{ "best_metric": 0.7431560204762965, "best_model_checkpoint": "/storage/elsayedissa/whisper-large-v2-japanese-24h/checkpoint-3000", "epoch": 22.900763358778626, "global_step": 3000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.19, "learning_rate": 4.6000000000000004e-07, "loss": 0.3048, "step": 25 }, { "epoch": 0.38, "learning_rate": 9.600000000000001e-07, "loss": 0.2896, "step": 50 }, { "epoch": 0.57, "learning_rate": 1.46e-06, "loss": 0.2433, "step": 75 }, { "epoch": 0.76, "learning_rate": 1.9600000000000003e-06, "loss": 0.2303, "step": 100 }, { "epoch": 0.95, "learning_rate": 2.4400000000000004e-06, "loss": 0.2237, "step": 125 }, { "epoch": 1.15, "learning_rate": 2.9400000000000002e-06, "loss": 0.1888, "step": 150 }, { "epoch": 1.34, "learning_rate": 3.44e-06, "loss": 0.1636, "step": 175 }, { "epoch": 1.53, "learning_rate": 3.94e-06, "loss": 0.1607, "step": 200 }, { "epoch": 1.72, "learning_rate": 4.440000000000001e-06, "loss": 0.1674, "step": 225 }, { "epoch": 1.91, "learning_rate": 4.94e-06, "loss": 0.1656, "step": 250 }, { "epoch": 2.1, "learning_rate": 5.4400000000000004e-06, "loss": 0.1375, "step": 275 }, { "epoch": 2.29, "learning_rate": 5.94e-06, "loss": 0.0927, "step": 300 }, { "epoch": 2.48, "learning_rate": 6.440000000000001e-06, "loss": 0.1047, "step": 325 }, { "epoch": 2.67, "learning_rate": 6.9400000000000005e-06, "loss": 0.1162, "step": 350 }, { "epoch": 2.86, "learning_rate": 7.420000000000001e-06, "loss": 0.1093, "step": 375 }, { "epoch": 3.05, "learning_rate": 7.92e-06, "loss": 0.0929, "step": 400 }, { "epoch": 3.24, "learning_rate": 8.42e-06, "loss": 0.0587, "step": 425 }, { "epoch": 3.44, "learning_rate": 8.920000000000001e-06, "loss": 0.0672, "step": 450 }, { "epoch": 3.63, "learning_rate": 9.42e-06, "loss": 0.0695, "step": 475 }, { "epoch": 3.82, "learning_rate": 9.920000000000002e-06, "loss": 0.0706, "step": 500 }, { "epoch": 4.01, "learning_rate": 9.953333333333333e-06, "loss": 0.0652, "step": 525 }, { "epoch": 4.2, "learning_rate": 9.89777777777778e-06, "loss": 0.0411, "step": 550 }, { "epoch": 4.39, "learning_rate": 9.842222222222223e-06, "loss": 0.043, "step": 575 }, { "epoch": 4.58, "learning_rate": 9.786666666666667e-06, "loss": 0.0422, "step": 600 }, { "epoch": 4.77, "learning_rate": 9.731111111111113e-06, "loss": 0.0424, "step": 625 }, { "epoch": 4.96, "learning_rate": 9.675555555555555e-06, "loss": 0.0447, "step": 650 }, { "epoch": 5.15, "learning_rate": 9.620000000000001e-06, "loss": 0.0294, "step": 675 }, { "epoch": 5.34, "learning_rate": 9.564444444444445e-06, "loss": 0.0252, "step": 700 }, { "epoch": 5.53, "learning_rate": 9.508888888888889e-06, "loss": 0.0272, "step": 725 }, { "epoch": 5.73, "learning_rate": 9.453333333333335e-06, "loss": 0.027, "step": 750 }, { "epoch": 5.92, "learning_rate": 9.397777777777779e-06, "loss": 0.0273, "step": 775 }, { "epoch": 6.11, "learning_rate": 9.342222222222223e-06, "loss": 0.0195, "step": 800 }, { "epoch": 6.3, "learning_rate": 9.286666666666667e-06, "loss": 0.0184, "step": 825 }, { "epoch": 6.49, "learning_rate": 9.231111111111111e-06, "loss": 0.0156, "step": 850 }, { "epoch": 6.68, "learning_rate": 9.175555555555557e-06, "loss": 0.0176, "step": 875 }, { "epoch": 6.87, "learning_rate": 9.12e-06, "loss": 0.0181, "step": 900 }, { "epoch": 7.06, "learning_rate": 9.064444444444447e-06, "loss": 0.0138, "step": 925 }, { "epoch": 7.25, "learning_rate": 9.008888888888889e-06, "loss": 0.0103, "step": 950 }, { "epoch": 7.44, "learning_rate": 8.953333333333335e-06, "loss": 0.0116, "step": 975 }, { "epoch": 7.63, "learning_rate": 8.897777777777779e-06, "loss": 0.0111, "step": 1000 }, { "epoch": 7.63, "eval_loss": 0.320997953414917, "eval_runtime": 1728.016, "eval_samples_per_second": 2.595, "eval_steps_per_second": 0.163, "eval_wer": 0.78878255063432, "step": 1000 }, { "epoch": 7.82, "learning_rate": 8.842222222222223e-06, "loss": 0.01, "step": 1025 }, { "epoch": 8.02, "learning_rate": 8.786666666666668e-06, "loss": 0.0094, "step": 1050 }, { "epoch": 8.21, "learning_rate": 8.73111111111111e-06, "loss": 0.0098, "step": 1075 }, { "epoch": 8.4, "learning_rate": 8.675555555555556e-06, "loss": 0.0097, "step": 1100 }, { "epoch": 8.59, "learning_rate": 8.62e-06, "loss": 0.008, "step": 1125 }, { "epoch": 8.78, "learning_rate": 8.564444444444445e-06, "loss": 0.0085, "step": 1150 }, { "epoch": 8.97, "learning_rate": 8.50888888888889e-06, "loss": 0.0081, "step": 1175 }, { "epoch": 9.16, "learning_rate": 8.453333333333334e-06, "loss": 0.0062, "step": 1200 }, { "epoch": 9.35, "learning_rate": 8.397777777777778e-06, "loss": 0.0063, "step": 1225 }, { "epoch": 9.54, "learning_rate": 8.342222222222222e-06, "loss": 0.0052, "step": 1250 }, { "epoch": 9.73, "learning_rate": 8.286666666666668e-06, "loss": 0.0048, "step": 1275 }, { "epoch": 9.92, "learning_rate": 8.231111111111112e-06, "loss": 0.0053, "step": 1300 }, { "epoch": 10.11, "learning_rate": 8.175555555555556e-06, "loss": 0.004, "step": 1325 }, { "epoch": 10.31, "learning_rate": 8.120000000000002e-06, "loss": 0.003, "step": 1350 }, { "epoch": 10.5, "learning_rate": 8.064444444444444e-06, "loss": 0.0044, "step": 1375 }, { "epoch": 10.69, "learning_rate": 8.00888888888889e-06, "loss": 0.0039, "step": 1400 }, { "epoch": 10.88, "learning_rate": 7.953333333333334e-06, "loss": 0.0038, "step": 1425 }, { "epoch": 11.07, "learning_rate": 7.897777777777778e-06, "loss": 0.0032, "step": 1450 }, { "epoch": 11.26, "learning_rate": 7.842222222222224e-06, "loss": 0.0017, "step": 1475 }, { "epoch": 11.45, "learning_rate": 7.786666666666666e-06, "loss": 0.0017, "step": 1500 }, { "epoch": 11.64, "learning_rate": 7.731111111111112e-06, "loss": 0.0024, "step": 1525 }, { "epoch": 11.83, "learning_rate": 7.675555555555556e-06, "loss": 0.0019, "step": 1550 }, { "epoch": 12.02, "learning_rate": 7.620000000000001e-06, "loss": 0.0027, "step": 1575 }, { "epoch": 12.21, "learning_rate": 7.564444444444446e-06, "loss": 0.0019, "step": 1600 }, { "epoch": 12.4, "learning_rate": 7.50888888888889e-06, "loss": 0.0027, "step": 1625 }, { "epoch": 12.6, "learning_rate": 7.453333333333334e-06, "loss": 0.0019, "step": 1650 }, { "epoch": 12.79, "learning_rate": 7.3977777777777786e-06, "loss": 0.0013, "step": 1675 }, { "epoch": 12.98, "learning_rate": 7.342222222222223e-06, "loss": 0.0013, "step": 1700 }, { "epoch": 13.17, "learning_rate": 7.2866666666666675e-06, "loss": 0.0011, "step": 1725 }, { "epoch": 13.36, "learning_rate": 7.231111111111112e-06, "loss": 0.0011, "step": 1750 }, { "epoch": 13.55, "learning_rate": 7.1755555555555556e-06, "loss": 0.0009, "step": 1775 }, { "epoch": 13.74, "learning_rate": 7.1200000000000004e-06, "loss": 0.0009, "step": 1800 }, { "epoch": 13.93, "learning_rate": 7.0644444444444445e-06, "loss": 0.0013, "step": 1825 }, { "epoch": 14.12, "learning_rate": 7.008888888888889e-06, "loss": 0.0014, "step": 1850 }, { "epoch": 14.31, "learning_rate": 6.953333333333334e-06, "loss": 0.0011, "step": 1875 }, { "epoch": 14.5, "learning_rate": 6.897777777777779e-06, "loss": 0.0007, "step": 1900 }, { "epoch": 14.69, "learning_rate": 6.842222222222222e-06, "loss": 0.0008, "step": 1925 }, { "epoch": 14.89, "learning_rate": 6.786666666666667e-06, "loss": 0.0009, "step": 1950 }, { "epoch": 15.08, "learning_rate": 6.731111111111111e-06, "loss": 0.0006, "step": 1975 }, { "epoch": 15.27, "learning_rate": 6.675555555555556e-06, "loss": 0.0007, "step": 2000 }, { "epoch": 15.27, "eval_loss": 0.35849472880363464, "eval_runtime": 1718.5005, "eval_samples_per_second": 2.61, "eval_steps_per_second": 0.164, "eval_wer": 0.7478299577119965, "step": 2000 }, { "epoch": 15.46, "learning_rate": 6.620000000000001e-06, "loss": 0.0007, "step": 2025 }, { "epoch": 15.65, "learning_rate": 6.564444444444446e-06, "loss": 0.0006, "step": 2050 }, { "epoch": 15.84, "learning_rate": 6.508888888888889e-06, "loss": 0.0005, "step": 2075 }, { "epoch": 16.03, "learning_rate": 6.453333333333334e-06, "loss": 0.0005, "step": 2100 }, { "epoch": 16.22, "learning_rate": 6.397777777777778e-06, "loss": 0.0005, "step": 2125 }, { "epoch": 16.41, "learning_rate": 6.342222222222223e-06, "loss": 0.0005, "step": 2150 }, { "epoch": 16.6, "learning_rate": 6.286666666666668e-06, "loss": 0.0004, "step": 2175 }, { "epoch": 16.79, "learning_rate": 6.231111111111111e-06, "loss": 0.0004, "step": 2200 }, { "epoch": 16.98, "learning_rate": 6.175555555555556e-06, "loss": 0.0004, "step": 2225 }, { "epoch": 17.18, "learning_rate": 6.120000000000001e-06, "loss": 0.0004, "step": 2250 }, { "epoch": 17.37, "learning_rate": 6.064444444444445e-06, "loss": 0.0004, "step": 2275 }, { "epoch": 17.56, "learning_rate": 6.00888888888889e-06, "loss": 0.0004, "step": 2300 }, { "epoch": 17.75, "learning_rate": 5.9533333333333345e-06, "loss": 0.0004, "step": 2325 }, { "epoch": 17.94, "learning_rate": 5.897777777777778e-06, "loss": 0.0004, "step": 2350 }, { "epoch": 18.13, "learning_rate": 5.8422222222222226e-06, "loss": 0.0004, "step": 2375 }, { "epoch": 18.32, "learning_rate": 5.7866666666666674e-06, "loss": 0.0003, "step": 2400 }, { "epoch": 18.51, "learning_rate": 5.7311111111111115e-06, "loss": 0.0003, "step": 2425 }, { "epoch": 18.7, "learning_rate": 5.675555555555556e-06, "loss": 0.0003, "step": 2450 }, { "epoch": 18.89, "learning_rate": 5.620000000000001e-06, "loss": 0.0004, "step": 2475 }, { "epoch": 19.08, "learning_rate": 5.5644444444444444e-06, "loss": 0.0003, "step": 2500 }, { "epoch": 19.27, "learning_rate": 5.508888888888889e-06, "loss": 0.0003, "step": 2525 }, { "epoch": 19.47, "learning_rate": 5.453333333333334e-06, "loss": 0.0003, "step": 2550 }, { "epoch": 19.66, "learning_rate": 5.397777777777778e-06, "loss": 0.0003, "step": 2575 }, { "epoch": 19.85, "learning_rate": 5.342222222222223e-06, "loss": 0.0003, "step": 2600 }, { "epoch": 20.04, "learning_rate": 5.286666666666666e-06, "loss": 0.0003, "step": 2625 }, { "epoch": 20.23, "learning_rate": 5.231111111111111e-06, "loss": 0.0003, "step": 2650 }, { "epoch": 20.42, "learning_rate": 5.175555555555556e-06, "loss": 0.0003, "step": 2675 }, { "epoch": 20.61, "learning_rate": 5.12e-06, "loss": 0.0003, "step": 2700 }, { "epoch": 20.8, "learning_rate": 5.064444444444445e-06, "loss": 0.0003, "step": 2725 }, { "epoch": 20.99, "learning_rate": 5.00888888888889e-06, "loss": 0.0003, "step": 2750 }, { "epoch": 21.18, "learning_rate": 4.953333333333334e-06, "loss": 0.0003, "step": 2775 }, { "epoch": 21.37, "learning_rate": 4.897777777777778e-06, "loss": 0.0003, "step": 2800 }, { "epoch": 21.56, "learning_rate": 4.842222222222223e-06, "loss": 0.0003, "step": 2825 }, { "epoch": 21.76, "learning_rate": 4.786666666666667e-06, "loss": 0.0003, "step": 2850 }, { "epoch": 21.95, "learning_rate": 4.731111111111112e-06, "loss": 0.0003, "step": 2875 }, { "epoch": 22.14, "learning_rate": 4.675555555555556e-06, "loss": 0.0003, "step": 2900 }, { "epoch": 22.33, "learning_rate": 4.620000000000001e-06, "loss": 0.0002, "step": 2925 }, { "epoch": 22.52, "learning_rate": 4.564444444444445e-06, "loss": 0.0002, "step": 2950 }, { "epoch": 22.71, "learning_rate": 4.50888888888889e-06, "loss": 0.0002, "step": 2975 }, { "epoch": 22.9, "learning_rate": 4.453333333333334e-06, "loss": 0.0003, "step": 3000 }, { "epoch": 22.9, "eval_loss": 0.3937181532382965, "eval_runtime": 1714.4778, "eval_samples_per_second": 2.616, "eval_steps_per_second": 0.164, "eval_wer": 0.7431560204762965, "step": 3000 } ], "max_steps": 5000, "num_train_epochs": 39, "total_flos": 3.16373713809408e+20, "trial_name": null, "trial_params": null }