| { | |
| "best_metric": 6.248270773771097, | |
| "best_model_checkpoint": "./whisper-large-id/checkpoint-6000", | |
| "epoch": 4.9776007964161275, | |
| "global_step": 10000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.2e-08, | |
| "loss": 0.6547, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9e-08, | |
| "loss": 0.7089, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.4e-07, | |
| "loss": 0.5781, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.8999999999999998e-07, | |
| "loss": 0.6924, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.4e-07, | |
| "loss": 0.4614, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9e-07, | |
| "loss": 0.4274, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 3.36e-07, | |
| "loss": 0.4134, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 3.86e-07, | |
| "loss": 0.402, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.36e-07, | |
| "loss": 0.3667, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.86e-07, | |
| "loss": 0.3072, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 5.36e-07, | |
| "loss": 0.2807, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 5.86e-07, | |
| "loss": 0.2892, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 6.36e-07, | |
| "loss": 0.2782, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 6.86e-07, | |
| "loss": 0.2528, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 7.359999999999999e-07, | |
| "loss": 0.2585, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 7.86e-07, | |
| "loss": 0.2218, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 8.359999999999999e-07, | |
| "loss": 0.2123, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 8.86e-07, | |
| "loss": 0.2104, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 9.36e-07, | |
| "loss": 0.1794, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 9.86e-07, | |
| "loss": 0.2228, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 9.981052631578947e-07, | |
| "loss": 0.1893, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 9.954736842105263e-07, | |
| "loss": 0.1997, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 9.928421052631579e-07, | |
| "loss": 0.1646, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 9.902105263157895e-07, | |
| "loss": 0.1854, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 9.87578947368421e-07, | |
| "loss": 0.1727, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 9.849473684210526e-07, | |
| "loss": 0.1898, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 9.823157894736842e-07, | |
| "loss": 0.1806, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 9.796842105263158e-07, | |
| "loss": 0.1486, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 9.770526315789474e-07, | |
| "loss": 0.1175, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 9.74421052631579e-07, | |
| "loss": 0.1601, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 9.717894736842105e-07, | |
| "loss": 0.1683, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 9.691578947368421e-07, | |
| "loss": 0.1464, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 9.665263157894737e-07, | |
| "loss": 0.1361, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 9.638947368421053e-07, | |
| "loss": 0.146, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 9.612631578947369e-07, | |
| "loss": 0.1278, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 9.586315789473682e-07, | |
| "loss": 0.1822, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 9.559999999999998e-07, | |
| "loss": 0.1407, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 9.533684210526315e-07, | |
| "loss": 0.1352, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 9.507368421052631e-07, | |
| "loss": 0.1229, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 9.481052631578948e-07, | |
| "loss": 0.1516, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_loss": 0.17297802865505219, | |
| "eval_runtime": 4131.6851, | |
| "eval_samples_per_second": 0.876, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.56644839988933, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 9.454736842105264e-07, | |
| "loss": 0.1443, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.428421052631578e-07, | |
| "loss": 0.1222, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.402105263157894e-07, | |
| "loss": 0.1649, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.37578947368421e-07, | |
| "loss": 0.1331, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 9.349473684210526e-07, | |
| "loss": 0.1113, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 9.323157894736842e-07, | |
| "loss": 0.1419, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 9.296842105263157e-07, | |
| "loss": 0.1105, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.270526315789473e-07, | |
| "loss": 0.1241, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.244210526315789e-07, | |
| "loss": 0.1182, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 9.217894736842105e-07, | |
| "loss": 0.1398, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 9.19157894736842e-07, | |
| "loss": 0.1304, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 9.165263157894736e-07, | |
| "loss": 0.1013, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 9.138947368421052e-07, | |
| "loss": 0.1168, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.112631578947368e-07, | |
| "loss": 0.1117, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.086315789473684e-07, | |
| "loss": 0.1101, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.06e-07, | |
| "loss": 0.1053, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 9.033684210526316e-07, | |
| "loss": 0.1093, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.007368421052632e-07, | |
| "loss": 0.1187, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.981052631578947e-07, | |
| "loss": 0.1555, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 8.954736842105263e-07, | |
| "loss": 0.1088, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 8.928421052631578e-07, | |
| "loss": 0.1395, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 8.902105263157894e-07, | |
| "loss": 0.1423, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 8.87578947368421e-07, | |
| "loss": 0.1037, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 8.849473684210525e-07, | |
| "loss": 0.1251, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 8.823157894736841e-07, | |
| "loss": 0.0935, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.796842105263157e-07, | |
| "loss": 0.1003, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 8.770526315789474e-07, | |
| "loss": 0.1063, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 8.74421052631579e-07, | |
| "loss": 0.1027, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 8.717894736842106e-07, | |
| "loss": 0.1056, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.691578947368421e-07, | |
| "loss": 0.1493, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.665263157894736e-07, | |
| "loss": 0.1028, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 8.638947368421052e-07, | |
| "loss": 0.1118, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.612631578947368e-07, | |
| "loss": 0.095, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 8.586315789473684e-07, | |
| "loss": 0.1203, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 8.559999999999999e-07, | |
| "loss": 0.1213, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 8.533684210526315e-07, | |
| "loss": 0.117, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.507368421052631e-07, | |
| "loss": 0.1088, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.481052631578947e-07, | |
| "loss": 0.1008, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 8.454736842105263e-07, | |
| "loss": 0.0982, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.42842105263158e-07, | |
| "loss": 0.1081, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.1638014167547226, | |
| "eval_runtime": 4131.4754, | |
| "eval_samples_per_second": 0.876, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.368163792308401, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 8.402105263157895e-07, | |
| "loss": 0.0708, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 8.37578947368421e-07, | |
| "loss": 0.0722, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 8.349473684210526e-07, | |
| "loss": 0.0681, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.323157894736842e-07, | |
| "loss": 0.059, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 8.296842105263158e-07, | |
| "loss": 0.063, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.270526315789473e-07, | |
| "loss": 0.0734, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 8.244210526315789e-07, | |
| "loss": 0.077, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 8.217894736842105e-07, | |
| "loss": 0.0845, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 8.191578947368421e-07, | |
| "loss": 0.0599, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 8.165263157894737e-07, | |
| "loss": 0.0734, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 8.138947368421051e-07, | |
| "loss": 0.0865, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 8.112631578947367e-07, | |
| "loss": 0.073, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.086315789473683e-07, | |
| "loss": 0.0708, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.06e-07, | |
| "loss": 0.0855, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.033684210526316e-07, | |
| "loss": 0.0758, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.007368421052632e-07, | |
| "loss": 0.1104, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 7.981052631578947e-07, | |
| "loss": 0.0801, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 7.954736842105263e-07, | |
| "loss": 0.1124, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 7.928421052631579e-07, | |
| "loss": 0.0607, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 7.902105263157895e-07, | |
| "loss": 0.0768, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 7.87578947368421e-07, | |
| "loss": 0.0801, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 7.849473684210525e-07, | |
| "loss": 0.0771, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 7.823157894736841e-07, | |
| "loss": 0.0752, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 7.796842105263157e-07, | |
| "loss": 0.0839, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 7.770526315789473e-07, | |
| "loss": 0.0844, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 7.744210526315789e-07, | |
| "loss": 0.0615, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 7.717894736842106e-07, | |
| "loss": 0.0885, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 7.691578947368421e-07, | |
| "loss": 0.0862, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 7.665263157894737e-07, | |
| "loss": 0.0638, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 7.638947368421053e-07, | |
| "loss": 0.0857, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 7.612631578947368e-07, | |
| "loss": 0.068, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 7.586315789473684e-07, | |
| "loss": 0.0747, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 7.559999999999999e-07, | |
| "loss": 0.0643, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 7.533684210526315e-07, | |
| "loss": 0.0788, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 7.507368421052631e-07, | |
| "loss": 0.0694, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 7.481052631578947e-07, | |
| "loss": 0.0664, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 7.454736842105263e-07, | |
| "loss": 0.0676, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 7.428421052631578e-07, | |
| "loss": 0.1094, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 7.402105263157894e-07, | |
| "loss": 0.0668, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 7.37578947368421e-07, | |
| "loss": 0.0715, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "eval_loss": 0.18033772706985474, | |
| "eval_runtime": 4134.4448, | |
| "eval_samples_per_second": 0.875, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.271327123489809, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 7.349473684210527e-07, | |
| "loss": 0.0673, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 7.323157894736842e-07, | |
| "loss": 0.0754, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 7.296842105263158e-07, | |
| "loss": 0.0567, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 7.270526315789473e-07, | |
| "loss": 0.0855, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 7.244210526315789e-07, | |
| "loss": 0.0854, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 7.217894736842105e-07, | |
| "loss": 0.0953, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 7.191578947368421e-07, | |
| "loss": 0.0657, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 7.165263157894737e-07, | |
| "loss": 0.0811, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 7.138947368421053e-07, | |
| "loss": 0.0604, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 7.112631578947368e-07, | |
| "loss": 0.0687, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 7.086315789473683e-07, | |
| "loss": 0.0723, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 7.059999999999999e-07, | |
| "loss": 0.0773, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 7.033684210526315e-07, | |
| "loss": 0.0748, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 7.007368421052631e-07, | |
| "loss": 0.079, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 6.981052631578947e-07, | |
| "loss": 0.0731, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 6.954736842105263e-07, | |
| "loss": 0.0996, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 6.928421052631579e-07, | |
| "loss": 0.0526, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 6.902105263157895e-07, | |
| "loss": 0.0781, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 6.875789473684211e-07, | |
| "loss": 0.0644, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 6.849473684210527e-07, | |
| "loss": 0.0686, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 6.823157894736841e-07, | |
| "loss": 0.0869, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 6.796842105263157e-07, | |
| "loss": 0.0687, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 6.770526315789473e-07, | |
| "loss": 0.0708, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 6.744210526315789e-07, | |
| "loss": 0.069, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 6.717894736842105e-07, | |
| "loss": 0.055, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 6.69157894736842e-07, | |
| "loss": 0.0698, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 6.665263157894736e-07, | |
| "loss": 0.0717, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 6.638947368421053e-07, | |
| "loss": 0.0729, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 6.612631578947369e-07, | |
| "loss": 0.0713, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 6.586315789473685e-07, | |
| "loss": 0.0539, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 6.56e-07, | |
| "loss": 0.0564, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 6.533684210526315e-07, | |
| "loss": 0.0627, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 6.507368421052631e-07, | |
| "loss": 0.0679, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 6.481052631578947e-07, | |
| "loss": 0.0778, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 6.454736842105263e-07, | |
| "loss": 0.0634, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 6.428421052631579e-07, | |
| "loss": 0.0773, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 6.402105263157894e-07, | |
| "loss": 0.0741, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 6.37578947368421e-07, | |
| "loss": 0.0506, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 6.349473684210526e-07, | |
| "loss": 0.0828, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 6.323157894736842e-07, | |
| "loss": 0.1009, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_loss": 0.17957788705825806, | |
| "eval_runtime": 4136.4423, | |
| "eval_samples_per_second": 0.875, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.266715853546066, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 6.296842105263157e-07, | |
| "loss": 0.0723, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 6.270526315789473e-07, | |
| "loss": 0.0392, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 6.244210526315789e-07, | |
| "loss": 0.0443, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 6.217894736842105e-07, | |
| "loss": 0.0443, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 6.191578947368421e-07, | |
| "loss": 0.0421, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 6.165263157894737e-07, | |
| "loss": 0.0372, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 6.138947368421053e-07, | |
| "loss": 0.0444, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 6.112631578947368e-07, | |
| "loss": 0.0596, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 6.086315789473684e-07, | |
| "loss": 0.0494, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.06e-07, | |
| "loss": 0.0605, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 6.034736842105263e-07, | |
| "loss": 0.0517, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 6.008421052631578e-07, | |
| "loss": 0.0391, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 5.982105263157895e-07, | |
| "loss": 0.0679, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 5.95578947368421e-07, | |
| "loss": 0.035, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 5.929473684210526e-07, | |
| "loss": 0.0375, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 5.903157894736842e-07, | |
| "loss": 0.0433, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 5.876842105263158e-07, | |
| "loss": 0.0494, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 5.850526315789474e-07, | |
| "loss": 0.0582, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 5.82421052631579e-07, | |
| "loss": 0.0437, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 5.797894736842105e-07, | |
| "loss": 0.0403, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 5.771578947368421e-07, | |
| "loss": 0.0522, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 5.745263157894736e-07, | |
| "loss": 0.0473, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 5.718947368421052e-07, | |
| "loss": 0.0522, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 5.692631578947368e-07, | |
| "loss": 0.0508, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 5.666315789473683e-07, | |
| "loss": 0.0481, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 5.639999999999999e-07, | |
| "loss": 0.0414, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 5.614736842105263e-07, | |
| "loss": 0.0664, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 5.588421052631579e-07, | |
| "loss": 0.0454, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 5.562105263157895e-07, | |
| "loss": 0.0427, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 5.535789473684211e-07, | |
| "loss": 0.0457, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 5.509473684210526e-07, | |
| "loss": 0.0507, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 5.483157894736842e-07, | |
| "loss": 0.0325, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 5.456842105263157e-07, | |
| "loss": 0.0386, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 5.430526315789473e-07, | |
| "loss": 0.0404, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 5.404210526315789e-07, | |
| "loss": 0.0486, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 5.377894736842104e-07, | |
| "loss": 0.0481, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 5.35157894736842e-07, | |
| "loss": 0.0577, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 5.325263157894737e-07, | |
| "loss": 0.0466, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 5.298947368421053e-07, | |
| "loss": 0.0528, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 5.272631578947369e-07, | |
| "loss": 0.0387, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_loss": 0.20537233352661133, | |
| "eval_runtime": 4140.2261, | |
| "eval_samples_per_second": 0.874, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.492668080789449, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 5.246315789473685e-07, | |
| "loss": 0.0434, | |
| "step": 5025 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 5.22e-07, | |
| "loss": 0.0484, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 5.193684210526315e-07, | |
| "loss": 0.0466, | |
| "step": 5075 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 5.167368421052631e-07, | |
| "loss": 0.0416, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 5.141052631578947e-07, | |
| "loss": 0.0561, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 5.114736842105263e-07, | |
| "loss": 0.0371, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.088421052631578e-07, | |
| "loss": 0.0467, | |
| "step": 5175 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.062105263157894e-07, | |
| "loss": 0.0363, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.03578947368421e-07, | |
| "loss": 0.0802, | |
| "step": 5225 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.009473684210526e-07, | |
| "loss": 0.0345, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 4.983157894736842e-07, | |
| "loss": 0.038, | |
| "step": 5275 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 4.956842105263158e-07, | |
| "loss": 0.0364, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 4.930526315789473e-07, | |
| "loss": 0.0658, | |
| "step": 5325 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 4.904210526315789e-07, | |
| "loss": 0.0605, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 4.877894736842105e-07, | |
| "loss": 0.0349, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 4.851578947368421e-07, | |
| "loss": 0.0599, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 4.825263157894737e-07, | |
| "loss": 0.0441, | |
| "step": 5425 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.798947368421052e-07, | |
| "loss": 0.0518, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.772631578947368e-07, | |
| "loss": 0.0501, | |
| "step": 5475 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.7463157894736836e-07, | |
| "loss": 0.0431, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.7199999999999994e-07, | |
| "loss": 0.0313, | |
| "step": 5525 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.6936842105263157e-07, | |
| "loss": 0.044, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.6673684210526316e-07, | |
| "loss": 0.04, | |
| "step": 5575 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.6410526315789474e-07, | |
| "loss": 0.0405, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.614736842105263e-07, | |
| "loss": 0.0485, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.5884210526315785e-07, | |
| "loss": 0.068, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.5621052631578943e-07, | |
| "loss": 0.0443, | |
| "step": 5675 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.53578947368421e-07, | |
| "loss": 0.0361, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.5094736842105265e-07, | |
| "loss": 0.042, | |
| "step": 5725 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 4.4831578947368423e-07, | |
| "loss": 0.0464, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 4.4568421052631576e-07, | |
| "loss": 0.038, | |
| "step": 5775 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 4.4305263157894734e-07, | |
| "loss": 0.0437, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 4.404210526315789e-07, | |
| "loss": 0.0509, | |
| "step": 5825 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 4.377894736842105e-07, | |
| "loss": 0.0453, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 4.3515789473684214e-07, | |
| "loss": 0.0496, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 4.3252631578947367e-07, | |
| "loss": 0.0382, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 4.2989473684210525e-07, | |
| "loss": 0.077, | |
| "step": 5925 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 4.2726315789473683e-07, | |
| "loss": 0.0663, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 4.246315789473684e-07, | |
| "loss": 0.04, | |
| "step": 5975 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 4.2199999999999994e-07, | |
| "loss": 0.0494, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "eval_loss": 0.20344047248363495, | |
| "eval_runtime": 4135.2447, | |
| "eval_samples_per_second": 0.875, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.248270773771097, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 4.193684210526315e-07, | |
| "loss": 0.0642, | |
| "step": 6025 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.1673684210526316e-07, | |
| "loss": 0.0283, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.1410526315789474e-07, | |
| "loss": 0.0368, | |
| "step": 6075 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 4.114736842105263e-07, | |
| "loss": 0.0372, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 4.088421052631579e-07, | |
| "loss": 0.0307, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 4.0621052631578943e-07, | |
| "loss": 0.0459, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 4.03578947368421e-07, | |
| "loss": 0.0327, | |
| "step": 6175 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 4.009473684210526e-07, | |
| "loss": 0.0312, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 3.9831578947368423e-07, | |
| "loss": 0.0338, | |
| "step": 6225 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.956842105263158e-07, | |
| "loss": 0.028, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 3.9305263157894734e-07, | |
| "loss": 0.0315, | |
| "step": 6275 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 3.904210526315789e-07, | |
| "loss": 0.0311, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 3.877894736842105e-07, | |
| "loss": 0.0372, | |
| "step": 6325 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 3.851578947368421e-07, | |
| "loss": 0.0255, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 3.825263157894736e-07, | |
| "loss": 0.0338, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 3.7989473684210525e-07, | |
| "loss": 0.0491, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 3.7726315789473683e-07, | |
| "loss": 0.0279, | |
| "step": 6425 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 3.746315789473684e-07, | |
| "loss": 0.031, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 3.72e-07, | |
| "loss": 0.044, | |
| "step": 6475 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 3.693684210526316e-07, | |
| "loss": 0.0345, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 3.667368421052631e-07, | |
| "loss": 0.0228, | |
| "step": 6525 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 3.641052631578947e-07, | |
| "loss": 0.0315, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 3.614736842105263e-07, | |
| "loss": 0.0377, | |
| "step": 6575 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 3.588421052631579e-07, | |
| "loss": 0.0339, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 3.562105263157895e-07, | |
| "loss": 0.0345, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 3.53578947368421e-07, | |
| "loss": 0.0312, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 3.509473684210526e-07, | |
| "loss": 0.0177, | |
| "step": 6675 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 3.483157894736842e-07, | |
| "loss": 0.041, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 3.456842105263158e-07, | |
| "loss": 0.0275, | |
| "step": 6725 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 3.430526315789474e-07, | |
| "loss": 0.0595, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 3.404210526315789e-07, | |
| "loss": 0.0376, | |
| "step": 6775 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 3.377894736842105e-07, | |
| "loss": 0.0303, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 3.351578947368421e-07, | |
| "loss": 0.0303, | |
| "step": 6825 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 3.3252631578947367e-07, | |
| "loss": 0.0286, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 3.298947368421052e-07, | |
| "loss": 0.0297, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 3.2726315789473683e-07, | |
| "loss": 0.0305, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 3.246315789473684e-07, | |
| "loss": 0.031, | |
| "step": 6925 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 3.22e-07, | |
| "loss": 0.0268, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 3.193684210526316e-07, | |
| "loss": 0.0309, | |
| "step": 6975 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 3.1673684210526316e-07, | |
| "loss": 0.0259, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "eval_loss": 0.22257116436958313, | |
| "eval_runtime": 4139.2156, | |
| "eval_samples_per_second": 0.874, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.3497187125334325, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 3.141052631578947e-07, | |
| "loss": 0.0286, | |
| "step": 7025 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 3.1147368421052627e-07, | |
| "loss": 0.0714, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 3.088421052631579e-07, | |
| "loss": 0.0465, | |
| "step": 7075 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 3.062105263157895e-07, | |
| "loss": 0.0244, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 3.0357894736842107e-07, | |
| "loss": 0.0277, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 3.009473684210526e-07, | |
| "loss": 0.037, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 2.983157894736842e-07, | |
| "loss": 0.0284, | |
| "step": 7175 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 2.9568421052631576e-07, | |
| "loss": 0.03, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 2.9305263157894735e-07, | |
| "loss": 0.0235, | |
| "step": 7225 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 2.90421052631579e-07, | |
| "loss": 0.0426, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 2.877894736842105e-07, | |
| "loss": 0.0296, | |
| "step": 7275 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 2.851578947368421e-07, | |
| "loss": 0.0268, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 2.8252631578947367e-07, | |
| "loss": 0.0336, | |
| "step": 7325 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 2.7989473684210525e-07, | |
| "loss": 0.0317, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 2.772631578947368e-07, | |
| "loss": 0.0279, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 2.746315789473684e-07, | |
| "loss": 0.0304, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 2.72e-07, | |
| "loss": 0.0296, | |
| "step": 7425 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.693684210526316e-07, | |
| "loss": 0.0249, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.6673684210526316e-07, | |
| "loss": 0.0313, | |
| "step": 7475 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.6410526315789475e-07, | |
| "loss": 0.0276, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.614736842105263e-07, | |
| "loss": 0.0318, | |
| "step": 7525 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 2.5884210526315786e-07, | |
| "loss": 0.025, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 2.562105263157895e-07, | |
| "loss": 0.0324, | |
| "step": 7575 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 2.5357894736842107e-07, | |
| "loss": 0.032, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 2.5094736842105266e-07, | |
| "loss": 0.0291, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 2.483157894736842e-07, | |
| "loss": 0.0314, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 2.4568421052631577e-07, | |
| "loss": 0.0286, | |
| "step": 7675 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 2.4305263157894735e-07, | |
| "loss": 0.0533, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 2.4042105263157893e-07, | |
| "loss": 0.0289, | |
| "step": 7725 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 2.377894736842105e-07, | |
| "loss": 0.0337, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 2.351578947368421e-07, | |
| "loss": 0.0462, | |
| "step": 7775 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 2.3252631578947368e-07, | |
| "loss": 0.0285, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 2.2989473684210526e-07, | |
| "loss": 0.0286, | |
| "step": 7825 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.272631578947368e-07, | |
| "loss": 0.0276, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.2463157894736842e-07, | |
| "loss": 0.0359, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 2.22e-07, | |
| "loss": 0.0273, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 2.1936842105263156e-07, | |
| "loss": 0.024, | |
| "step": 7925 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 2.1673684210526314e-07, | |
| "loss": 0.0292, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 2.1410526315789472e-07, | |
| "loss": 0.0364, | |
| "step": 7975 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.114736842105263e-07, | |
| "loss": 0.0265, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "eval_loss": 0.22743414342403412, | |
| "eval_runtime": 4139.2969, | |
| "eval_samples_per_second": 0.874, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.400442681914599, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 2.0884210526315789e-07, | |
| "loss": 0.0411, | |
| "step": 8025 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 2.0621052631578947e-07, | |
| "loss": 0.0389, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 2.0357894736842105e-07, | |
| "loss": 0.022, | |
| "step": 8075 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 2.009473684210526e-07, | |
| "loss": 0.0121, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 1.983157894736842e-07, | |
| "loss": 0.0248, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 1.956842105263158e-07, | |
| "loss": 0.0258, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 1.9305263157894735e-07, | |
| "loss": 0.0281, | |
| "step": 8175 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 1.9042105263157893e-07, | |
| "loss": 0.0169, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.8778947368421051e-07, | |
| "loss": 0.0194, | |
| "step": 8225 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 1.851578947368421e-07, | |
| "loss": 0.0204, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 1.8252631578947368e-07, | |
| "loss": 0.0257, | |
| "step": 8275 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 1.7989473684210526e-07, | |
| "loss": 0.0234, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 1.7726315789473684e-07, | |
| "loss": 0.0249, | |
| "step": 8325 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 1.746315789473684e-07, | |
| "loss": 0.0177, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 1.7210526315789472e-07, | |
| "loss": 0.0358, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 1.694736842105263e-07, | |
| "loss": 0.0393, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 1.668421052631579e-07, | |
| "loss": 0.0337, | |
| "step": 8425 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 1.6421052631578946e-07, | |
| "loss": 0.0238, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 1.6157894736842104e-07, | |
| "loss": 0.0384, | |
| "step": 8475 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 1.5894736842105263e-07, | |
| "loss": 0.0282, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 1.563157894736842e-07, | |
| "loss": 0.0182, | |
| "step": 8525 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 1.536842105263158e-07, | |
| "loss": 0.0189, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 1.5105263157894734e-07, | |
| "loss": 0.0302, | |
| "step": 8575 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 1.4842105263157895e-07, | |
| "loss": 0.0449, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 1.457894736842105e-07, | |
| "loss": 0.0204, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 1.431578947368421e-07, | |
| "loss": 0.02, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 1.405263157894737e-07, | |
| "loss": 0.0227, | |
| "step": 8675 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 1.3789473684210525e-07, | |
| "loss": 0.0226, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 1.3526315789473684e-07, | |
| "loss": 0.0402, | |
| "step": 8725 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 1.3263157894736842e-07, | |
| "loss": 0.0257, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 1.3e-07, | |
| "loss": 0.0205, | |
| "step": 8775 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 1.2736842105263158e-07, | |
| "loss": 0.0261, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 1.2473684210526316e-07, | |
| "loss": 0.0224, | |
| "step": 8825 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 1.2210526315789472e-07, | |
| "loss": 0.014, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 1.194736842105263e-07, | |
| "loss": 0.0249, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 1.168421052631579e-07, | |
| "loss": 0.0144, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 1.1421052631578946e-07, | |
| "loss": 0.0163, | |
| "step": 8925 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 1.1157894736842105e-07, | |
| "loss": 0.0242, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 1.0894736842105263e-07, | |
| "loss": 0.024, | |
| "step": 8975 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 1.0631578947368421e-07, | |
| "loss": 0.0232, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "eval_loss": 0.2443310171365738, | |
| "eval_runtime": 4137.0789, | |
| "eval_samples_per_second": 0.875, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.561837129945587, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 1.0368421052631579e-07, | |
| "loss": 0.0191, | |
| "step": 9025 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 1.0105263157894736e-07, | |
| "loss": 0.0227, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 9.842105263157894e-08, | |
| "loss": 0.0269, | |
| "step": 9075 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 9.578947368421052e-08, | |
| "loss": 0.0211, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 9.31578947368421e-08, | |
| "loss": 0.0237, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 9.052631578947369e-08, | |
| "loss": 0.0243, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 8.789473684210526e-08, | |
| "loss": 0.0202, | |
| "step": 9175 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 8.526315789473684e-08, | |
| "loss": 0.0306, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 8.263157894736842e-08, | |
| "loss": 0.025, | |
| "step": 9225 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 8e-08, | |
| "loss": 0.0323, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 7.736842105263157e-08, | |
| "loss": 0.0238, | |
| "step": 9275 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 7.473684210526315e-08, | |
| "loss": 0.0149, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 7.210526315789473e-08, | |
| "loss": 0.0241, | |
| "step": 9325 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 6.947368421052632e-08, | |
| "loss": 0.026, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 6.68421052631579e-08, | |
| "loss": 0.0171, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 6.421052631578947e-08, | |
| "loss": 0.0219, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 6.157894736842105e-08, | |
| "loss": 0.0225, | |
| "step": 9425 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 5.894736842105263e-08, | |
| "loss": 0.0201, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 5.6315789473684205e-08, | |
| "loss": 0.0218, | |
| "step": 9475 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 5.368421052631579e-08, | |
| "loss": 0.0227, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 5.105263157894737e-08, | |
| "loss": 0.0187, | |
| "step": 9525 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 4.8421052631578944e-08, | |
| "loss": 0.0214, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 4.5789473684210526e-08, | |
| "loss": 0.0309, | |
| "step": 9575 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 4.31578947368421e-08, | |
| "loss": 0.0228, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 4.0526315789473677e-08, | |
| "loss": 0.0185, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 3.7894736842105265e-08, | |
| "loss": 0.0157, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 3.526315789473684e-08, | |
| "loss": 0.0396, | |
| "step": 9675 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 3.263157894736842e-08, | |
| "loss": 0.0177, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 3e-08, | |
| "loss": 0.0197, | |
| "step": 9725 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 2.736842105263158e-08, | |
| "loss": 0.0213, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 2.4736842105263158e-08, | |
| "loss": 0.0185, | |
| "step": 9775 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 2.2105263157894733e-08, | |
| "loss": 0.0288, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 1.9473684210526315e-08, | |
| "loss": 0.0176, | |
| "step": 9825 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 1.6842105263157893e-08, | |
| "loss": 0.0198, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 1.4210526315789474e-08, | |
| "loss": 0.0201, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 1.1578947368421052e-08, | |
| "loss": 0.0253, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 8.94736842105263e-09, | |
| "loss": 0.0324, | |
| "step": 9925 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 6.31578947368421e-09, | |
| "loss": 0.0529, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 3.6842105263157894e-09, | |
| "loss": 0.0237, | |
| "step": 9975 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 1.0526315789473683e-09, | |
| "loss": 0.015, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "eval_loss": 0.24133874475955963, | |
| "eval_runtime": 4136.3131, | |
| "eval_samples_per_second": 0.875, | |
| "eval_steps_per_second": 0.073, | |
| "eval_wer": 6.492668080789449, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "step": 10000, | |
| "total_flos": 2.54695595655168e+20, | |
| "train_loss": 0.07414095425903797, | |
| "train_runtime": 91144.998, | |
| "train_samples_per_second": 1.317, | |
| "train_steps_per_second": 0.11 | |
| } | |
| ], | |
| "max_steps": 10000, | |
| "num_train_epochs": 5, | |
| "total_flos": 2.54695595655168e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |