": 0.0, "step": 14425 }, { "epoch": 70.0, "learning_rate": 2.847692307692308e-06, "loss": 0.0, "step": 14450 }, { "epoch": 70.0, "learning_rate": 2.8348717948717952e-06, "loss": 0.0, "step": 14475 }, { "epoch": 70.0, "learning_rate": 2.8220512820512825e-06, "loss": 0.0, "step": 14500 }, { "epoch": 70.01, "learning_rate": 2.8092307692307697e-06, "loss": 0.0, "step": 14525 }, { "epoch": 70.01, "learning_rate": 2.796410256410257e-06, "loss": 0.0, "step": 14550 }, { "epoch": 70.01, "learning_rate": 2.7835897435897434e-06, "loss": 0.0, "step": 14575 }, { "epoch": 70.01, "learning_rate": 2.7707692307692306e-06, "loss": 0.0, "step": 14600 }, { "epoch": 70.01, "learning_rate": 2.757948717948718e-06, "loss": 0.0, "step": 14625 }, { "epoch": 71.0, "learning_rate": 2.745128205128205e-06, "loss": 0.0, "step": 14650 }, { "epoch": 71.0, "learning_rate": 2.7323076923076924e-06, "loss": 0.0, "step": 14675 }, { "epoch": 71.0, "learning_rate": 2.7194871794871796e-06, "loss": 0.0, "step": 14700 }, { "epoch": 71.0, "learning_rate": 2.706666666666667e-06, "loss": 0.0, "step": 14725 }, { "epoch": 71.01, "learning_rate": 2.693846153846154e-06, "loss": 0.0, "step": 14750 }, { "epoch": 71.01, "learning_rate": 2.6810256410256414e-06, "loss": 0.0, "step": 14775 }, { "epoch": 71.01, "learning_rate": 2.6682051282051286e-06, "loss": 0.0, "step": 14800 }, { "epoch": 71.01, "learning_rate": 2.655384615384616e-06, "loss": 0.0, "step": 14825 }, { "epoch": 72.0, "learning_rate": 2.642564102564103e-06, "loss": 0.0, "step": 14850 }, { "epoch": 72.0, "learning_rate": 2.6297435897435904e-06, "loss": 0.0, "step": 14875 }, { "epoch": 72.0, "learning_rate": 2.6169230769230768e-06, "loss": 0.0, "step": 14900 }, { "epoch": 72.0, "learning_rate": 2.604102564102564e-06, "loss": 0.0, "step": 14925 }, { "epoch": 72.01, "learning_rate": 2.5912820512820513e-06, "loss": 0.0, "step": 14950 }, { "epoch": 72.01, "learning_rate": 2.5784615384615385e-06, "loss": 0.0, "step": 14975 }, { "epoch": 72.01, "learning_rate": 2.5656410256410257e-06, "loss": 0.0, "step": 15000 }, { "epoch": 72.01, "eval_loss": 0.4287641942501068, "eval_runtime": 2294.9192, "eval_samples_per_second": 2.872, "eval_steps_per_second": 0.18, "eval_wer": 13.222487747579894, "step": 15000 }, { "epoch": 72.01, "learning_rate": 2.552820512820513e-06, "loss": 0.0, "step": 15025 }, { "epoch": 73.0, "learning_rate": 2.5400000000000002e-06, "loss": 0.0, "step": 15050 }, { "epoch": 73.0, "learning_rate": 2.5271794871794875e-06, "loss": 0.0, "step": 15075 }, { "epoch": 73.0, "learning_rate": 2.5143589743589747e-06, "loss": 0.0, "step": 15100 }, { "epoch": 73.0, "learning_rate": 2.501538461538462e-06, "loss": 0.0, "step": 15125 }, { "epoch": 73.01, "learning_rate": 2.488717948717949e-06, "loss": 0.0, "step": 15150 }, { "epoch": 73.01, "learning_rate": 2.475897435897436e-06, "loss": 0.0, "step": 15175 }, { "epoch": 73.01, "learning_rate": 2.4630769230769233e-06, "loss": 0.0, "step": 15200 }, { "epoch": 73.01, "learning_rate": 2.4502564102564106e-06, "loss": 0.0, "step": 15225 }, { "epoch": 74.0, "learning_rate": 2.437435897435898e-06, "loss": 0.0, "step": 15250 }, { "epoch": 74.0, "learning_rate": 2.4246153846153846e-06, "loss": 0.0, "step": 15275 }, { "epoch": 74.0, "learning_rate": 2.411794871794872e-06, "loss": 0.0, "step": 15300 }, { "epoch": 74.0, "learning_rate": 2.398974358974359e-06, "loss": 0.0, "step": 15325 }, { "epoch": 74.01, "learning_rate": 2.3861538461538464e-06, "loss": 0.0, "step": 15350 }, { "epoch": 74.01, "learning_rate": 2.3733333333333336e-06, "loss": 0.0, "step": 15375 }, { "epoch": 74.01, "learning_rate": 2.360512820512821e-06, "loss": 0.0, "step": 15400 }, { "epoch": 74.01, "learning_rate": 2.3476923076923077e-06, "loss": 0.0, "step": 15425 }, { "epoch": 74.01, "learning_rate": 2.334871794871795e-06, "loss": 0.0, "step": 15450 }, { "epoch": 75.0, "learning_rate": 2.322051282051282e-06, "loss": 0.0, "step": 15475 }, { "epoch": 75.0, "learning_rate": 2.3092307692307694e-06, "loss": 0.0, "step": 15500 }, { "epoch": 75.0, "learning_rate": 2.2964102564102567e-06, "loss": 0.0, "step": 15525 }, { "epoch": 75.0, "learning_rate": 2.283589743589744e-06, "loss": 0.0, "step": 15550 }, { "epoch": 75.01, "learning_rate": 2.2707692307692308e-06, "loss": 0.0, "step": 15575 }, { "epoch": 75.01, "learning_rate": 2.257948717948718e-06, "loss": 0.0, "step": 15600 }, { "epoch": 75.01, "learning_rate": 2.2451282051282053e-06, "loss": 0.0, "step": 15625 }, { "epoch": 75.01, "learning_rate": 2.2323076923076925e-06, "loss": 0.0, "step": 15650 }, { "epoch": 76.0, "learning_rate": 2.2194871794871797e-06, "loss": 0.0, "step": 15675 }, { "epoch": 76.0, "learning_rate": 2.206666666666667e-06, "loss": 0.0, "step": 15700 }, { "epoch": 76.0, "learning_rate": 2.193846153846154e-06, "loss": 0.0, "step": 15725 }, { "epoch": 76.0, "learning_rate": 2.181025641025641e-06, "loss": 0.0, "step": 15750 }, { "epoch": 76.01, "learning_rate": 2.1682051282051283e-06, "loss": 0.0, "step": 15775 }, { "epoch": 76.01, "learning_rate": 2.1553846153846156e-06, "loss": 0.0, "step": 15800 }, { "epoch": 76.01, "learning_rate": 2.142564102564103e-06, "loss": 0.0, "step": 15825 }, { "epoch": 76.01, "learning_rate": 2.12974358974359e-06, "loss": 0.0, "step": 15850 }, { "epoch": 77.0, "learning_rate": 2.116923076923077e-06, "loss": 0.0, "step": 15875 }, { "epoch": 77.0, "learning_rate": 2.104102564102564e-06, "loss": 0.0, "step": 15900 }, { "epoch": 77.0, "learning_rate": 2.0912820512820514e-06, "loss": 0.0, "step": 15925 }, { "epoch": 77.0, "learning_rate": 2.0784615384615386e-06, "loss": 0.0, "step": 15950 }, { "epoch": 77.01, "learning_rate": 2.065641025641026e-06, "loss": 0.0, "step": 15975 }, { "epoch": 77.01, "learning_rate": 2.052820512820513e-06, "loss": 0.0, "step": 16000 }, { "epoch": 77.01, "eval_loss": 0.43614599108695984, "eval_runtime": 2286.8558, "eval_samples_per_second": 2.882, "eval_steps_per_second": 0.18, "eval_wer": 13.151605978370934, "step": 16000 }, { "epoch": 77.01, "learning_rate": 2.04e-06, "loss": 0.0, "step": 16025 }, { "epoch": 77.01, "learning_rate": 2.027179487179487e-06, "loss": 0.0, "step": 16050 }, { "epoch": 78.0, "learning_rate": 2.0143589743589744e-06, "loss": 0.0, "step": 16075 }, { "epoch": 78.0, "learning_rate": 2.0015384615384617e-06, "loss": 0.0, "step": 16100 }, { "epoch": 78.0, "learning_rate": 1.988717948717949e-06, "loss": 0.0, "step": 16125 }, { "epoch": 78.0, "learning_rate": 1.975897435897436e-06, "loss": 0.0, "step": 16150 }, { "epoch": 78.01, "learning_rate": 1.963076923076923e-06, "loss": 0.0, "step": 16175 }, { "epoch": 78.01, "learning_rate": 1.9502564102564103e-06, "loss": 0.0, "step": 16200 }, { "epoch": 78.01, "learning_rate": 1.9374358974358975e-06, "loss": 0.0, "step": 16225 }, { "epoch": 78.01, "learning_rate": 1.9246153846153848e-06, "loss": 0.0, "step": 16250 }, { "epoch": 79.0, "learning_rate": 1.911794871794872e-06, "loss": 0.0, "step": 16275 }, { "epoch": 79.0, "learning_rate": 1.8989743589743593e-06, "loss": 0.0, "step": 16300 }, { "epoch": 79.0, "learning_rate": 1.8861538461538465e-06, "loss": 0.0, "step": 16325 }, { "epoch": 79.0, "learning_rate": 1.8733333333333333e-06, "loss": 0.0, "step": 16350 }, { "epoch": 79.01, "learning_rate": 1.8605128205128206e-06, "loss": 0.0, "step": 16375 }, { "epoch": 79.01, "learning_rate": 1.8476923076923078e-06, "loss": 0.0, "step": 16400 }, { "epoch": 79.01, "learning_rate": 1.834871794871795e-06, "loss": 0.0, "step": 16425 }, { "epoch": 79.01, "learning_rate": 1.8220512820512823e-06, "loss": 0.0, "step": 16450 }, { "epoch": 79.01, "learning_rate": 1.8092307692307696e-06, "loss": 0.0, "step": 16475 }, { "epoch": 80.0, "learning_rate": 1.7964102564102564e-06, "loss": 0.0, "step": 16500 }, { "epoch": 80.0, "learning_rate": 1.7835897435897436e-06, "loss": 0.0, "step": 16525 }, { "epoch": 80.0, "learning_rate": 1.7707692307692309e-06, "loss": 0.0, "step": 16550 }, { "epoch": 80.0, "learning_rate": 1.7579487179487181e-06, "loss": 0.0, "step": 16575 }, { "epoch": 80.01, "learning_rate": 1.7451282051282054e-06, "loss": 0.0, "step": 16600 }, { "epoch": 80.01, "learning_rate": 1.7323076923076926e-06, "loss": 0.0, "step": 16625 }, { "epoch": 80.01, "learning_rate": 1.7194871794871795e-06, "loss": 0.0, "step": 16650 }, { "epoch": 80.01, "learning_rate": 1.7066666666666667e-06, "loss": 0.0, "step": 16675 }, { "epoch": 81.0, "learning_rate": 1.693846153846154e-06, "loss": 0.0, "step": 16700 }, { "epoch": 81.0, "learning_rate": 1.6810256410256412e-06, "loss": 0.0, "step": 16725 }, { "epoch": 81.0, "learning_rate": 1.6682051282051284e-06, "loss": 0.0, "step": 16750 }, { "epoch": 81.0, "learning_rate": 1.6553846153846157e-06, "loss": 0.0, "step": 16775 }, { "epoch": 81.01, "learning_rate": 1.6425641025641025e-06, "loss": 0.0, "step": 16800 }, { "epoch": 81.01, "learning_rate": 1.6297435897435898e-06, "loss": 0.0, "step": 16825 }, { "epoch": 81.01, "learning_rate": 1.616923076923077e-06, "loss": 0.0, "step": 16850 }, { "epoch": 81.01, "learning_rate": 1.6041025641025643e-06, "loss": 0.0, "step": 16875 }, { "epoch": 82.0, "learning_rate": 1.5912820512820515e-06, "loss": 0.0, "step": 16900 }, { "epoch": 82.0, "learning_rate": 1.5784615384615388e-06, "loss": 0.0, "step": 16925 }, { "epoch": 82.0, "learning_rate": 1.5656410256410256e-06, "loss": 0.0, "step": 16950 }, { "epoch": 82.0, "learning_rate": 1.5528205128205128e-06, "loss": 0.0, "step": 16975 }, { "epoch": 82.01, "learning_rate": 1.54e-06, "loss": 0.0, "step": 17000 }, { "epoch": 82.01, "eval_loss": 0.4427773058414459, "eval_runtime": 2300.5112, "eval_samples_per_second": 2.865, "eval_steps_per_second": 0.179, "eval_wer": 13.133379237717202, "step": 17000 }, { "epoch": 82.01, "learning_rate": 1.5271794871794873e-06, "loss": 0.0, "step": 17025 }, { "epoch": 82.01, "learning_rate": 1.5143589743589746e-06, "loss": 0.0, "step": 17050 }, { "epoch": 82.01, "learning_rate": 1.5015384615384618e-06, "loss": 0.0, "step": 17075 }, { "epoch": 83.0, "learning_rate": 1.4887179487179486e-06, "loss": 0.0, "step": 17100 }, { "epoch": 83.0, "learning_rate": 1.475897435897436e-06, "loss": 0.0, "step": 17125 }, { "epoch": 83.0, "learning_rate": 1.4630769230769231e-06, "loss": 0.0, "step": 17150 }, { "epoch": 83.0, "learning_rate": 1.4502564102564104e-06, "loss": 0.0, "step": 17175 }, { "epoch": 83.01, "learning_rate": 1.4374358974358976e-06, "loss": 0.0, "step": 17200 }, { "epoch": 83.01, "learning_rate": 1.4246153846153849e-06, "loss": 0.0, "step": 17225 }, { "epoch": 83.01, "learning_rate": 1.4117948717948717e-06, "loss": 0.0, "step": 17250 }, { "epoch": 83.01, "learning_rate": 1.398974358974359e-06, "loss": 0.0, "step": 17275 }, { "epoch": 83.01, "learning_rate": 1.3861538461538462e-06, "loss": 0.0, "step": 17300 }, { "epoch": 84.0, "learning_rate": 1.3733333333333335e-06, "loss": 0.0, "step": 17325 }, { "epoch": 84.0, "learning_rate": 1.3605128205128207e-06, "loss": 0.0, "step": 17350 }, { "epoch": 84.0, "learning_rate": 1.347692307692308e-06, "loss": 0.0, "step": 17375 }, { "epoch": 84.0, "learning_rate": 1.3348717948717948e-06, "loss": 0.0, "step": 17400 }, { "epoch": 84.01, "learning_rate": 1.322051282051282e-06, "loss": 0.0, "step": 17425 }, { "epoch": 84.01, "learning_rate": 1.3092307692307693e-06, "loss": 0.0, "step": 17450 }, { "epoch": 84.01, "learning_rate": 1.2964102564102565e-06, "loss": 0.0, "step": 17475 }, { "epoch": 84.01, "learning_rate": 1.2835897435897438e-06, "loss": 0.0, "step": 17500 }, { "epoch": 85.0, "learning_rate": 1.270769230769231e-06, "loss": 0.0, "step": 17525 }, { "epoch": 85.0, "learning_rate": 1.2579487179487178e-06, "loss": 0.0, "step": 17550 }, { "epoch": 85.0, "learning_rate": 1.2451282051282053e-06, "loss": 0.0, "step": 17575 }, { "epoch": 85.0, "learning_rate": 1.2323076923076923e-06, "loss": 0.0, "step": 17600 }, { "epoch": 85.01, "learning_rate": 1.2194871794871796e-06, "loss": 0.0, "step": 17625 }, { "epoch": 85.01, "learning_rate": 1.2066666666666668e-06, "loss": 0.0, "step": 17650 }, { "epoch": 85.01, "learning_rate": 1.1938461538461539e-06, "loss": 0.0, "step": 17675 }, { "epoch": 85.01, "learning_rate": 1.1810256410256411e-06, "loss": 0.0, "step": 17700 }, { "epoch": 86.0, "learning_rate": 1.1682051282051284e-06, "loss": 0.0, "step": 17725 }, { "epoch": 86.0, "learning_rate": 1.1553846153846154e-06, "loss": 0.0, "step": 17750 }, { "epoch": 86.0, "learning_rate": 1.1425641025641026e-06, "loss": 0.0, "step": 17775 }, { "epoch": 86.0, "learning_rate": 1.12974358974359e-06, "loss": 0.0, "step": 17800 }, { "epoch": 86.01, "learning_rate": 1.116923076923077e-06, "loss": 0.0, "step": 17825 }, { "epoch": 86.01, "learning_rate": 1.1041025641025642e-06, "loss": 0.0, "step": 17850 }, { "epoch": 86.01, "learning_rate": 1.0912820512820514e-06, "loss": 0.0, "step": 17875 }, { "epoch": 86.01, "learning_rate": 1.0784615384615385e-06, "loss": 0.0, "step": 17900 }, { "epoch": 87.0, "learning_rate": 1.0656410256410257e-06, "loss": 0.0, "step": 17925 }, { "epoch": 87.0, "learning_rate": 1.052820512820513e-06, "loss": 0.0, "step": 17950 }, { "epoch": 87.0, "learning_rate": 1.04e-06, "loss": 0.0, "step": 17975 }, { "epoch": 87.0, "learning_rate": 1.0271794871794872e-06, "loss": 0.0, "step": 18000 }, { "epoch": 87.0, "eval_loss": 0.4487149715423584, "eval_runtime": 2241.1505, "eval_samples_per_second": 2.941, "eval_steps_per_second": 0.184, "eval_wer": 13.133379237717202, "step": 18000 }, { "epoch": 87.01, "learning_rate": 1.0143589743589745e-06, "loss": 0.0, "step": 18025 }, { "epoch": 87.01, "learning_rate": 1.0015384615384615e-06, "loss": 0.0, "step": 18050 }, { "epoch": 87.01, "learning_rate": 9.887179487179488e-07, "loss": 0.0, "step": 18075 }, { "epoch": 87.01, "learning_rate": 9.75897435897436e-07, "loss": 0.0, "step": 18100 }, { "epoch": 87.01, "learning_rate": 9.63076923076923e-07, "loss": 0.0, "step": 18125 }, { "epoch": 88.0, "learning_rate": 9.502564102564103e-07, "loss": 0.0, "step": 18150 }, { "epoch": 88.0, "learning_rate": 9.374358974358976e-07, "loss": 0.0, "step": 18175 }, { "epoch": 88.0, "learning_rate": 9.246153846153846e-07, "loss": 0.0, "step": 18200 }, { "epoch": 88.0, "learning_rate": 9.117948717948718e-07, "loss": 0.0, "step": 18225 }, { "epoch": 88.01, "learning_rate": 8.989743589743591e-07, "loss": 0.0, "step": 18250 }, { "epoch": 88.01, "learning_rate": 8.861538461538461e-07, "loss": 0.0, "step": 18275 }, { "epoch": 88.01, "learning_rate": 8.733333333333334e-07, "loss": 0.0, "step": 18300 }, { "epoch": 88.01, "learning_rate": 8.605128205128206e-07, "loss": 0.0, "step": 18325 }, { "epoch": 89.0, "learning_rate": 8.476923076923077e-07, "loss": 0.0, "step": 18350 }, { "epoch": 89.0, "learning_rate": 8.348717948717949e-07, "loss": 0.0, "step": 18375 }, { "epoch": 89.0, "learning_rate": 8.220512820512822e-07, "loss": 0.0, "step": 18400 }, { "epoch": 89.0, "learning_rate": 8.092307692307692e-07, "loss": 0.0, "step": 18425 }, { "epoch": 89.01, "learning_rate": 7.964102564102564e-07, "loss": 0.0, "step": 18450 }, { "epoch": 89.01, "learning_rate": 7.835897435897437e-07, "loss": 0.0, "step": 18475 }, { "epoch": 89.01, "learning_rate": 7.707692307692307e-07, "loss": 0.0, "step": 18500 }, { "epoch": 89.01, "learning_rate": 7.57948717948718e-07, "loss": 0.0, "step": 18525 }, { "epoch": 90.0, "learning_rate": 7.451282051282052e-07, "loss": 0.0, "step": 18550 }, { "epoch": 90.0, "learning_rate": 7.323076923076923e-07, "loss": 0.0, "step": 18575 }, { "epoch": 90.0, "learning_rate": 7.194871794871795e-07, "loss": 0.0, "step": 18600 }, { "epoch": 90.0, "learning_rate": 7.066666666666667e-07, "loss": 0.0, "step": 18625 }, { "epoch": 90.01, "learning_rate": 6.938461538461538e-07, "loss": 0.0, "step": 18650 }, { "epoch": 90.01, "learning_rate": 6.81025641025641e-07, "loss": 0.0, "step": 18675 }, { "epoch": 90.01, "learning_rate": 6.682051282051283e-07, "loss": 0.0, "step": 18700 }, { "epoch": 90.01, "learning_rate": 6.553846153846154e-07, "loss": 0.0, "step": 18725 }, { "epoch": 91.0, "learning_rate": 6.425641025641026e-07, "loss": 0.0, "step": 18750 }, { "epoch": 91.0, "learning_rate": 6.297435897435898e-07, "loss": 0.0, "step": 18775 }, { "epoch": 91.0, "learning_rate": 6.16923076923077e-07, "loss": 0.0, "step": 18800 }, { "epoch": 91.0, "learning_rate": 6.041025641025641e-07, "loss": 0.0, "step": 18825 }, { "epoch": 91.01, "learning_rate": 5.912820512820513e-07, "loss": 0.0, "step": 18850 }, { "epoch": 91.01, "learning_rate": 5.784615384615385e-07, "loss": 0.0, "step": 18875 }, { "epoch": 91.01, "learning_rate": 5.656410256410256e-07, "loss": 0.0, "step": 18900 }, { "epoch": 91.01, "learning_rate": 5.528205128205129e-07, "loss": 0.0, "step": 18925 }, { "epoch": 91.01, "learning_rate": 5.4e-07, "loss": 0.0, "step": 18950 }, { "epoch": 92.0, "learning_rate": 5.271794871794872e-07, "loss": 0.0, "step": 18975 }, { "epoch": 92.0, "learning_rate": 5.143589743589744e-07, "loss": 0.0, "step": 19000 }, { "epoch": 92.0, "eval_loss": 0.4530625343322754, "eval_runtime": 2288.6602, "eval_samples_per_second": 2.88, "eval_steps_per_second": 0.18, "eval_wer": 12.989590505893311, "step": 19000 }, { "epoch": 92.0, "learning_rate": 5.015384615384615e-07, "loss": 0.0, "step": 19025 }, { "epoch": 92.0, "learning_rate": 4.887179487179487e-07, "loss": 0.0, "step": 19050 }, { "epoch": 92.01, "learning_rate": 4.7589743589743594e-07, "loss": 0.0, "step": 19075 }, { "epoch": 92.01, "learning_rate": 4.630769230769231e-07, "loss": 0.0, "step": 19100 }, { "epoch": 92.01, "learning_rate": 4.502564102564103e-07, "loss": 0.0, "step": 19125 }, { "epoch": 92.01, "learning_rate": 4.3743589743589747e-07, "loss": 0.0, "step": 19150 }, { "epoch": 93.0, "learning_rate": 4.246153846153846e-07, "loss": 0.0, "step": 19175 }, { "epoch": 93.0, "learning_rate": 4.1179487179487186e-07, "loss": 0.0, "step": 19200 }, { "epoch": 93.0, "learning_rate": 3.98974358974359e-07, "loss": 0.0, "step": 19225 }, { "epoch": 93.0, "learning_rate": 3.8615384615384615e-07, "loss": 0.0, "step": 19250 }, { "epoch": 93.01, "learning_rate": 3.733333333333334e-07, "loss": 0.0, "step": 19275 }, { "epoch": 93.01, "learning_rate": 3.6051282051282054e-07, "loss": 0.0, "step": 19300 }, { "epoch": 93.01, "learning_rate": 3.476923076923077e-07, "loss": 0.0, "step": 19325 }, { "epoch": 93.01, "learning_rate": 3.348717948717949e-07, "loss": 0.0, "step": 19350 }, { "epoch": 94.0, "learning_rate": 3.2205128205128207e-07, "loss": 0.0, "step": 19375 }, { "epoch": 94.0, "learning_rate": 3.0923076923076926e-07, "loss": 0.0, "step": 19400 }, { "epoch": 94.0, "learning_rate": 2.964102564102564e-07, "loss": 0.0, "step": 19425 }, { "epoch": 94.0, "learning_rate": 2.835897435897436e-07, "loss": 0.0, "step": 19450 }, { "epoch": 94.01, "learning_rate": 2.707692307692308e-07, "loss": 0.0, "step": 19475 }, { "epoch": 94.01, "learning_rate": 2.5794871794871794e-07, "loss": 0.0, "step": 19500 }, { "epoch": 94.01, "learning_rate": 2.4512820512820513e-07, "loss": 0.0, "step": 19525 }, { "epoch": 94.01, "learning_rate": 2.3230769230769233e-07, "loss": 0.0, "step": 19550 }, { "epoch": 95.0, "learning_rate": 2.194871794871795e-07, "loss": 0.0, "step": 19575 }, { "epoch": 95.0, "learning_rate": 2.066666666666667e-07, "loss": 0.0, "step": 19600 }, { "epoch": 95.0, "learning_rate": 1.9384615384615386e-07, "loss": 0.0, "step": 19625 }, { "epoch": 95.0, "learning_rate": 1.8102564102564103e-07, "loss": 0.0, "step": 19650 }, { "epoch": 95.01, "learning_rate": 1.6820512820512822e-07, "loss": 0.0, "step": 19675 }, { "epoch": 95.01, "learning_rate": 1.553846153846154e-07, "loss": 0.0, "step": 19700 }, { "epoch": 95.01, "learning_rate": 1.4256410256410259e-07, "loss": 0.0, "step": 19725 }, { "epoch": 95.01, "learning_rate": 1.2974358974358975e-07, "loss": 0.0, "step": 19750 }, { "epoch": 95.01, "learning_rate": 1.1692307692307694e-07, "loss": 0.0, "step": 19775 }, { "epoch": 96.0, "learning_rate": 1.041025641025641e-07, "loss": 0.0, "step": 19800 }, { "epoch": 96.0, "learning_rate": 9.128205128205129e-08, "loss": 0.0, "step": 19825 }, { "epoch": 96.0, "learning_rate": 7.846153846153847e-08, "loss": 0.0, "step": 19850 }, { "epoch": 96.0, "learning_rate": 6.564102564102564e-08, "loss": 0.0, "step": 19875 }, { "epoch": 96.01, "learning_rate": 5.2820512820512826e-08, "loss": 0.0, "step": 19900 }, { "epoch": 96.01, "learning_rate": 4e-08, "loss": 0.0, "step": 19925 }, { "epoch": 96.01, "learning_rate": 2.717948717948718e-08, "loss": 0.0, "step": 19950 }, { "epoch": 96.01, "learning_rate": 1.4358974358974361e-08, "loss": 0.0, "step": 19975 }, { "epoch": 97.0, "learning_rate": 1.5384615384615387e-09, "loss": 0.0, "step": 20000 }, { "epoch": 97.0, "eval_loss": 0.4549092948436737, "eval_runtime": 2295.1962, "eval_samples_per_second": 2.872, "eval_steps_per_second": 0.18, "eval_wer": 12.981489732269432, "step": 20000 }, { "epoch": 97.0, "step": 20000, "total_flos": 2.7172470114459614e+21, "train_loss": 0.014709487067163355, "train_runtime": 388431.4847, "train_samples_per_second": 3.295, "train_steps_per_second": 0.051 } ], "logging_steps": 25, "max_steps": 20000, "num_train_epochs": 9223372036854775807, "save_steps": 1000, "total_flos": 2.7172470114459614e+21, "trial_name": null, "trial_params": null }