| { | |
| "best_metric": 13.228527607361965, | |
| "best_model_checkpoint": "./ywhisper-small-th/checkpoint-5000", | |
| "epoch": 3.1048, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.2000000000000004e-06, | |
| "loss": 6.2965, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.2e-06, | |
| "loss": 3.963, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.42e-05, | |
| "loss": 2.8448, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.9200000000000003e-05, | |
| "loss": 1.9325, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.4200000000000002e-05, | |
| "loss": 1.4301, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.9199999999999998e-05, | |
| "loss": 1.0656, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.4200000000000005e-05, | |
| "loss": 0.8654, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.9200000000000004e-05, | |
| "loss": 0.7382, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.4200000000000004e-05, | |
| "loss": 0.6436, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.92e-05, | |
| "loss": 0.5363, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.420000000000001e-05, | |
| "loss": 0.4902, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.92e-05, | |
| "loss": 0.4315, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.42e-05, | |
| "loss": 0.4001, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.92e-05, | |
| "loss": 0.3658, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 7.42e-05, | |
| "loss": 0.3635, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 7.920000000000001e-05, | |
| "loss": 0.3169, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.42e-05, | |
| "loss": 0.2862, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.92e-05, | |
| "loss": 0.2887, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.42e-05, | |
| "loss": 0.2692, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.92e-05, | |
| "loss": 0.2394, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.953333333333333e-05, | |
| "loss": 0.2128, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.897777777777778e-05, | |
| "loss": 0.2443, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.842222222222223e-05, | |
| "loss": 0.2289, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.786666666666667e-05, | |
| "loss": 0.1987, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.731111111111112e-05, | |
| "loss": 0.1817, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.675555555555556e-05, | |
| "loss": 0.1755, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.620000000000001e-05, | |
| "loss": 0.209, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.564444444444444e-05, | |
| "loss": 0.1909, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 9.50888888888889e-05, | |
| "loss": 0.1735, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.453333333333335e-05, | |
| "loss": 0.1224, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 9.397777777777778e-05, | |
| "loss": 0.1284, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.342222222222222e-05, | |
| "loss": 0.1251, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.286666666666667e-05, | |
| "loss": 0.1162, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.231111111111112e-05, | |
| "loss": 0.1041, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 9.175555555555556e-05, | |
| "loss": 0.12, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.120000000000001e-05, | |
| "loss": 0.1502, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 9.064444444444446e-05, | |
| "loss": 0.1342, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 9.00888888888889e-05, | |
| "loss": 0.1284, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.953333333333333e-05, | |
| "loss": 0.1151, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.897777777777778e-05, | |
| "loss": 0.0909, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_loss": 0.3373394012451172, | |
| "eval_runtime": 131.4366, | |
| "eval_samples_per_second": 2.435, | |
| "eval_steps_per_second": 0.076, | |
| "eval_wer": 25.57515337423313, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 8.842222222222223e-05, | |
| "loss": 0.0814, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.786666666666667e-05, | |
| "loss": 0.0767, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.731111111111111e-05, | |
| "loss": 0.0713, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.675555555555556e-05, | |
| "loss": 0.0964, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.620000000000001e-05, | |
| "loss": 0.0801, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.564444444444445e-05, | |
| "loss": 0.0758, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 8.50888888888889e-05, | |
| "loss": 0.0778, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.453333333333335e-05, | |
| "loss": 0.0733, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.397777777777778e-05, | |
| "loss": 0.0734, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.342222222222222e-05, | |
| "loss": 0.0715, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.286666666666667e-05, | |
| "loss": 0.0717, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.231111111111112e-05, | |
| "loss": 0.0636, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.175555555555556e-05, | |
| "loss": 0.0596, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.120000000000001e-05, | |
| "loss": 0.0517, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.064444444444445e-05, | |
| "loss": 0.0606, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 8.00888888888889e-05, | |
| "loss": 0.0586, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 7.953333333333333e-05, | |
| "loss": 0.0565, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.897777777777778e-05, | |
| "loss": 0.0572, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.842222222222223e-05, | |
| "loss": 0.0554, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 7.786666666666667e-05, | |
| "loss": 0.0628, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 7.731111111111111e-05, | |
| "loss": 0.071, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 7.675555555555556e-05, | |
| "loss": 0.0964, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 7.620000000000001e-05, | |
| "loss": 0.0815, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 7.564444444444445e-05, | |
| "loss": 0.0901, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 7.50888888888889e-05, | |
| "loss": 0.0856, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 7.453333333333333e-05, | |
| "loss": 0.0649, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 7.397777777777778e-05, | |
| "loss": 0.0626, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 7.342222222222222e-05, | |
| "loss": 0.0713, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 7.286666666666667e-05, | |
| "loss": 0.0578, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 7.231111111111112e-05, | |
| "loss": 0.0589, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 7.175555555555556e-05, | |
| "loss": 0.0622, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 7.12e-05, | |
| "loss": 0.05, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 7.064444444444445e-05, | |
| "loss": 0.062, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 7.00888888888889e-05, | |
| "loss": 0.0601, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 6.953333333333333e-05, | |
| "loss": 0.0758, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 6.897777777777778e-05, | |
| "loss": 0.0534, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 6.842222222222222e-05, | |
| "loss": 0.0512, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 6.786666666666667e-05, | |
| "loss": 0.061, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 6.731111111111111e-05, | |
| "loss": 0.0528, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 6.675555555555556e-05, | |
| "loss": 0.0426, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "eval_loss": 0.25401371717453003, | |
| "eval_runtime": 127.2978, | |
| "eval_samples_per_second": 2.514, | |
| "eval_steps_per_second": 0.079, | |
| "eval_wer": 20.9739263803681, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 6.620000000000001e-05, | |
| "loss": 0.0403, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 6.564444444444445e-05, | |
| "loss": 0.0666, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 6.508888888888888e-05, | |
| "loss": 0.0561, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 6.453333333333333e-05, | |
| "loss": 0.0446, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 6.397777777777778e-05, | |
| "loss": 0.0441, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 6.342222222222222e-05, | |
| "loss": 0.0562, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 6.286666666666667e-05, | |
| "loss": 0.0657, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 6.231111111111111e-05, | |
| "loss": 0.0691, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 6.175555555555556e-05, | |
| "loss": 0.0518, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 6.12e-05, | |
| "loss": 0.0347, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 6.064444444444445e-05, | |
| "loss": 0.031, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 6.00888888888889e-05, | |
| "loss": 0.0361, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 5.953333333333334e-05, | |
| "loss": 0.0297, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 5.897777777777778e-05, | |
| "loss": 0.0337, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 5.842222222222222e-05, | |
| "loss": 0.0447, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 5.7866666666666666e-05, | |
| "loss": 0.0594, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 5.7311111111111117e-05, | |
| "loss": 0.0547, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 5.675555555555556e-05, | |
| "loss": 0.0575, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 5.620000000000001e-05, | |
| "loss": 0.0373, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 5.564444444444444e-05, | |
| "loss": 0.0275, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 5.508888888888889e-05, | |
| "loss": 0.0258, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 5.4533333333333335e-05, | |
| "loss": 0.0214, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 5.3977777777777786e-05, | |
| "loss": 0.0224, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 5.342222222222223e-05, | |
| "loss": 0.0314, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 5.2866666666666666e-05, | |
| "loss": 0.0258, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 5.231111111111111e-05, | |
| "loss": 0.0237, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 5.1755555555555554e-05, | |
| "loss": 0.0249, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 5.1200000000000004e-05, | |
| "loss": 0.0262, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 5.064444444444445e-05, | |
| "loss": 0.0272, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 5.00888888888889e-05, | |
| "loss": 0.0251, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 4.9533333333333336e-05, | |
| "loss": 0.0254, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 4.897777777777778e-05, | |
| "loss": 0.0257, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 4.842222222222222e-05, | |
| "loss": 0.0212, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 4.7866666666666674e-05, | |
| "loss": 0.0201, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 4.731111111111111e-05, | |
| "loss": 0.0232, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 4.675555555555556e-05, | |
| "loss": 0.0223, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 4.6200000000000005e-05, | |
| "loss": 0.0218, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 4.564444444444444e-05, | |
| "loss": 0.0229, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 4.508888888888889e-05, | |
| "loss": 0.0217, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.4533333333333336e-05, | |
| "loss": 0.0267, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.22101850807666779, | |
| "eval_runtime": 129.0103, | |
| "eval_samples_per_second": 2.48, | |
| "eval_steps_per_second": 0.078, | |
| "eval_wer": 17.4079754601227, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 4.397777777777778e-05, | |
| "loss": 0.0406, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 4.3422222222222224e-05, | |
| "loss": 0.0358, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 4.286666666666667e-05, | |
| "loss": 0.036, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 4.231111111111111e-05, | |
| "loss": 0.0479, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 4.175555555555556e-05, | |
| "loss": 0.0395, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 4.12e-05, | |
| "loss": 0.0316, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 4.064444444444445e-05, | |
| "loss": 0.0261, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 4.008888888888889e-05, | |
| "loss": 0.0312, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 3.9533333333333337e-05, | |
| "loss": 0.0231, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 3.897777777777778e-05, | |
| "loss": 0.0329, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 3.8422222222222224e-05, | |
| "loss": 0.0273, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 3.786666666666667e-05, | |
| "loss": 0.0224, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 3.731111111111111e-05, | |
| "loss": 0.027, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 3.675555555555556e-05, | |
| "loss": 0.0333, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 3.62e-05, | |
| "loss": 0.0358, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 3.564444444444445e-05, | |
| "loss": 0.026, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 3.5088888888888886e-05, | |
| "loss": 0.0228, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 3.453333333333334e-05, | |
| "loss": 0.0294, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 3.397777777777778e-05, | |
| "loss": 0.023, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 3.3422222222222224e-05, | |
| "loss": 0.0195, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 3.286666666666667e-05, | |
| "loss": 0.0231, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 3.231111111111111e-05, | |
| "loss": 0.0322, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 3.1755555555555556e-05, | |
| "loss": 0.0282, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 3.12e-05, | |
| "loss": 0.0169, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 3.064444444444445e-05, | |
| "loss": 0.0251, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 3.008888888888889e-05, | |
| "loss": 0.0282, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 2.9533333333333334e-05, | |
| "loss": 0.0329, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 2.897777777777778e-05, | |
| "loss": 0.0363, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 2.842222222222222e-05, | |
| "loss": 0.0193, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 2.786666666666667e-05, | |
| "loss": 0.0174, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 2.7311111111111116e-05, | |
| "loss": 0.0139, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 2.6755555555555556e-05, | |
| "loss": 0.0158, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 2.6200000000000003e-05, | |
| "loss": 0.0147, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 2.5644444444444444e-05, | |
| "loss": 0.0165, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 2.508888888888889e-05, | |
| "loss": 0.0294, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 2.4533333333333334e-05, | |
| "loss": 0.027, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 2.3977777777777778e-05, | |
| "loss": 0.0278, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 2.3422222222222222e-05, | |
| "loss": 0.0277, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 2.2866666666666666e-05, | |
| "loss": 0.0158, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 2.2311111111111113e-05, | |
| "loss": 0.0145, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "eval_loss": 0.21337780356407166, | |
| "eval_runtime": 143.6043, | |
| "eval_samples_per_second": 2.228, | |
| "eval_steps_per_second": 0.07, | |
| "eval_wer": 15.567484662576685, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 2.1755555555555557e-05, | |
| "loss": 0.0128, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 2.12e-05, | |
| "loss": 0.0094, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 2.0644444444444447e-05, | |
| "loss": 0.0137, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 2.008888888888889e-05, | |
| "loss": 0.0135, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.9533333333333335e-05, | |
| "loss": 0.0109, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.897777777777778e-05, | |
| "loss": 0.0112, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.8422222222222222e-05, | |
| "loss": 0.0105, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.7866666666666666e-05, | |
| "loss": 0.0133, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.731111111111111e-05, | |
| "loss": 0.013, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.6755555555555557e-05, | |
| "loss": 0.0121, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.62e-05, | |
| "loss": 0.0127, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.5644444444444444e-05, | |
| "loss": 0.0129, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.5088888888888888e-05, | |
| "loss": 0.0087, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.4533333333333335e-05, | |
| "loss": 0.0108, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.3977777777777779e-05, | |
| "loss": 0.0122, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.3422222222222223e-05, | |
| "loss": 0.0108, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.2866666666666668e-05, | |
| "loss": 0.0106, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.2311111111111112e-05, | |
| "loss": 0.0111, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.1755555555555556e-05, | |
| "loss": 0.0118, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 1.1200000000000001e-05, | |
| "loss": 0.0147, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 1.0644444444444445e-05, | |
| "loss": 0.0249, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 1.0088888888888889e-05, | |
| "loss": 0.0173, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 9.533333333333334e-06, | |
| "loss": 0.0202, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.977777777777778e-06, | |
| "loss": 0.0215, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.422222222222223e-06, | |
| "loss": 0.0136, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 7.866666666666667e-06, | |
| "loss": 0.0142, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 7.311111111111112e-06, | |
| "loss": 0.015, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 6.755555555555555e-06, | |
| "loss": 0.014, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 6.2e-06, | |
| "loss": 0.0108, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 5.6444444444444445e-06, | |
| "loss": 0.0152, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 5.088888888888889e-06, | |
| "loss": 0.0131, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 4.533333333333334e-06, | |
| "loss": 0.0104, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 3.977777777777778e-06, | |
| "loss": 0.0132, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 3.4222222222222224e-06, | |
| "loss": 0.0206, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.8666666666666666e-06, | |
| "loss": 0.0141, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.311111111111111e-06, | |
| "loss": 0.0117, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 1.7555555555555558e-06, | |
| "loss": 0.0135, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 1.2000000000000002e-06, | |
| "loss": 0.0148, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 6.444444444444444e-07, | |
| "loss": 0.0108, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 8.88888888888889e-08, | |
| "loss": 0.0099, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "eval_loss": 0.18412813544273376, | |
| "eval_runtime": 137.2691, | |
| "eval_samples_per_second": 2.331, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 13.228527607361965, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "step": 5000, | |
| "total_flos": 9.23429993213952e+19, | |
| "train_loss": 0.16319971153736115, | |
| "train_runtime": 45749.6514, | |
| "train_samples_per_second": 6.995, | |
| "train_steps_per_second": 0.109 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 9.23429993213952e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |