{ "best_metric": 70.1955074875208, "best_model_checkpoint": "./whisper-it/checkpoint-300", "epoch": 5.916030534351145, "eval_steps": 100, "global_step": 3100, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.05, "learning_rate": 4.4e-07, "loss": 3.4673, "step": 25 }, { "epoch": 0.1, "learning_rate": 9.400000000000001e-07, "loss": 2.7417, "step": 50 }, { "epoch": 0.14, "learning_rate": 1.44e-06, "loss": 2.0307, "step": 75 }, { "epoch": 0.19, "learning_rate": 1.94e-06, "loss": 1.2496, "step": 100 }, { "epoch": 0.19, "eval_loss": 1.233077049255371, "eval_runtime": 1969.1907, "eval_samples_per_second": 0.762, "eval_steps_per_second": 0.095, "eval_wer": 176.42124237382143, "step": 100 }, { "epoch": 0.24, "learning_rate": 2.4000000000000003e-06, "loss": 0.9794, "step": 125 }, { "epoch": 0.29, "learning_rate": 2.9e-06, "loss": 0.8642, "step": 150 }, { "epoch": 0.33, "learning_rate": 3.4000000000000005e-06, "loss": 0.8044, "step": 175 }, { "epoch": 0.38, "learning_rate": 3.88e-06, "loss": 0.7389, "step": 200 }, { "epoch": 0.38, "eval_loss": 0.8330782055854797, "eval_runtime": 1755.11, "eval_samples_per_second": 0.855, "eval_steps_per_second": 0.107, "eval_wer": 80.49084858569051, "step": 200 }, { "epoch": 0.43, "learning_rate": 4.38e-06, "loss": 0.6293, "step": 225 }, { "epoch": 0.48, "learning_rate": 4.880000000000001e-06, "loss": 0.5066, "step": 250 }, { "epoch": 0.52, "learning_rate": 5.380000000000001e-06, "loss": 0.3526, "step": 275 }, { "epoch": 0.57, "learning_rate": 5.8800000000000005e-06, "loss": 0.2951, "step": 300 }, { "epoch": 0.57, "eval_loss": 0.4260523319244385, "eval_runtime": 1708.1953, "eval_samples_per_second": 0.878, "eval_steps_per_second": 0.11, "eval_wer": 70.1955074875208, "step": 300 }, { "epoch": 0.62, "learning_rate": 6.380000000000001e-06, "loss": 0.2553, "step": 325 }, { "epoch": 0.67, "learning_rate": 6.88e-06, "loss": 0.2744, "step": 350 }, { "epoch": 0.72, "learning_rate": 7.3800000000000005e-06, "loss": 0.279, "step": 375 }, { "epoch": 0.76, "learning_rate": 7.88e-06, "loss": 0.2703, "step": 400 }, { "epoch": 0.76, "eval_loss": 0.40512609481811523, "eval_runtime": 1849.1681, "eval_samples_per_second": 0.811, "eval_steps_per_second": 0.102, "eval_wer": 101.62922906267333, "step": 400 }, { "epoch": 0.81, "learning_rate": 8.380000000000001e-06, "loss": 0.2514, "step": 425 }, { "epoch": 0.86, "learning_rate": 8.880000000000001e-06, "loss": 0.2826, "step": 450 }, { "epoch": 0.91, "learning_rate": 9.38e-06, "loss": 0.2381, "step": 475 }, { "epoch": 0.95, "learning_rate": 9.88e-06, "loss": 0.2491, "step": 500 }, { "epoch": 0.95, "eval_loss": 0.3922654092311859, "eval_runtime": 1735.7941, "eval_samples_per_second": 0.864, "eval_steps_per_second": 0.108, "eval_wer": 112.23655019412091, "step": 500 }, { "epoch": 1.0, "learning_rate": 9.945714285714286e-06, "loss": 0.2621, "step": 525 }, { "epoch": 1.05, "learning_rate": 9.874285714285715e-06, "loss": 0.1729, "step": 550 }, { "epoch": 1.1, "learning_rate": 9.802857142857144e-06, "loss": 0.1725, "step": 575 }, { "epoch": 1.15, "learning_rate": 9.731428571428573e-06, "loss": 0.17, "step": 600 }, { "epoch": 1.15, "eval_loss": 0.3860141336917877, "eval_runtime": 1739.7902, "eval_samples_per_second": 0.862, "eval_steps_per_second": 0.108, "eval_wer": 107.05074875207987, "step": 600 }, { "epoch": 1.19, "learning_rate": 9.66e-06, "loss": 0.1744, "step": 625 }, { "epoch": 1.24, "learning_rate": 9.58857142857143e-06, "loss": 0.1843, "step": 650 }, { "epoch": 1.29, "learning_rate": 9.517142857142859e-06, "loss": 0.1708, "step": 675 }, { "epoch": 1.34, "learning_rate": 9.445714285714288e-06, "loss": 0.1603, "step": 700 }, { "epoch": 1.34, "eval_loss": 0.38356906175613403, "eval_runtime": 1678.1272, "eval_samples_per_second": 0.894, "eval_steps_per_second": 0.112, "eval_wer": 90.3563505268996, "step": 700 }, { "epoch": 1.38, "learning_rate": 9.374285714285715e-06, "loss": 0.1626, "step": 725 }, { "epoch": 1.43, "learning_rate": 9.302857142857144e-06, "loss": 0.1674, "step": 750 }, { "epoch": 1.48, "learning_rate": 9.231428571428573e-06, "loss": 0.1717, "step": 775 }, { "epoch": 1.53, "learning_rate": 9.16e-06, "loss": 0.1607, "step": 800 }, { "epoch": 1.53, "eval_loss": 0.3789268732070923, "eval_runtime": 1834.2673, "eval_samples_per_second": 0.818, "eval_steps_per_second": 0.102, "eval_wer": 134.98336106489182, "step": 800 }, { "epoch": 1.57, "learning_rate": 9.08857142857143e-06, "loss": 0.1636, "step": 825 }, { "epoch": 1.62, "learning_rate": 9.017142857142858e-06, "loss": 0.1625, "step": 850 }, { "epoch": 1.67, "learning_rate": 8.945714285714286e-06, "loss": 0.1762, "step": 875 }, { "epoch": 1.72, "learning_rate": 8.874285714285715e-06, "loss": 0.154, "step": 900 }, { "epoch": 1.72, "eval_loss": 0.3783314824104309, "eval_runtime": 1787.7259, "eval_samples_per_second": 0.839, "eval_steps_per_second": 0.105, "eval_wer": 99.0501941209096, "step": 900 }, { "epoch": 1.77, "learning_rate": 8.802857142857144e-06, "loss": 0.1703, "step": 925 }, { "epoch": 1.81, "learning_rate": 8.731428571428571e-06, "loss": 0.1593, "step": 950 }, { "epoch": 1.86, "learning_rate": 8.66e-06, "loss": 0.1648, "step": 975 }, { "epoch": 1.91, "learning_rate": 8.588571428571429e-06, "loss": 0.1562, "step": 1000 }, { "epoch": 1.91, "eval_loss": 0.36686423420906067, "eval_runtime": 1815.5283, "eval_samples_per_second": 0.826, "eval_steps_per_second": 0.104, "eval_wer": 98.31530782029951, "step": 1000 }, { "epoch": 1.96, "learning_rate": 8.517142857142858e-06, "loss": 0.1751, "step": 1025 }, { "epoch": 2.0, "learning_rate": 8.445714285714285e-06, "loss": 0.1593, "step": 1050 }, { "epoch": 2.05, "learning_rate": 8.374285714285714e-06, "loss": 0.0693, "step": 1075 }, { "epoch": 2.1, "learning_rate": 8.302857142857143e-06, "loss": 0.0723, "step": 1100 }, { "epoch": 2.1, "eval_loss": 0.3756558299064636, "eval_runtime": 1775.3191, "eval_samples_per_second": 0.845, "eval_steps_per_second": 0.106, "eval_wer": 158.88103161397672, "step": 1100 }, { "epoch": 2.15, "learning_rate": 8.231428571428572e-06, "loss": 0.075, "step": 1125 }, { "epoch": 2.19, "learning_rate": 8.16e-06, "loss": 0.0701, "step": 1150 }, { "epoch": 2.24, "learning_rate": 8.088571428571429e-06, "loss": 0.0732, "step": 1175 }, { "epoch": 2.29, "learning_rate": 8.017142857142858e-06, "loss": 0.0769, "step": 1200 }, { "epoch": 2.29, "eval_loss": 0.3788917660713196, "eval_runtime": 1861.7459, "eval_samples_per_second": 0.806, "eval_steps_per_second": 0.101, "eval_wer": 215.1552967276761, "step": 1200 }, { "epoch": 2.34, "learning_rate": 7.945714285714287e-06, "loss": 0.0747, "step": 1225 }, { "epoch": 2.39, "learning_rate": 7.874285714285716e-06, "loss": 0.0737, "step": 1250 }, { "epoch": 2.43, "learning_rate": 7.802857142857143e-06, "loss": 0.0774, "step": 1275 }, { "epoch": 2.48, "learning_rate": 7.731428571428572e-06, "loss": 0.0814, "step": 1300 }, { "epoch": 2.48, "eval_loss": 0.3779347538948059, "eval_runtime": 1837.4869, "eval_samples_per_second": 0.816, "eval_steps_per_second": 0.102, "eval_wer": 170.4520244037715, "step": 1300 }, { "epoch": 2.53, "learning_rate": 7.660000000000001e-06, "loss": 0.0782, "step": 1325 }, { "epoch": 2.58, "learning_rate": 7.588571428571429e-06, "loss": 0.0732, "step": 1350 }, { "epoch": 2.62, "learning_rate": 7.5171428571428575e-06, "loss": 0.078, "step": 1375 }, { "epoch": 2.67, "learning_rate": 7.445714285714286e-06, "loss": 0.0786, "step": 1400 }, { "epoch": 2.67, "eval_loss": 0.37702885270118713, "eval_runtime": 1750.66, "eval_samples_per_second": 0.857, "eval_steps_per_second": 0.107, "eval_wer": 140.5851358846367, "step": 1400 }, { "epoch": 2.72, "learning_rate": 7.374285714285715e-06, "loss": 0.0774, "step": 1425 }, { "epoch": 2.77, "learning_rate": 7.302857142857144e-06, "loss": 0.0717, "step": 1450 }, { "epoch": 2.81, "learning_rate": 7.231428571428573e-06, "loss": 0.07, "step": 1475 }, { "epoch": 2.86, "learning_rate": 7.16e-06, "loss": 0.0673, "step": 1500 }, { "epoch": 2.86, "eval_loss": 0.37770169973373413, "eval_runtime": 1736.5025, "eval_samples_per_second": 0.864, "eval_steps_per_second": 0.108, "eval_wer": 137.09095951192455, "step": 1500 }, { "epoch": 2.91, "learning_rate": 7.088571428571429e-06, "loss": 0.0753, "step": 1525 }, { "epoch": 2.96, "learning_rate": 7.017142857142858e-06, "loss": 0.0767, "step": 1550 }, { "epoch": 3.01, "learning_rate": 6.945714285714287e-06, "loss": 0.0676, "step": 1575 }, { "epoch": 3.05, "learning_rate": 6.874285714285714e-06, "loss": 0.0339, "step": 1600 }, { "epoch": 3.05, "eval_loss": 0.3892403542995453, "eval_runtime": 1709.1218, "eval_samples_per_second": 0.878, "eval_steps_per_second": 0.11, "eval_wer": 166.51414309484193, "step": 1600 }, { "epoch": 3.1, "learning_rate": 6.8028571428571434e-06, "loss": 0.0353, "step": 1625 }, { "epoch": 3.15, "learning_rate": 6.7314285714285724e-06, "loss": 0.0366, "step": 1650 }, { "epoch": 3.2, "learning_rate": 6.660000000000001e-06, "loss": 0.034, "step": 1675 }, { "epoch": 3.24, "learning_rate": 6.588571428571429e-06, "loss": 0.0324, "step": 1700 }, { "epoch": 3.24, "eval_loss": 0.3962928354740143, "eval_runtime": 1676.3692, "eval_samples_per_second": 0.895, "eval_steps_per_second": 0.112, "eval_wer": 170.93039378813089, "step": 1700 }, { "epoch": 3.29, "learning_rate": 6.517142857142858e-06, "loss": 0.0332, "step": 1725 }, { "epoch": 3.34, "learning_rate": 6.445714285714286e-06, "loss": 0.0342, "step": 1750 }, { "epoch": 3.39, "learning_rate": 6.374285714285715e-06, "loss": 0.0353, "step": 1775 }, { "epoch": 3.44, "learning_rate": 6.302857142857144e-06, "loss": 0.0348, "step": 1800 }, { "epoch": 3.44, "eval_loss": 0.40044304728507996, "eval_runtime": 1648.6224, "eval_samples_per_second": 0.91, "eval_steps_per_second": 0.114, "eval_wer": 163.39434276206325, "step": 1800 }, { "epoch": 3.48, "learning_rate": 6.231428571428571e-06, "loss": 0.0348, "step": 1825 }, { "epoch": 3.53, "learning_rate": 6.16e-06, "loss": 0.0336, "step": 1850 }, { "epoch": 3.58, "learning_rate": 6.088571428571429e-06, "loss": 0.0347, "step": 1875 }, { "epoch": 3.63, "learning_rate": 6.017142857142858e-06, "loss": 0.0345, "step": 1900 }, { "epoch": 3.63, "eval_loss": 0.40161246061325073, "eval_runtime": 1812.1091, "eval_samples_per_second": 0.828, "eval_steps_per_second": 0.104, "eval_wer": 158.56905158069884, "step": 1900 }, { "epoch": 3.67, "learning_rate": 5.945714285714286e-06, "loss": 0.0327, "step": 1925 }, { "epoch": 3.72, "learning_rate": 5.874285714285715e-06, "loss": 0.0363, "step": 1950 }, { "epoch": 3.77, "learning_rate": 5.802857142857144e-06, "loss": 0.0328, "step": 1975 }, { "epoch": 3.82, "learning_rate": 5.731428571428572e-06, "loss": 0.0346, "step": 2000 }, { "epoch": 3.82, "eval_loss": 0.40199828147888184, "eval_runtime": 1770.9946, "eval_samples_per_second": 0.847, "eval_steps_per_second": 0.106, "eval_wer": 176.11619523017194, "step": 2000 }, { "epoch": 3.86, "learning_rate": 5.66e-06, "loss": 0.0363, "step": 2025 }, { "epoch": 3.91, "learning_rate": 5.588571428571429e-06, "loss": 0.0346, "step": 2050 }, { "epoch": 3.96, "learning_rate": 5.517142857142857e-06, "loss": 0.0342, "step": 2075 }, { "epoch": 4.01, "learning_rate": 5.445714285714286e-06, "loss": 0.0317, "step": 2100 }, { "epoch": 4.01, "eval_loss": 0.400082528591156, "eval_runtime": 1696.9765, "eval_samples_per_second": 0.884, "eval_steps_per_second": 0.111, "eval_wer": 134.65058236272878, "step": 2100 }, { "epoch": 4.06, "learning_rate": 5.374285714285715e-06, "loss": 0.018, "step": 2125 }, { "epoch": 4.1, "learning_rate": 5.3028571428571425e-06, "loss": 0.0165, "step": 2150 }, { "epoch": 4.15, "learning_rate": 5.2314285714285716e-06, "loss": 0.0157, "step": 2175 }, { "epoch": 4.2, "learning_rate": 5.1600000000000006e-06, "loss": 0.0173, "step": 2200 }, { "epoch": 4.2, "eval_loss": 0.4140649735927582, "eval_runtime": 1834.1524, "eval_samples_per_second": 0.818, "eval_steps_per_second": 0.102, "eval_wer": 189.33721575152524, "step": 2200 }, { "epoch": 4.25, "learning_rate": 5.08857142857143e-06, "loss": 0.0157, "step": 2225 }, { "epoch": 4.29, "learning_rate": 5.017142857142857e-06, "loss": 0.015, "step": 2250 }, { "epoch": 4.34, "learning_rate": 4.945714285714286e-06, "loss": 0.0146, "step": 2275 }, { "epoch": 4.39, "learning_rate": 4.874285714285715e-06, "loss": 0.0174, "step": 2300 }, { "epoch": 4.39, "eval_loss": 0.41064801812171936, "eval_runtime": 1771.4749, "eval_samples_per_second": 0.847, "eval_steps_per_second": 0.106, "eval_wer": 175.0207986688852, "step": 2300 }, { "epoch": 4.44, "learning_rate": 4.802857142857143e-06, "loss": 0.0164, "step": 2325 }, { "epoch": 4.48, "learning_rate": 4.731428571428572e-06, "loss": 0.0156, "step": 2350 }, { "epoch": 4.53, "learning_rate": 4.66e-06, "loss": 0.0145, "step": 2375 }, { "epoch": 4.58, "learning_rate": 4.588571428571429e-06, "loss": 0.0165, "step": 2400 }, { "epoch": 4.58, "eval_loss": 0.4204367697238922, "eval_runtime": 1768.9183, "eval_samples_per_second": 0.848, "eval_steps_per_second": 0.106, "eval_wer": 179.64503605102607, "step": 2400 }, { "epoch": 4.63, "learning_rate": 4.5171428571428575e-06, "loss": 0.0165, "step": 2425 }, { "epoch": 4.68, "learning_rate": 4.445714285714286e-06, "loss": 0.0159, "step": 2450 }, { "epoch": 4.72, "learning_rate": 4.374285714285715e-06, "loss": 0.0161, "step": 2475 }, { "epoch": 4.77, "learning_rate": 4.302857142857143e-06, "loss": 0.0172, "step": 2500 }, { "epoch": 4.77, "eval_loss": 0.418465793132782, "eval_runtime": 1826.6437, "eval_samples_per_second": 0.821, "eval_steps_per_second": 0.103, "eval_wer": 186.09262340543538, "step": 2500 }, { "epoch": 4.82, "learning_rate": 4.231428571428572e-06, "loss": 0.016, "step": 2525 }, { "epoch": 4.87, "learning_rate": 4.16e-06, "loss": 0.0163, "step": 2550 }, { "epoch": 4.91, "learning_rate": 4.088571428571429e-06, "loss": 0.0163, "step": 2575 }, { "epoch": 4.96, "learning_rate": 4.017142857142857e-06, "loss": 0.0142, "step": 2600 }, { "epoch": 4.96, "eval_loss": 0.41750070452690125, "eval_runtime": 1819.7416, "eval_samples_per_second": 0.824, "eval_steps_per_second": 0.103, "eval_wer": 181.14254021075985, "step": 2600 }, { "epoch": 5.01, "learning_rate": 3.945714285714286e-06, "loss": 0.0156, "step": 2625 }, { "epoch": 5.06, "learning_rate": 3.874285714285715e-06, "loss": 0.0086, "step": 2650 }, { "epoch": 5.1, "learning_rate": 3.802857142857143e-06, "loss": 0.0075, "step": 2675 }, { "epoch": 5.15, "learning_rate": 3.731428571428572e-06, "loss": 0.009, "step": 2700 }, { "epoch": 5.15, "eval_loss": 0.43247538805007935, "eval_runtime": 1622.3539, "eval_samples_per_second": 0.925, "eval_steps_per_second": 0.116, "eval_wer": 161.67498613422075, "step": 2700 }, { "epoch": 5.2, "learning_rate": 3.66e-06, "loss": 0.0074, "step": 2725 }, { "epoch": 5.25, "learning_rate": 3.588571428571429e-06, "loss": 0.0096, "step": 2750 }, { "epoch": 5.3, "learning_rate": 3.5171428571428573e-06, "loss": 0.0098, "step": 2775 }, { "epoch": 5.34, "learning_rate": 3.4457142857142863e-06, "loss": 0.0069, "step": 2800 }, { "epoch": 5.34, "eval_loss": 0.43616336584091187, "eval_runtime": 1644.4427, "eval_samples_per_second": 0.912, "eval_steps_per_second": 0.114, "eval_wer": 161.19661674986133, "step": 2800 }, { "epoch": 5.39, "learning_rate": 3.3742857142857145e-06, "loss": 0.0103, "step": 2825 }, { "epoch": 5.44, "learning_rate": 3.302857142857143e-06, "loss": 0.0083, "step": 2850 }, { "epoch": 5.49, "learning_rate": 3.2314285714285716e-06, "loss": 0.0084, "step": 2875 }, { "epoch": 5.53, "learning_rate": 3.1600000000000002e-06, "loss": 0.0093, "step": 2900 }, { "epoch": 5.53, "eval_loss": 0.43420636653900146, "eval_runtime": 1623.2282, "eval_samples_per_second": 0.924, "eval_steps_per_second": 0.116, "eval_wer": 157.52218524681086, "step": 2900 }, { "epoch": 5.58, "learning_rate": 3.0885714285714284e-06, "loss": 0.009, "step": 2925 }, { "epoch": 5.63, "learning_rate": 3.0171428571428574e-06, "loss": 0.0092, "step": 2950 }, { "epoch": 5.68, "learning_rate": 2.945714285714286e-06, "loss": 0.0086, "step": 2975 }, { "epoch": 5.73, "learning_rate": 2.8742857142857146e-06, "loss": 0.0076, "step": 3000 }, { "epoch": 5.73, "eval_loss": 0.43520399928092957, "eval_runtime": 1605.7035, "eval_samples_per_second": 0.934, "eval_steps_per_second": 0.117, "eval_wer": 154.5410427066001, "step": 3000 }, { "epoch": 5.77, "learning_rate": 2.802857142857143e-06, "loss": 0.0089, "step": 3025 }, { "epoch": 5.82, "learning_rate": 2.7314285714285714e-06, "loss": 0.0074, "step": 3050 }, { "epoch": 5.87, "learning_rate": 2.6600000000000004e-06, "loss": 0.0084, "step": 3075 }, { "epoch": 5.92, "learning_rate": 2.5885714285714285e-06, "loss": 0.0089, "step": 3100 }, { "epoch": 5.92, "eval_loss": 0.4393894076347351, "eval_runtime": 1680.0289, "eval_samples_per_second": 0.893, "eval_steps_per_second": 0.112, "eval_wer": 184.29007210205214, "step": 3100 } ], "logging_steps": 25, "max_steps": 4000, "num_input_tokens_seen": 0, "num_train_epochs": 8, "save_steps": 100, "total_flos": 2.86175712227328e+19, "trial_name": null, "trial_params": null }