| { | |
| "best_metric": 10.128807544441889, | |
| "best_model_checkpoint": "output/checkpoint-4000", | |
| "epoch": 12.0512, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.2000000000000006e-07, | |
| "loss": 5.0258, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.200000000000001e-07, | |
| "loss": 3.7575, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.42e-06, | |
| "loss": 2.3527, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.9200000000000003e-06, | |
| "loss": 1.1445, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.42e-06, | |
| "loss": 0.7122, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.92e-06, | |
| "loss": 0.5839, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.4200000000000007e-06, | |
| "loss": 0.5938, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.920000000000001e-06, | |
| "loss": 0.6077, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.42e-06, | |
| "loss": 0.4504, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.92e-06, | |
| "loss": 0.3166, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.420000000000001e-06, | |
| "loss": 0.2072, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 5.92e-06, | |
| "loss": 0.4215, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 6.42e-06, | |
| "loss": 0.1434, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 6.92e-06, | |
| "loss": 0.0385, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 7.420000000000001e-06, | |
| "loss": 0.0226, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 7.92e-06, | |
| "loss": 0.0368, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 8.42e-06, | |
| "loss": 0.035, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 8.920000000000001e-06, | |
| "loss": 0.0329, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.42e-06, | |
| "loss": 0.0366, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.920000000000002e-06, | |
| "loss": 0.1362, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.953333333333333e-06, | |
| "loss": 0.1407, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 9.89777777777778e-06, | |
| "loss": 0.13, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 9.842222222222223e-06, | |
| "loss": 0.1173, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 9.786666666666667e-06, | |
| "loss": 0.1724, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 9.731111111111113e-06, | |
| "loss": 0.1211, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.675555555555555e-06, | |
| "loss": 0.0179, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.620000000000001e-06, | |
| "loss": 0.0133, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.564444444444445e-06, | |
| "loss": 0.012, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.508888888888889e-06, | |
| "loss": 0.0133, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.453333333333335e-06, | |
| "loss": 0.0139, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.397777777777779e-06, | |
| "loss": 0.0096, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 9.342222222222223e-06, | |
| "loss": 0.0357, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 9.286666666666667e-06, | |
| "loss": 0.063, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 9.231111111111111e-06, | |
| "loss": 0.0552, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 9.175555555555557e-06, | |
| "loss": 0.0544, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 9.12e-06, | |
| "loss": 0.0461, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 9.064444444444447e-06, | |
| "loss": 0.0768, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.008888888888889e-06, | |
| "loss": 0.0263, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 8.953333333333335e-06, | |
| "loss": 0.0068, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 8.897777777777779e-06, | |
| "loss": 0.0068, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "eval_loss": 0.2931811213493347, | |
| "eval_runtime": 1642.4208, | |
| "eval_samples_per_second": 5.524, | |
| "eval_steps_per_second": 0.173, | |
| "eval_wer": 10.348963296421648, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.842222222222223e-06, | |
| "loss": 0.0072, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.786666666666668e-06, | |
| "loss": 0.0058, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.73111111111111e-06, | |
| "loss": 0.007, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.675555555555556e-06, | |
| "loss": 0.0084, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 8.62e-06, | |
| "loss": 0.024, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 8.564444444444445e-06, | |
| "loss": 0.0227, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 8.50888888888889e-06, | |
| "loss": 0.0223, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 8.453333333333334e-06, | |
| "loss": 0.0188, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 8.397777777777778e-06, | |
| "loss": 0.031, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 8.342222222222222e-06, | |
| "loss": 0.0213, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.286666666666668e-06, | |
| "loss": 0.005, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.231111111111112e-06, | |
| "loss": 0.005, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.175555555555556e-06, | |
| "loss": 0.0054, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.120000000000002e-06, | |
| "loss": 0.0039, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.064444444444444e-06, | |
| "loss": 0.0051, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.00888888888889e-06, | |
| "loss": 0.0038, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 7.953333333333334e-06, | |
| "loss": 0.0082, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 7.897777777777778e-06, | |
| "loss": 0.0121, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 7.842222222222224e-06, | |
| "loss": 0.0098, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 7.786666666666666e-06, | |
| "loss": 0.0113, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 7.731111111111112e-06, | |
| "loss": 0.0085, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 7.675555555555556e-06, | |
| "loss": 0.0137, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 7.620000000000001e-06, | |
| "loss": 0.0051, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 7.564444444444446e-06, | |
| "loss": 0.0029, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 7.50888888888889e-06, | |
| "loss": 0.0038, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 7.453333333333334e-06, | |
| "loss": 0.0029, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 7.3977777777777786e-06, | |
| "loss": 0.0039, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 7.342222222222223e-06, | |
| "loss": 0.0036, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 7.2866666666666675e-06, | |
| "loss": 0.0031, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 7.231111111111112e-06, | |
| "loss": 0.0077, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 7.1755555555555556e-06, | |
| "loss": 0.0056, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 7.1200000000000004e-06, | |
| "loss": 0.0055, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 7.0644444444444445e-06, | |
| "loss": 0.0051, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 7.008888888888889e-06, | |
| "loss": 0.0089, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 6.953333333333334e-06, | |
| "loss": 0.0058, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 6.897777777777779e-06, | |
| "loss": 0.0025, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 6.842222222222222e-06, | |
| "loss": 0.0026, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 6.786666666666667e-06, | |
| "loss": 0.0023, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 6.731111111111111e-06, | |
| "loss": 0.0022, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 6.675555555555556e-06, | |
| "loss": 0.0027, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "eval_loss": 0.3447674810886383, | |
| "eval_runtime": 1616.4648, | |
| "eval_samples_per_second": 5.612, | |
| "eval_steps_per_second": 0.176, | |
| "eval_wer": 10.391680084119214, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 6.620000000000001e-06, | |
| "loss": 0.0016, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 6.564444444444446e-06, | |
| "loss": 0.0036, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 6.508888888888889e-06, | |
| "loss": 0.0037, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 6.453333333333334e-06, | |
| "loss": 0.003, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 6.397777777777778e-06, | |
| "loss": 0.0039, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 6.342222222222223e-06, | |
| "loss": 0.0037, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 6.286666666666668e-06, | |
| "loss": 0.0051, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 6.231111111111111e-06, | |
| "loss": 0.0019, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 6.175555555555556e-06, | |
| "loss": 0.0018, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 6.120000000000001e-06, | |
| "loss": 0.0018, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 6.064444444444445e-06, | |
| "loss": 0.0023, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 6.00888888888889e-06, | |
| "loss": 0.0031, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 5.9533333333333345e-06, | |
| "loss": 0.0017, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 5.897777777777778e-06, | |
| "loss": 0.0019, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 5.8422222222222226e-06, | |
| "loss": 0.003, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 5.7866666666666674e-06, | |
| "loss": 0.0028, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 5.7311111111111115e-06, | |
| "loss": 0.0031, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 5.675555555555556e-06, | |
| "loss": 0.0025, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 5.620000000000001e-06, | |
| "loss": 0.0031, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 5.5644444444444444e-06, | |
| "loss": 0.002, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 5.508888888888889e-06, | |
| "loss": 0.0018, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 5.453333333333334e-06, | |
| "loss": 0.0009, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 5.397777777777778e-06, | |
| "loss": 0.0012, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 5.342222222222223e-06, | |
| "loss": 0.0013, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 5.286666666666666e-06, | |
| "loss": 0.0013, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 5.231111111111111e-06, | |
| "loss": 0.0014, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 5.175555555555556e-06, | |
| "loss": 0.0019, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 5.12e-06, | |
| "loss": 0.0017, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 5.064444444444445e-06, | |
| "loss": 0.002, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 5.00888888888889e-06, | |
| "loss": 0.0022, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 4.953333333333334e-06, | |
| "loss": 0.0021, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 4.897777777777778e-06, | |
| "loss": 0.0023, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 4.842222222222223e-06, | |
| "loss": 0.001, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 4.786666666666667e-06, | |
| "loss": 0.0007, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 4.731111111111112e-06, | |
| "loss": 0.0007, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 4.675555555555556e-06, | |
| "loss": 0.0008, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 4.620000000000001e-06, | |
| "loss": 0.0008, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 4.564444444444445e-06, | |
| "loss": 0.001, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 4.50888888888889e-06, | |
| "loss": 0.0009, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 4.453333333333334e-06, | |
| "loss": 0.0014, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "eval_loss": 0.35735899209976196, | |
| "eval_runtime": 1591.4923, | |
| "eval_samples_per_second": 5.7, | |
| "eval_steps_per_second": 0.178, | |
| "eval_wer": 10.217527026582985, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 4.397777777777778e-06, | |
| "loss": 0.0015, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 4.3422222222222225e-06, | |
| "loss": 0.002, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 4.2866666666666666e-06, | |
| "loss": 0.0012, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 4.2311111111111114e-06, | |
| "loss": 0.0018, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 4.175555555555556e-06, | |
| "loss": 0.0013, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 4.12e-06, | |
| "loss": 0.0006, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 4.064444444444444e-06, | |
| "loss": 0.0007, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 4.008888888888889e-06, | |
| "loss": 0.0008, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 3.953333333333333e-06, | |
| "loss": 0.0007, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 3.897777777777778e-06, | |
| "loss": 0.0006, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 3.842222222222223e-06, | |
| "loss": 0.0006, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 3.7866666666666667e-06, | |
| "loss": 0.0009, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 3.7311111111111116e-06, | |
| "loss": 0.0011, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 10.05, | |
| "learning_rate": 3.675555555555556e-06, | |
| "loss": 0.001, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 10.05, | |
| "learning_rate": 3.62e-06, | |
| "loss": 0.0014, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 10.06, | |
| "learning_rate": 3.564444444444445e-06, | |
| "loss": 0.001, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 10.06, | |
| "learning_rate": 3.508888888888889e-06, | |
| "loss": 0.0014, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 3.4533333333333334e-06, | |
| "loss": 0.0007, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 3.3977777777777783e-06, | |
| "loss": 0.0005, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 3.3422222222222224e-06, | |
| "loss": 0.0006, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 3.286666666666667e-06, | |
| "loss": 0.0005, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 3.2311111111111117e-06, | |
| "loss": 0.0006, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 3.1755555555555557e-06, | |
| "loss": 0.0005, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 3.12e-06, | |
| "loss": 0.0006, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "learning_rate": 3.064444444444445e-06, | |
| "loss": 0.0008, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "learning_rate": 3.008888888888889e-06, | |
| "loss": 0.0009, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 11.05, | |
| "learning_rate": 2.9533333333333336e-06, | |
| "loss": 0.0009, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 11.05, | |
| "learning_rate": 2.8977777777777785e-06, | |
| "loss": 0.0008, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 11.06, | |
| "learning_rate": 2.8422222222222225e-06, | |
| "loss": 0.0013, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 2.786666666666667e-06, | |
| "loss": 0.0008, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 2.7311111111111114e-06, | |
| "loss": 0.0005, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 2.675555555555556e-06, | |
| "loss": 0.0005, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 2.6200000000000003e-06, | |
| "loss": 0.0005, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 2.5644444444444444e-06, | |
| "loss": 0.0005, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 2.5088888888888892e-06, | |
| "loss": 0.0005, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 2.4533333333333333e-06, | |
| "loss": 0.0004, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 12.04, | |
| "learning_rate": 2.397777777777778e-06, | |
| "loss": 0.0006, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 12.04, | |
| "learning_rate": 2.342222222222222e-06, | |
| "loss": 0.0008, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 12.05, | |
| "learning_rate": 2.2866666666666667e-06, | |
| "loss": 0.0007, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 12.05, | |
| "learning_rate": 2.2311111111111115e-06, | |
| "loss": 0.0008, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 12.05, | |
| "eval_loss": 0.36404192447662354, | |
| "eval_runtime": 1602.1344, | |
| "eval_samples_per_second": 5.662, | |
| "eval_steps_per_second": 0.177, | |
| "eval_wer": 10.128807544441889, | |
| "step": 4000 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 7.38086019858432e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |