|
{ |
|
"best_metric": 54.08, |
|
"best_model_checkpoint": "./checkpoint-4000", |
|
"epoch": 15.0205, |
|
"global_step": 10000, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.0, |
|
"learning_rate": 4.4e-07, |
|
"loss": 0.9674, |
|
"step": 25 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 9.400000000000001e-07, |
|
"loss": 0.8775, |
|
"step": 50 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 1.44e-06, |
|
"loss": 0.6455, |
|
"step": 75 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 1.94e-06, |
|
"loss": 0.4408, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 2.4400000000000004e-06, |
|
"loss": 0.4532, |
|
"step": 125 |
|
}, |
|
{ |
|
"epoch": 0.01, |
|
"learning_rate": 2.9400000000000002e-06, |
|
"loss": 0.3882, |
|
"step": 150 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 3.44e-06, |
|
"loss": 0.4273, |
|
"step": 175 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 3.94e-06, |
|
"loss": 0.4366, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 0.02, |
|
"learning_rate": 4.440000000000001e-06, |
|
"loss": 0.4325, |
|
"step": 225 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 4.94e-06, |
|
"loss": 0.385, |
|
"step": 250 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 5.4400000000000004e-06, |
|
"loss": 0.3172, |
|
"step": 275 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 5.94e-06, |
|
"loss": 0.2816, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 0.03, |
|
"learning_rate": 6.440000000000001e-06, |
|
"loss": 0.2695, |
|
"step": 325 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 6.9400000000000005e-06, |
|
"loss": 0.302, |
|
"step": 350 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 7.440000000000001e-06, |
|
"loss": 0.2971, |
|
"step": 375 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 7.94e-06, |
|
"loss": 0.2968, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 8.44e-06, |
|
"loss": 0.3452, |
|
"step": 425 |
|
}, |
|
{ |
|
"epoch": 0.04, |
|
"learning_rate": 8.94e-06, |
|
"loss": 0.356, |
|
"step": 450 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 9.440000000000001e-06, |
|
"loss": 0.3285, |
|
"step": 475 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 9.940000000000001e-06, |
|
"loss": 0.2805, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 0.05, |
|
"learning_rate": 9.976842105263158e-06, |
|
"loss": 0.3373, |
|
"step": 525 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 9.950526315789475e-06, |
|
"loss": 0.2881, |
|
"step": 550 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 9.92421052631579e-06, |
|
"loss": 0.256, |
|
"step": 575 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 9.897894736842107e-06, |
|
"loss": 0.2397, |
|
"step": 600 |
|
}, |
|
{ |
|
"epoch": 0.06, |
|
"learning_rate": 9.871578947368422e-06, |
|
"loss": 0.2604, |
|
"step": 625 |
|
}, |
|
{ |
|
"epoch": 0.07, |
|
"learning_rate": 9.845263157894738e-06, |
|
"loss": 0.3038, |
|
"step": 650 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 9.818947368421053e-06, |
|
"loss": 0.2369, |
|
"step": 675 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 9.79263157894737e-06, |
|
"loss": 0.2556, |
|
"step": 700 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.766315789473685e-06, |
|
"loss": 0.2723, |
|
"step": 725 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.74e-06, |
|
"loss": 0.2401, |
|
"step": 750 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.713684210526317e-06, |
|
"loss": 0.2657, |
|
"step": 775 |
|
}, |
|
{ |
|
"epoch": 1.01, |
|
"learning_rate": 9.687368421052632e-06, |
|
"loss": 0.2259, |
|
"step": 800 |
|
}, |
|
{ |
|
"epoch": 1.02, |
|
"learning_rate": 9.661052631578948e-06, |
|
"loss": 0.2452, |
|
"step": 825 |
|
}, |
|
{ |
|
"epoch": 1.02, |
|
"learning_rate": 9.634736842105265e-06, |
|
"loss": 0.261, |
|
"step": 850 |
|
}, |
|
{ |
|
"epoch": 1.02, |
|
"learning_rate": 9.60842105263158e-06, |
|
"loss": 0.256, |
|
"step": 875 |
|
}, |
|
{ |
|
"epoch": 1.02, |
|
"learning_rate": 9.582105263157897e-06, |
|
"loss": 0.218, |
|
"step": 900 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 9.555789473684211e-06, |
|
"loss": 0.1978, |
|
"step": 925 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 9.529473684210528e-06, |
|
"loss": 0.1739, |
|
"step": 950 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 9.503157894736843e-06, |
|
"loss": 0.1523, |
|
"step": 975 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"learning_rate": 9.476842105263158e-06, |
|
"loss": 0.1885, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 1.03, |
|
"eval_loss": 0.39497342705726624, |
|
"eval_runtime": 1606.636, |
|
"eval_samples_per_second": 6.498, |
|
"eval_steps_per_second": 3.249, |
|
"eval_wer": 66.44, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 1.04, |
|
"learning_rate": 9.450526315789475e-06, |
|
"loss": 0.1694, |
|
"step": 1025 |
|
}, |
|
{ |
|
"epoch": 1.04, |
|
"learning_rate": 9.42421052631579e-06, |
|
"loss": 0.1653, |
|
"step": 1050 |
|
}, |
|
{ |
|
"epoch": 1.04, |
|
"learning_rate": 9.397894736842106e-06, |
|
"loss": 0.1952, |
|
"step": 1075 |
|
}, |
|
{ |
|
"epoch": 1.04, |
|
"learning_rate": 9.371578947368421e-06, |
|
"loss": 0.2001, |
|
"step": 1100 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 9.345263157894738e-06, |
|
"loss": 0.1774, |
|
"step": 1125 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 9.318947368421053e-06, |
|
"loss": 0.1526, |
|
"step": 1150 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 9.292631578947368e-06, |
|
"loss": 0.1961, |
|
"step": 1175 |
|
}, |
|
{ |
|
"epoch": 1.05, |
|
"learning_rate": 9.266315789473685e-06, |
|
"loss": 0.1678, |
|
"step": 1200 |
|
}, |
|
{ |
|
"epoch": 1.06, |
|
"learning_rate": 9.240000000000001e-06, |
|
"loss": 0.1325, |
|
"step": 1225 |
|
}, |
|
{ |
|
"epoch": 1.06, |
|
"learning_rate": 9.213684210526316e-06, |
|
"loss": 0.1348, |
|
"step": 1250 |
|
}, |
|
{ |
|
"epoch": 1.06, |
|
"learning_rate": 9.187368421052633e-06, |
|
"loss": 0.1362, |
|
"step": 1275 |
|
}, |
|
{ |
|
"epoch": 1.06, |
|
"learning_rate": 9.161052631578948e-06, |
|
"loss": 0.1693, |
|
"step": 1300 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 9.134736842105265e-06, |
|
"loss": 0.1395, |
|
"step": 1325 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 9.10842105263158e-06, |
|
"loss": 0.1346, |
|
"step": 1350 |
|
}, |
|
{ |
|
"epoch": 2.01, |
|
"learning_rate": 9.082105263157896e-06, |
|
"loss": 0.1554, |
|
"step": 1375 |
|
}, |
|
{ |
|
"epoch": 2.01, |
|
"learning_rate": 9.055789473684211e-06, |
|
"loss": 0.1491, |
|
"step": 1400 |
|
}, |
|
{ |
|
"epoch": 2.01, |
|
"learning_rate": 9.029473684210526e-06, |
|
"loss": 0.1484, |
|
"step": 1425 |
|
}, |
|
{ |
|
"epoch": 2.01, |
|
"learning_rate": 9.003157894736843e-06, |
|
"loss": 0.1389, |
|
"step": 1450 |
|
}, |
|
{ |
|
"epoch": 2.02, |
|
"learning_rate": 8.976842105263158e-06, |
|
"loss": 0.129, |
|
"step": 1475 |
|
}, |
|
{ |
|
"epoch": 2.02, |
|
"learning_rate": 8.950526315789474e-06, |
|
"loss": 0.1424, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 2.02, |
|
"learning_rate": 8.92421052631579e-06, |
|
"loss": 0.1502, |
|
"step": 1525 |
|
}, |
|
{ |
|
"epoch": 2.02, |
|
"learning_rate": 8.897894736842106e-06, |
|
"loss": 0.1328, |
|
"step": 1550 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 8.871578947368423e-06, |
|
"loss": 0.1153, |
|
"step": 1575 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 8.845263157894738e-06, |
|
"loss": 0.1021, |
|
"step": 1600 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 8.818947368421054e-06, |
|
"loss": 0.0937, |
|
"step": 1625 |
|
}, |
|
{ |
|
"epoch": 2.03, |
|
"learning_rate": 8.79263157894737e-06, |
|
"loss": 0.1058, |
|
"step": 1650 |
|
}, |
|
{ |
|
"epoch": 2.04, |
|
"learning_rate": 8.766315789473684e-06, |
|
"loss": 0.092, |
|
"step": 1675 |
|
}, |
|
{ |
|
"epoch": 2.04, |
|
"learning_rate": 8.740000000000001e-06, |
|
"loss": 0.0931, |
|
"step": 1700 |
|
}, |
|
{ |
|
"epoch": 2.04, |
|
"learning_rate": 8.713684210526316e-06, |
|
"loss": 0.1078, |
|
"step": 1725 |
|
}, |
|
{ |
|
"epoch": 2.04, |
|
"learning_rate": 8.687368421052633e-06, |
|
"loss": 0.1102, |
|
"step": 1750 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 8.661052631578948e-06, |
|
"loss": 0.1078, |
|
"step": 1775 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 8.634736842105264e-06, |
|
"loss": 0.0957, |
|
"step": 1800 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 8.60842105263158e-06, |
|
"loss": 0.1067, |
|
"step": 1825 |
|
}, |
|
{ |
|
"epoch": 2.05, |
|
"learning_rate": 8.582105263157894e-06, |
|
"loss": 0.0987, |
|
"step": 1850 |
|
}, |
|
{ |
|
"epoch": 2.06, |
|
"learning_rate": 8.555789473684213e-06, |
|
"loss": 0.0822, |
|
"step": 1875 |
|
}, |
|
{ |
|
"epoch": 2.06, |
|
"learning_rate": 8.529473684210528e-06, |
|
"loss": 0.0772, |
|
"step": 1900 |
|
}, |
|
{ |
|
"epoch": 2.06, |
|
"learning_rate": 8.503157894736843e-06, |
|
"loss": 0.077, |
|
"step": 1925 |
|
}, |
|
{ |
|
"epoch": 2.06, |
|
"learning_rate": 8.47684210526316e-06, |
|
"loss": 0.0946, |
|
"step": 1950 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"learning_rate": 8.450526315789474e-06, |
|
"loss": 0.082, |
|
"step": 1975 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"learning_rate": 8.42421052631579e-06, |
|
"loss": 0.0794, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_loss": 0.3950011134147644, |
|
"eval_runtime": 1547.0055, |
|
"eval_samples_per_second": 6.749, |
|
"eval_steps_per_second": 3.374, |
|
"eval_wer": 58.550666666666665, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 3.01, |
|
"learning_rate": 8.397894736842106e-06, |
|
"loss": 0.0874, |
|
"step": 2025 |
|
}, |
|
{ |
|
"epoch": 3.01, |
|
"learning_rate": 8.371578947368422e-06, |
|
"loss": 0.0826, |
|
"step": 2050 |
|
}, |
|
{ |
|
"epoch": 3.01, |
|
"learning_rate": 8.345263157894737e-06, |
|
"loss": 0.0874, |
|
"step": 2075 |
|
}, |
|
{ |
|
"epoch": 3.01, |
|
"learning_rate": 8.318947368421052e-06, |
|
"loss": 0.0838, |
|
"step": 2100 |
|
}, |
|
{ |
|
"epoch": 3.02, |
|
"learning_rate": 8.292631578947369e-06, |
|
"loss": 0.0687, |
|
"step": 2125 |
|
}, |
|
{ |
|
"epoch": 3.02, |
|
"learning_rate": 8.266315789473684e-06, |
|
"loss": 0.0784, |
|
"step": 2150 |
|
}, |
|
{ |
|
"epoch": 3.02, |
|
"learning_rate": 8.24e-06, |
|
"loss": 0.0813, |
|
"step": 2175 |
|
}, |
|
{ |
|
"epoch": 3.02, |
|
"learning_rate": 8.213684210526316e-06, |
|
"loss": 0.0741, |
|
"step": 2200 |
|
}, |
|
{ |
|
"epoch": 3.03, |
|
"learning_rate": 8.187368421052632e-06, |
|
"loss": 0.0727, |
|
"step": 2225 |
|
}, |
|
{ |
|
"epoch": 3.03, |
|
"learning_rate": 8.161052631578949e-06, |
|
"loss": 0.0569, |
|
"step": 2250 |
|
}, |
|
{ |
|
"epoch": 3.03, |
|
"learning_rate": 8.134736842105264e-06, |
|
"loss": 0.0551, |
|
"step": 2275 |
|
}, |
|
{ |
|
"epoch": 3.03, |
|
"learning_rate": 8.10842105263158e-06, |
|
"loss": 0.0561, |
|
"step": 2300 |
|
}, |
|
{ |
|
"epoch": 3.04, |
|
"learning_rate": 8.082105263157896e-06, |
|
"loss": 0.0503, |
|
"step": 2325 |
|
}, |
|
{ |
|
"epoch": 3.04, |
|
"learning_rate": 8.05578947368421e-06, |
|
"loss": 0.0511, |
|
"step": 2350 |
|
}, |
|
{ |
|
"epoch": 3.04, |
|
"learning_rate": 8.029473684210527e-06, |
|
"loss": 0.0571, |
|
"step": 2375 |
|
}, |
|
{ |
|
"epoch": 3.04, |
|
"learning_rate": 8.003157894736842e-06, |
|
"loss": 0.0632, |
|
"step": 2400 |
|
}, |
|
{ |
|
"epoch": 3.05, |
|
"learning_rate": 7.976842105263159e-06, |
|
"loss": 0.0566, |
|
"step": 2425 |
|
}, |
|
{ |
|
"epoch": 3.05, |
|
"learning_rate": 7.950526315789474e-06, |
|
"loss": 0.0514, |
|
"step": 2450 |
|
}, |
|
{ |
|
"epoch": 3.05, |
|
"learning_rate": 7.92421052631579e-06, |
|
"loss": 0.0641, |
|
"step": 2475 |
|
}, |
|
{ |
|
"epoch": 3.05, |
|
"learning_rate": 7.897894736842106e-06, |
|
"loss": 0.059, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 3.06, |
|
"learning_rate": 7.87157894736842e-06, |
|
"loss": 0.0465, |
|
"step": 2525 |
|
}, |
|
{ |
|
"epoch": 3.06, |
|
"learning_rate": 7.845263157894739e-06, |
|
"loss": 0.0423, |
|
"step": 2550 |
|
}, |
|
{ |
|
"epoch": 3.06, |
|
"learning_rate": 7.818947368421054e-06, |
|
"loss": 0.0465, |
|
"step": 2575 |
|
}, |
|
{ |
|
"epoch": 3.06, |
|
"learning_rate": 7.792631578947369e-06, |
|
"loss": 0.0538, |
|
"step": 2600 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"learning_rate": 7.766315789473685e-06, |
|
"loss": 0.0471, |
|
"step": 2625 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"learning_rate": 7.74e-06, |
|
"loss": 0.0423, |
|
"step": 2650 |
|
}, |
|
{ |
|
"epoch": 4.01, |
|
"learning_rate": 7.713684210526317e-06, |
|
"loss": 0.0478, |
|
"step": 2675 |
|
}, |
|
{ |
|
"epoch": 4.01, |
|
"learning_rate": 7.687368421052632e-06, |
|
"loss": 0.041, |
|
"step": 2700 |
|
}, |
|
{ |
|
"epoch": 4.01, |
|
"learning_rate": 7.661052631578949e-06, |
|
"loss": 0.0502, |
|
"step": 2725 |
|
}, |
|
{ |
|
"epoch": 4.01, |
|
"learning_rate": 7.634736842105264e-06, |
|
"loss": 0.046, |
|
"step": 2750 |
|
}, |
|
{ |
|
"epoch": 4.02, |
|
"learning_rate": 7.6084210526315795e-06, |
|
"loss": 0.0403, |
|
"step": 2775 |
|
}, |
|
{ |
|
"epoch": 4.02, |
|
"learning_rate": 7.582105263157895e-06, |
|
"loss": 0.0432, |
|
"step": 2800 |
|
}, |
|
{ |
|
"epoch": 4.02, |
|
"learning_rate": 7.555789473684211e-06, |
|
"loss": 0.0439, |
|
"step": 2825 |
|
}, |
|
{ |
|
"epoch": 4.02, |
|
"learning_rate": 7.529473684210526e-06, |
|
"loss": 0.0403, |
|
"step": 2850 |
|
}, |
|
{ |
|
"epoch": 4.03, |
|
"learning_rate": 7.503157894736842e-06, |
|
"loss": 0.035, |
|
"step": 2875 |
|
}, |
|
{ |
|
"epoch": 4.03, |
|
"learning_rate": 7.476842105263159e-06, |
|
"loss": 0.0314, |
|
"step": 2900 |
|
}, |
|
{ |
|
"epoch": 4.03, |
|
"learning_rate": 7.4505263157894744e-06, |
|
"loss": 0.0302, |
|
"step": 2925 |
|
}, |
|
{ |
|
"epoch": 4.03, |
|
"learning_rate": 7.42421052631579e-06, |
|
"loss": 0.0311, |
|
"step": 2950 |
|
}, |
|
{ |
|
"epoch": 4.04, |
|
"learning_rate": 7.397894736842106e-06, |
|
"loss": 0.0279, |
|
"step": 2975 |
|
}, |
|
{ |
|
"epoch": 4.04, |
|
"learning_rate": 7.371578947368422e-06, |
|
"loss": 0.0286, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 4.04, |
|
"eval_loss": 0.4602390229701996, |
|
"eval_runtime": 1605.9238, |
|
"eval_samples_per_second": 6.501, |
|
"eval_steps_per_second": 3.25, |
|
"eval_wer": 63.88, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 4.04, |
|
"learning_rate": 7.345263157894738e-06, |
|
"loss": 0.0259, |
|
"step": 3025 |
|
}, |
|
{ |
|
"epoch": 4.04, |
|
"learning_rate": 7.3189473684210535e-06, |
|
"loss": 0.0373, |
|
"step": 3050 |
|
}, |
|
{ |
|
"epoch": 4.05, |
|
"learning_rate": 7.292631578947369e-06, |
|
"loss": 0.0301, |
|
"step": 3075 |
|
}, |
|
{ |
|
"epoch": 4.05, |
|
"learning_rate": 7.266315789473684e-06, |
|
"loss": 0.0266, |
|
"step": 3100 |
|
}, |
|
{ |
|
"epoch": 4.05, |
|
"learning_rate": 7.24e-06, |
|
"loss": 0.0321, |
|
"step": 3125 |
|
}, |
|
{ |
|
"epoch": 4.05, |
|
"learning_rate": 7.213684210526316e-06, |
|
"loss": 0.0365, |
|
"step": 3150 |
|
}, |
|
{ |
|
"epoch": 4.06, |
|
"learning_rate": 7.187368421052632e-06, |
|
"loss": 0.0256, |
|
"step": 3175 |
|
}, |
|
{ |
|
"epoch": 4.06, |
|
"learning_rate": 7.161052631578948e-06, |
|
"loss": 0.023, |
|
"step": 3200 |
|
}, |
|
{ |
|
"epoch": 4.06, |
|
"learning_rate": 7.134736842105264e-06, |
|
"loss": 0.0276, |
|
"step": 3225 |
|
}, |
|
{ |
|
"epoch": 4.06, |
|
"learning_rate": 7.10842105263158e-06, |
|
"loss": 0.0271, |
|
"step": 3250 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"learning_rate": 7.082105263157896e-06, |
|
"loss": 0.0254, |
|
"step": 3275 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"learning_rate": 7.055789473684212e-06, |
|
"loss": 0.025, |
|
"step": 3300 |
|
}, |
|
{ |
|
"epoch": 5.01, |
|
"learning_rate": 7.0294736842105275e-06, |
|
"loss": 0.0255, |
|
"step": 3325 |
|
}, |
|
{ |
|
"epoch": 5.01, |
|
"learning_rate": 7.0031578947368425e-06, |
|
"loss": 0.0224, |
|
"step": 3350 |
|
}, |
|
{ |
|
"epoch": 5.01, |
|
"learning_rate": 6.976842105263158e-06, |
|
"loss": 0.0222, |
|
"step": 3375 |
|
}, |
|
{ |
|
"epoch": 5.01, |
|
"learning_rate": 6.950526315789474e-06, |
|
"loss": 0.0256, |
|
"step": 3400 |
|
}, |
|
{ |
|
"epoch": 5.02, |
|
"learning_rate": 6.92421052631579e-06, |
|
"loss": 0.0214, |
|
"step": 3425 |
|
}, |
|
{ |
|
"epoch": 5.02, |
|
"learning_rate": 6.897894736842106e-06, |
|
"loss": 0.0231, |
|
"step": 3450 |
|
}, |
|
{ |
|
"epoch": 5.02, |
|
"learning_rate": 6.871578947368422e-06, |
|
"loss": 0.0226, |
|
"step": 3475 |
|
}, |
|
{ |
|
"epoch": 5.02, |
|
"learning_rate": 6.845263157894737e-06, |
|
"loss": 0.0217, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 5.03, |
|
"learning_rate": 6.818947368421052e-06, |
|
"loss": 0.0177, |
|
"step": 3525 |
|
}, |
|
{ |
|
"epoch": 5.03, |
|
"learning_rate": 6.792631578947368e-06, |
|
"loss": 0.0167, |
|
"step": 3550 |
|
}, |
|
{ |
|
"epoch": 5.03, |
|
"learning_rate": 6.766315789473685e-06, |
|
"loss": 0.016, |
|
"step": 3575 |
|
}, |
|
{ |
|
"epoch": 5.03, |
|
"learning_rate": 6.740000000000001e-06, |
|
"loss": 0.0193, |
|
"step": 3600 |
|
}, |
|
{ |
|
"epoch": 5.04, |
|
"learning_rate": 6.7136842105263165e-06, |
|
"loss": 0.0139, |
|
"step": 3625 |
|
}, |
|
{ |
|
"epoch": 5.04, |
|
"learning_rate": 6.687368421052632e-06, |
|
"loss": 0.0127, |
|
"step": 3650 |
|
}, |
|
{ |
|
"epoch": 5.04, |
|
"learning_rate": 6.661052631578948e-06, |
|
"loss": 0.0152, |
|
"step": 3675 |
|
}, |
|
{ |
|
"epoch": 5.04, |
|
"learning_rate": 6.634736842105264e-06, |
|
"loss": 0.0217, |
|
"step": 3700 |
|
}, |
|
{ |
|
"epoch": 5.05, |
|
"learning_rate": 6.60842105263158e-06, |
|
"loss": 0.0155, |
|
"step": 3725 |
|
}, |
|
{ |
|
"epoch": 5.05, |
|
"learning_rate": 6.582105263157896e-06, |
|
"loss": 0.0135, |
|
"step": 3750 |
|
}, |
|
{ |
|
"epoch": 5.05, |
|
"learning_rate": 6.5557894736842106e-06, |
|
"loss": 0.0197, |
|
"step": 3775 |
|
}, |
|
{ |
|
"epoch": 5.05, |
|
"learning_rate": 6.529473684210526e-06, |
|
"loss": 0.0201, |
|
"step": 3800 |
|
}, |
|
{ |
|
"epoch": 5.06, |
|
"learning_rate": 6.503157894736842e-06, |
|
"loss": 0.0156, |
|
"step": 3825 |
|
}, |
|
{ |
|
"epoch": 5.06, |
|
"learning_rate": 6.476842105263158e-06, |
|
"loss": 0.0155, |
|
"step": 3850 |
|
}, |
|
{ |
|
"epoch": 5.06, |
|
"learning_rate": 6.450526315789474e-06, |
|
"loss": 0.0129, |
|
"step": 3875 |
|
}, |
|
{ |
|
"epoch": 5.06, |
|
"learning_rate": 6.42421052631579e-06, |
|
"loss": 0.0134, |
|
"step": 3900 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"learning_rate": 6.397894736842106e-06, |
|
"loss": 0.0142, |
|
"step": 3925 |
|
}, |
|
{ |
|
"epoch": 6.0, |
|
"learning_rate": 6.371578947368422e-06, |
|
"loss": 0.0144, |
|
"step": 3950 |
|
}, |
|
{ |
|
"epoch": 6.01, |
|
"learning_rate": 6.345263157894738e-06, |
|
"loss": 0.014, |
|
"step": 3975 |
|
}, |
|
{ |
|
"epoch": 6.01, |
|
"learning_rate": 6.318947368421054e-06, |
|
"loss": 0.0128, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 6.01, |
|
"eval_loss": 0.4948364794254303, |
|
"eval_runtime": 1573.6399, |
|
"eval_samples_per_second": 6.634, |
|
"eval_steps_per_second": 3.317, |
|
"eval_wer": 54.08, |
|
"step": 4000 |
|
}, |
|
{ |
|
"epoch": 6.01, |
|
"learning_rate": 6.292631578947369e-06, |
|
"loss": 0.0114, |
|
"step": 4025 |
|
}, |
|
{ |
|
"epoch": 6.01, |
|
"learning_rate": 6.266315789473685e-06, |
|
"loss": 0.0136, |
|
"step": 4050 |
|
}, |
|
{ |
|
"epoch": 6.02, |
|
"learning_rate": 6.24e-06, |
|
"loss": 0.014, |
|
"step": 4075 |
|
}, |
|
{ |
|
"epoch": 6.02, |
|
"learning_rate": 6.213684210526316e-06, |
|
"loss": 0.0116, |
|
"step": 4100 |
|
}, |
|
{ |
|
"epoch": 6.02, |
|
"learning_rate": 6.187368421052632e-06, |
|
"loss": 0.0157, |
|
"step": 4125 |
|
}, |
|
{ |
|
"epoch": 6.02, |
|
"learning_rate": 6.161052631578948e-06, |
|
"loss": 0.0123, |
|
"step": 4150 |
|
}, |
|
{ |
|
"epoch": 6.03, |
|
"learning_rate": 6.134736842105264e-06, |
|
"loss": 0.0095, |
|
"step": 4175 |
|
}, |
|
{ |
|
"epoch": 6.03, |
|
"learning_rate": 6.108421052631579e-06, |
|
"loss": 0.0111, |
|
"step": 4200 |
|
}, |
|
{ |
|
"epoch": 6.03, |
|
"learning_rate": 6.0821052631578945e-06, |
|
"loss": 0.0115, |
|
"step": 4225 |
|
}, |
|
{ |
|
"epoch": 6.03, |
|
"learning_rate": 6.055789473684212e-06, |
|
"loss": 0.0099, |
|
"step": 4250 |
|
}, |
|
{ |
|
"epoch": 6.04, |
|
"learning_rate": 6.029473684210527e-06, |
|
"loss": 0.0085, |
|
"step": 4275 |
|
}, |
|
{ |
|
"epoch": 6.04, |
|
"learning_rate": 6.003157894736843e-06, |
|
"loss": 0.0069, |
|
"step": 4300 |
|
}, |
|
{ |
|
"epoch": 6.04, |
|
"learning_rate": 5.976842105263159e-06, |
|
"loss": 0.0072, |
|
"step": 4325 |
|
}, |
|
{ |
|
"epoch": 6.04, |
|
"learning_rate": 5.950526315789474e-06, |
|
"loss": 0.0105, |
|
"step": 4350 |
|
}, |
|
{ |
|
"epoch": 6.05, |
|
"learning_rate": 5.92421052631579e-06, |
|
"loss": 0.0093, |
|
"step": 4375 |
|
}, |
|
{ |
|
"epoch": 6.05, |
|
"learning_rate": 5.897894736842106e-06, |
|
"loss": 0.008, |
|
"step": 4400 |
|
}, |
|
{ |
|
"epoch": 6.05, |
|
"learning_rate": 5.871578947368422e-06, |
|
"loss": 0.01, |
|
"step": 4425 |
|
}, |
|
{ |
|
"epoch": 6.05, |
|
"learning_rate": 5.845263157894737e-06, |
|
"loss": 0.0115, |
|
"step": 4450 |
|
}, |
|
{ |
|
"epoch": 6.06, |
|
"learning_rate": 5.818947368421053e-06, |
|
"loss": 0.0101, |
|
"step": 4475 |
|
}, |
|
{ |
|
"epoch": 6.06, |
|
"learning_rate": 5.7926315789473685e-06, |
|
"loss": 0.0082, |
|
"step": 4500 |
|
}, |
|
{ |
|
"epoch": 6.06, |
|
"learning_rate": 5.766315789473684e-06, |
|
"loss": 0.0071, |
|
"step": 4525 |
|
}, |
|
{ |
|
"epoch": 6.06, |
|
"learning_rate": 5.74e-06, |
|
"loss": 0.0084, |
|
"step": 4550 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"learning_rate": 5.713684210526316e-06, |
|
"loss": 0.0091, |
|
"step": 4575 |
|
}, |
|
{ |
|
"epoch": 7.0, |
|
"learning_rate": 5.687368421052633e-06, |
|
"loss": 0.0071, |
|
"step": 4600 |
|
}, |
|
{ |
|
"epoch": 7.01, |
|
"learning_rate": 5.661052631578948e-06, |
|
"loss": 0.0084, |
|
"step": 4625 |
|
}, |
|
{ |
|
"epoch": 7.01, |
|
"learning_rate": 5.634736842105264e-06, |
|
"loss": 0.0069, |
|
"step": 4650 |
|
}, |
|
{ |
|
"epoch": 7.01, |
|
"learning_rate": 5.60842105263158e-06, |
|
"loss": 0.0063, |
|
"step": 4675 |
|
}, |
|
{ |
|
"epoch": 7.01, |
|
"learning_rate": 5.582105263157895e-06, |
|
"loss": 0.0075, |
|
"step": 4700 |
|
}, |
|
{ |
|
"epoch": 7.02, |
|
"learning_rate": 5.555789473684211e-06, |
|
"loss": 0.008, |
|
"step": 4725 |
|
}, |
|
{ |
|
"epoch": 7.02, |
|
"learning_rate": 5.529473684210527e-06, |
|
"loss": 0.0076, |
|
"step": 4750 |
|
}, |
|
{ |
|
"epoch": 7.02, |
|
"learning_rate": 5.5031578947368425e-06, |
|
"loss": 0.0074, |
|
"step": 4775 |
|
}, |
|
{ |
|
"epoch": 7.02, |
|
"learning_rate": 5.476842105263158e-06, |
|
"loss": 0.0074, |
|
"step": 4800 |
|
}, |
|
{ |
|
"epoch": 7.03, |
|
"learning_rate": 5.450526315789474e-06, |
|
"loss": 0.0057, |
|
"step": 4825 |
|
}, |
|
{ |
|
"epoch": 7.03, |
|
"learning_rate": 5.42421052631579e-06, |
|
"loss": 0.006, |
|
"step": 4850 |
|
}, |
|
{ |
|
"epoch": 7.03, |
|
"learning_rate": 5.397894736842105e-06, |
|
"loss": 0.0058, |
|
"step": 4875 |
|
}, |
|
{ |
|
"epoch": 7.03, |
|
"learning_rate": 5.371578947368421e-06, |
|
"loss": 0.0065, |
|
"step": 4900 |
|
}, |
|
{ |
|
"epoch": 7.04, |
|
"learning_rate": 5.345263157894738e-06, |
|
"loss": 0.0047, |
|
"step": 4925 |
|
}, |
|
{ |
|
"epoch": 7.04, |
|
"learning_rate": 5.318947368421053e-06, |
|
"loss": 0.005, |
|
"step": 4950 |
|
}, |
|
{ |
|
"epoch": 7.04, |
|
"learning_rate": 5.292631578947369e-06, |
|
"loss": 0.0056, |
|
"step": 4975 |
|
}, |
|
{ |
|
"epoch": 7.04, |
|
"learning_rate": 5.266315789473685e-06, |
|
"loss": 0.0048, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 7.04, |
|
"eval_loss": 0.5465821027755737, |
|
"eval_runtime": 1578.2947, |
|
"eval_samples_per_second": 6.615, |
|
"eval_steps_per_second": 3.307, |
|
"eval_wer": 57.986666666666665, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 7.05, |
|
"learning_rate": 5.240000000000001e-06, |
|
"loss": 0.0053, |
|
"step": 5025 |
|
}, |
|
{ |
|
"epoch": 7.05, |
|
"learning_rate": 5.2136842105263165e-06, |
|
"loss": 0.0046, |
|
"step": 5050 |
|
}, |
|
{ |
|
"epoch": 7.05, |
|
"learning_rate": 5.187368421052632e-06, |
|
"loss": 0.005, |
|
"step": 5075 |
|
}, |
|
{ |
|
"epoch": 7.05, |
|
"learning_rate": 5.161052631578948e-06, |
|
"loss": 0.0082, |
|
"step": 5100 |
|
}, |
|
{ |
|
"epoch": 7.06, |
|
"learning_rate": 5.134736842105263e-06, |
|
"loss": 0.0059, |
|
"step": 5125 |
|
}, |
|
{ |
|
"epoch": 7.06, |
|
"learning_rate": 5.108421052631579e-06, |
|
"loss": 0.0056, |
|
"step": 5150 |
|
}, |
|
{ |
|
"epoch": 7.06, |
|
"learning_rate": 5.082105263157895e-06, |
|
"loss": 0.0046, |
|
"step": 5175 |
|
}, |
|
{ |
|
"epoch": 7.06, |
|
"learning_rate": 5.0557894736842105e-06, |
|
"loss": 0.0048, |
|
"step": 5200 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"learning_rate": 5.029473684210526e-06, |
|
"loss": 0.0046, |
|
"step": 5225 |
|
}, |
|
{ |
|
"epoch": 8.0, |
|
"learning_rate": 5.003157894736842e-06, |
|
"loss": 0.0044, |
|
"step": 5250 |
|
}, |
|
{ |
|
"epoch": 8.01, |
|
"learning_rate": 4.976842105263158e-06, |
|
"loss": 0.0048, |
|
"step": 5275 |
|
}, |
|
{ |
|
"epoch": 8.01, |
|
"learning_rate": 4.950526315789474e-06, |
|
"loss": 0.004, |
|
"step": 5300 |
|
}, |
|
{ |
|
"epoch": 8.01, |
|
"learning_rate": 4.92421052631579e-06, |
|
"loss": 0.0038, |
|
"step": 5325 |
|
}, |
|
{ |
|
"epoch": 8.01, |
|
"learning_rate": 4.897894736842106e-06, |
|
"loss": 0.0059, |
|
"step": 5350 |
|
}, |
|
{ |
|
"epoch": 8.02, |
|
"learning_rate": 4.871578947368421e-06, |
|
"loss": 0.0054, |
|
"step": 5375 |
|
}, |
|
{ |
|
"epoch": 8.02, |
|
"learning_rate": 4.845263157894737e-06, |
|
"loss": 0.0047, |
|
"step": 5400 |
|
}, |
|
{ |
|
"epoch": 8.02, |
|
"learning_rate": 4.818947368421053e-06, |
|
"loss": 0.0045, |
|
"step": 5425 |
|
}, |
|
{ |
|
"epoch": 8.02, |
|
"learning_rate": 4.792631578947369e-06, |
|
"loss": 0.0038, |
|
"step": 5450 |
|
}, |
|
{ |
|
"epoch": 8.03, |
|
"learning_rate": 4.7663157894736845e-06, |
|
"loss": 0.0031, |
|
"step": 5475 |
|
}, |
|
{ |
|
"epoch": 8.03, |
|
"learning_rate": 4.74e-06, |
|
"loss": 0.0038, |
|
"step": 5500 |
|
}, |
|
{ |
|
"epoch": 8.03, |
|
"learning_rate": 4.713684210526316e-06, |
|
"loss": 0.003, |
|
"step": 5525 |
|
}, |
|
{ |
|
"epoch": 8.03, |
|
"learning_rate": 4.687368421052632e-06, |
|
"loss": 0.0043, |
|
"step": 5550 |
|
}, |
|
{ |
|
"epoch": 8.04, |
|
"learning_rate": 4.661052631578948e-06, |
|
"loss": 0.0033, |
|
"step": 5575 |
|
}, |
|
{ |
|
"epoch": 8.04, |
|
"learning_rate": 4.634736842105264e-06, |
|
"loss": 0.0028, |
|
"step": 5600 |
|
}, |
|
{ |
|
"epoch": 8.04, |
|
"learning_rate": 4.6084210526315794e-06, |
|
"loss": 0.003, |
|
"step": 5625 |
|
}, |
|
{ |
|
"epoch": 8.04, |
|
"learning_rate": 4.582105263157895e-06, |
|
"loss": 0.0041, |
|
"step": 5650 |
|
}, |
|
{ |
|
"epoch": 8.05, |
|
"learning_rate": 4.55578947368421e-06, |
|
"loss": 0.0029, |
|
"step": 5675 |
|
}, |
|
{ |
|
"epoch": 8.05, |
|
"learning_rate": 4.529473684210527e-06, |
|
"loss": 0.0036, |
|
"step": 5700 |
|
}, |
|
{ |
|
"epoch": 8.05, |
|
"learning_rate": 4.503157894736843e-06, |
|
"loss": 0.0026, |
|
"step": 5725 |
|
}, |
|
{ |
|
"epoch": 8.05, |
|
"learning_rate": 4.4768421052631585e-06, |
|
"loss": 0.0051, |
|
"step": 5750 |
|
}, |
|
{ |
|
"epoch": 8.06, |
|
"learning_rate": 4.450526315789474e-06, |
|
"loss": 0.0033, |
|
"step": 5775 |
|
}, |
|
{ |
|
"epoch": 8.06, |
|
"learning_rate": 4.424210526315789e-06, |
|
"loss": 0.0028, |
|
"step": 5800 |
|
}, |
|
{ |
|
"epoch": 8.06, |
|
"learning_rate": 4.397894736842105e-06, |
|
"loss": 0.003, |
|
"step": 5825 |
|
}, |
|
{ |
|
"epoch": 8.06, |
|
"learning_rate": 4.371578947368421e-06, |
|
"loss": 0.0032, |
|
"step": 5850 |
|
}, |
|
{ |
|
"epoch": 8.07, |
|
"learning_rate": 4.345263157894738e-06, |
|
"loss": 0.0036, |
|
"step": 5875 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"learning_rate": 4.3189473684210535e-06, |
|
"loss": 0.0031, |
|
"step": 5900 |
|
}, |
|
{ |
|
"epoch": 9.0, |
|
"learning_rate": 4.2926315789473684e-06, |
|
"loss": 0.0031, |
|
"step": 5925 |
|
}, |
|
{ |
|
"epoch": 9.01, |
|
"learning_rate": 4.266315789473684e-06, |
|
"loss": 0.003, |
|
"step": 5950 |
|
}, |
|
{ |
|
"epoch": 9.01, |
|
"learning_rate": 4.24e-06, |
|
"loss": 0.0023, |
|
"step": 5975 |
|
}, |
|
{ |
|
"epoch": 9.01, |
|
"learning_rate": 4.213684210526316e-06, |
|
"loss": 0.0029, |
|
"step": 6000 |
|
}, |
|
{ |
|
"epoch": 9.01, |
|
"eval_loss": 0.5710424780845642, |
|
"eval_runtime": 1579.3618, |
|
"eval_samples_per_second": 6.61, |
|
"eval_steps_per_second": 3.305, |
|
"eval_wer": 55.41466666666667, |
|
"step": 6000 |
|
}, |
|
{ |
|
"epoch": 9.01, |
|
"learning_rate": 4.1873684210526325e-06, |
|
"loss": 0.0025, |
|
"step": 6025 |
|
}, |
|
{ |
|
"epoch": 9.02, |
|
"learning_rate": 4.1610526315789475e-06, |
|
"loss": 0.0027, |
|
"step": 6050 |
|
}, |
|
{ |
|
"epoch": 9.02, |
|
"learning_rate": 4.134736842105263e-06, |
|
"loss": 0.0025, |
|
"step": 6075 |
|
}, |
|
{ |
|
"epoch": 9.02, |
|
"learning_rate": 4.108421052631579e-06, |
|
"loss": 0.0027, |
|
"step": 6100 |
|
}, |
|
{ |
|
"epoch": 9.02, |
|
"learning_rate": 4.082105263157895e-06, |
|
"loss": 0.0023, |
|
"step": 6125 |
|
}, |
|
{ |
|
"epoch": 9.03, |
|
"learning_rate": 4.055789473684211e-06, |
|
"loss": 0.002, |
|
"step": 6150 |
|
}, |
|
{ |
|
"epoch": 9.03, |
|
"learning_rate": 4.029473684210527e-06, |
|
"loss": 0.002, |
|
"step": 6175 |
|
}, |
|
{ |
|
"epoch": 9.03, |
|
"learning_rate": 4.0031578947368424e-06, |
|
"loss": 0.0022, |
|
"step": 6200 |
|
}, |
|
{ |
|
"epoch": 9.03, |
|
"learning_rate": 3.976842105263158e-06, |
|
"loss": 0.0018, |
|
"step": 6225 |
|
}, |
|
{ |
|
"epoch": 9.04, |
|
"learning_rate": 3.950526315789474e-06, |
|
"loss": 0.0016, |
|
"step": 6250 |
|
}, |
|
{ |
|
"epoch": 9.04, |
|
"learning_rate": 3.92421052631579e-06, |
|
"loss": 0.0026, |
|
"step": 6275 |
|
}, |
|
{ |
|
"epoch": 9.04, |
|
"learning_rate": 3.897894736842106e-06, |
|
"loss": 0.0015, |
|
"step": 6300 |
|
}, |
|
{ |
|
"epoch": 9.04, |
|
"learning_rate": 3.8715789473684215e-06, |
|
"loss": 0.0016, |
|
"step": 6325 |
|
}, |
|
{ |
|
"epoch": 9.05, |
|
"learning_rate": 3.845263157894737e-06, |
|
"loss": 0.0019, |
|
"step": 6350 |
|
}, |
|
{ |
|
"epoch": 9.05, |
|
"learning_rate": 3.818947368421053e-06, |
|
"loss": 0.0019, |
|
"step": 6375 |
|
}, |
|
{ |
|
"epoch": 9.05, |
|
"learning_rate": 3.792631578947369e-06, |
|
"loss": 0.0022, |
|
"step": 6400 |
|
}, |
|
{ |
|
"epoch": 9.05, |
|
"learning_rate": 3.766315789473685e-06, |
|
"loss": 0.0024, |
|
"step": 6425 |
|
}, |
|
{ |
|
"epoch": 9.06, |
|
"learning_rate": 3.74e-06, |
|
"loss": 0.0021, |
|
"step": 6450 |
|
}, |
|
{ |
|
"epoch": 9.06, |
|
"learning_rate": 3.713684210526316e-06, |
|
"loss": 0.0015, |
|
"step": 6475 |
|
}, |
|
{ |
|
"epoch": 9.06, |
|
"learning_rate": 3.687368421052632e-06, |
|
"loss": 0.0019, |
|
"step": 6500 |
|
}, |
|
{ |
|
"epoch": 9.06, |
|
"learning_rate": 3.6610526315789472e-06, |
|
"loss": 0.0016, |
|
"step": 6525 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"learning_rate": 3.6347368421052635e-06, |
|
"loss": 0.0032, |
|
"step": 6550 |
|
}, |
|
{ |
|
"epoch": 10.0, |
|
"learning_rate": 3.6084210526315793e-06, |
|
"loss": 0.0016, |
|
"step": 6575 |
|
}, |
|
{ |
|
"epoch": 10.01, |
|
"learning_rate": 3.582105263157895e-06, |
|
"loss": 0.0017, |
|
"step": 6600 |
|
}, |
|
{ |
|
"epoch": 10.01, |
|
"learning_rate": 3.555789473684211e-06, |
|
"loss": 0.0017, |
|
"step": 6625 |
|
}, |
|
{ |
|
"epoch": 10.01, |
|
"learning_rate": 3.5294736842105263e-06, |
|
"loss": 0.0014, |
|
"step": 6650 |
|
}, |
|
{ |
|
"epoch": 10.01, |
|
"learning_rate": 3.503157894736842e-06, |
|
"loss": 0.0021, |
|
"step": 6675 |
|
}, |
|
{ |
|
"epoch": 10.02, |
|
"learning_rate": 3.476842105263158e-06, |
|
"loss": 0.0014, |
|
"step": 6700 |
|
}, |
|
{ |
|
"epoch": 10.02, |
|
"learning_rate": 3.450526315789474e-06, |
|
"loss": 0.002, |
|
"step": 6725 |
|
}, |
|
{ |
|
"epoch": 10.02, |
|
"learning_rate": 3.42421052631579e-06, |
|
"loss": 0.0014, |
|
"step": 6750 |
|
}, |
|
{ |
|
"epoch": 10.02, |
|
"learning_rate": 3.3978947368421054e-06, |
|
"loss": 0.0019, |
|
"step": 6775 |
|
}, |
|
{ |
|
"epoch": 10.03, |
|
"learning_rate": 3.3715789473684212e-06, |
|
"loss": 0.0017, |
|
"step": 6800 |
|
}, |
|
{ |
|
"epoch": 10.03, |
|
"learning_rate": 3.345263157894737e-06, |
|
"loss": 0.0012, |
|
"step": 6825 |
|
}, |
|
{ |
|
"epoch": 10.03, |
|
"learning_rate": 3.318947368421053e-06, |
|
"loss": 0.0013, |
|
"step": 6850 |
|
}, |
|
{ |
|
"epoch": 10.03, |
|
"learning_rate": 3.292631578947369e-06, |
|
"loss": 0.0012, |
|
"step": 6875 |
|
}, |
|
{ |
|
"epoch": 10.04, |
|
"learning_rate": 3.2663157894736845e-06, |
|
"loss": 0.0009, |
|
"step": 6900 |
|
}, |
|
{ |
|
"epoch": 10.04, |
|
"learning_rate": 3.2400000000000003e-06, |
|
"loss": 0.0014, |
|
"step": 6925 |
|
}, |
|
{ |
|
"epoch": 10.04, |
|
"learning_rate": 3.213684210526316e-06, |
|
"loss": 0.0009, |
|
"step": 6950 |
|
}, |
|
{ |
|
"epoch": 10.04, |
|
"learning_rate": 3.187368421052632e-06, |
|
"loss": 0.0012, |
|
"step": 6975 |
|
}, |
|
{ |
|
"epoch": 10.05, |
|
"learning_rate": 3.1610526315789474e-06, |
|
"loss": 0.0013, |
|
"step": 7000 |
|
}, |
|
{ |
|
"epoch": 10.05, |
|
"eval_loss": 0.5995678305625916, |
|
"eval_runtime": 1578.9889, |
|
"eval_samples_per_second": 6.612, |
|
"eval_steps_per_second": 3.306, |
|
"eval_wer": 58.77066666666667, |
|
"step": 7000 |
|
}, |
|
{ |
|
"epoch": 10.05, |
|
"learning_rate": 3.134736842105263e-06, |
|
"loss": 0.0013, |
|
"step": 7025 |
|
}, |
|
{ |
|
"epoch": 10.05, |
|
"learning_rate": 3.1084210526315794e-06, |
|
"loss": 0.0021, |
|
"step": 7050 |
|
}, |
|
{ |
|
"epoch": 10.05, |
|
"learning_rate": 3.0821052631578952e-06, |
|
"loss": 0.0016, |
|
"step": 7075 |
|
}, |
|
{ |
|
"epoch": 10.06, |
|
"learning_rate": 3.055789473684211e-06, |
|
"loss": 0.0009, |
|
"step": 7100 |
|
}, |
|
{ |
|
"epoch": 10.06, |
|
"learning_rate": 3.0294736842105264e-06, |
|
"loss": 0.0018, |
|
"step": 7125 |
|
}, |
|
{ |
|
"epoch": 10.06, |
|
"learning_rate": 3.0031578947368423e-06, |
|
"loss": 0.0009, |
|
"step": 7150 |
|
}, |
|
{ |
|
"epoch": 10.06, |
|
"learning_rate": 2.976842105263158e-06, |
|
"loss": 0.001, |
|
"step": 7175 |
|
}, |
|
{ |
|
"epoch": 11.0, |
|
"learning_rate": 2.9505263157894735e-06, |
|
"loss": 0.0012, |
|
"step": 7200 |
|
}, |
|
{ |
|
"epoch": 11.0, |
|
"learning_rate": 2.92421052631579e-06, |
|
"loss": 0.0017, |
|
"step": 7225 |
|
}, |
|
{ |
|
"epoch": 11.01, |
|
"learning_rate": 2.8978947368421055e-06, |
|
"loss": 0.0008, |
|
"step": 7250 |
|
}, |
|
{ |
|
"epoch": 11.01, |
|
"learning_rate": 2.8715789473684214e-06, |
|
"loss": 0.0008, |
|
"step": 7275 |
|
}, |
|
{ |
|
"epoch": 11.01, |
|
"learning_rate": 2.845263157894737e-06, |
|
"loss": 0.0011, |
|
"step": 7300 |
|
}, |
|
{ |
|
"epoch": 11.01, |
|
"learning_rate": 2.8189473684210526e-06, |
|
"loss": 0.0011, |
|
"step": 7325 |
|
}, |
|
{ |
|
"epoch": 11.02, |
|
"learning_rate": 2.7926315789473684e-06, |
|
"loss": 0.0008, |
|
"step": 7350 |
|
}, |
|
{ |
|
"epoch": 11.02, |
|
"learning_rate": 2.766315789473684e-06, |
|
"loss": 0.0009, |
|
"step": 7375 |
|
}, |
|
{ |
|
"epoch": 11.02, |
|
"learning_rate": 2.7400000000000004e-06, |
|
"loss": 0.0011, |
|
"step": 7400 |
|
}, |
|
{ |
|
"epoch": 11.02, |
|
"learning_rate": 2.7136842105263163e-06, |
|
"loss": 0.0007, |
|
"step": 7425 |
|
}, |
|
{ |
|
"epoch": 11.03, |
|
"learning_rate": 2.6873684210526317e-06, |
|
"loss": 0.001, |
|
"step": 7450 |
|
}, |
|
{ |
|
"epoch": 11.03, |
|
"learning_rate": 2.6610526315789475e-06, |
|
"loss": 0.0007, |
|
"step": 7475 |
|
}, |
|
{ |
|
"epoch": 11.03, |
|
"learning_rate": 2.6347368421052633e-06, |
|
"loss": 0.0012, |
|
"step": 7500 |
|
}, |
|
{ |
|
"epoch": 11.03, |
|
"learning_rate": 2.608421052631579e-06, |
|
"loss": 0.0012, |
|
"step": 7525 |
|
}, |
|
{ |
|
"epoch": 11.04, |
|
"learning_rate": 2.5821052631578954e-06, |
|
"loss": 0.0005, |
|
"step": 7550 |
|
}, |
|
{ |
|
"epoch": 11.04, |
|
"learning_rate": 2.5557894736842108e-06, |
|
"loss": 0.0011, |
|
"step": 7575 |
|
}, |
|
{ |
|
"epoch": 11.04, |
|
"learning_rate": 2.5294736842105266e-06, |
|
"loss": 0.0006, |
|
"step": 7600 |
|
}, |
|
{ |
|
"epoch": 11.04, |
|
"learning_rate": 2.5031578947368424e-06, |
|
"loss": 0.0012, |
|
"step": 7625 |
|
}, |
|
{ |
|
"epoch": 11.05, |
|
"learning_rate": 2.476842105263158e-06, |
|
"loss": 0.0007, |
|
"step": 7650 |
|
}, |
|
{ |
|
"epoch": 11.05, |
|
"learning_rate": 2.4505263157894736e-06, |
|
"loss": 0.0007, |
|
"step": 7675 |
|
}, |
|
{ |
|
"epoch": 11.05, |
|
"learning_rate": 2.42421052631579e-06, |
|
"loss": 0.001, |
|
"step": 7700 |
|
}, |
|
{ |
|
"epoch": 11.05, |
|
"learning_rate": 2.3978947368421052e-06, |
|
"loss": 0.0013, |
|
"step": 7725 |
|
}, |
|
{ |
|
"epoch": 11.06, |
|
"learning_rate": 2.371578947368421e-06, |
|
"loss": 0.0007, |
|
"step": 7750 |
|
}, |
|
{ |
|
"epoch": 11.06, |
|
"learning_rate": 2.3452631578947373e-06, |
|
"loss": 0.0006, |
|
"step": 7775 |
|
}, |
|
{ |
|
"epoch": 11.06, |
|
"learning_rate": 2.3189473684210527e-06, |
|
"loss": 0.0006, |
|
"step": 7800 |
|
}, |
|
{ |
|
"epoch": 11.06, |
|
"learning_rate": 2.2926315789473685e-06, |
|
"loss": 0.0006, |
|
"step": 7825 |
|
}, |
|
{ |
|
"epoch": 12.0, |
|
"learning_rate": 2.2663157894736843e-06, |
|
"loss": 0.0009, |
|
"step": 7850 |
|
}, |
|
{ |
|
"epoch": 12.0, |
|
"learning_rate": 2.24e-06, |
|
"loss": 0.0006, |
|
"step": 7875 |
|
}, |
|
{ |
|
"epoch": 12.01, |
|
"learning_rate": 2.213684210526316e-06, |
|
"loss": 0.0006, |
|
"step": 7900 |
|
}, |
|
{ |
|
"epoch": 12.01, |
|
"learning_rate": 2.187368421052632e-06, |
|
"loss": 0.0006, |
|
"step": 7925 |
|
}, |
|
{ |
|
"epoch": 12.01, |
|
"learning_rate": 2.1610526315789476e-06, |
|
"loss": 0.0008, |
|
"step": 7950 |
|
}, |
|
{ |
|
"epoch": 12.01, |
|
"learning_rate": 2.1347368421052634e-06, |
|
"loss": 0.0007, |
|
"step": 7975 |
|
}, |
|
{ |
|
"epoch": 12.02, |
|
"learning_rate": 2.1084210526315792e-06, |
|
"loss": 0.0008, |
|
"step": 8000 |
|
}, |
|
{ |
|
"epoch": 12.02, |
|
"eval_loss": 0.6178876161575317, |
|
"eval_runtime": 1574.1035, |
|
"eval_samples_per_second": 6.632, |
|
"eval_steps_per_second": 3.316, |
|
"eval_wer": 54.748, |
|
"step": 8000 |
|
}, |
|
{ |
|
"epoch": 12.02, |
|
"learning_rate": 2.082105263157895e-06, |
|
"loss": 0.001, |
|
"step": 8025 |
|
}, |
|
{ |
|
"epoch": 12.02, |
|
"learning_rate": 2.055789473684211e-06, |
|
"loss": 0.0011, |
|
"step": 8050 |
|
}, |
|
{ |
|
"epoch": 12.02, |
|
"learning_rate": 2.0294736842105263e-06, |
|
"loss": 0.0006, |
|
"step": 8075 |
|
}, |
|
{ |
|
"epoch": 12.03, |
|
"learning_rate": 2.003157894736842e-06, |
|
"loss": 0.0006, |
|
"step": 8100 |
|
}, |
|
{ |
|
"epoch": 12.03, |
|
"learning_rate": 1.976842105263158e-06, |
|
"loss": 0.0005, |
|
"step": 8125 |
|
}, |
|
{ |
|
"epoch": 12.03, |
|
"learning_rate": 1.9505263157894737e-06, |
|
"loss": 0.0005, |
|
"step": 8150 |
|
}, |
|
{ |
|
"epoch": 12.03, |
|
"learning_rate": 1.9242105263157896e-06, |
|
"loss": 0.0006, |
|
"step": 8175 |
|
}, |
|
{ |
|
"epoch": 12.04, |
|
"learning_rate": 1.8978947368421056e-06, |
|
"loss": 0.0004, |
|
"step": 8200 |
|
}, |
|
{ |
|
"epoch": 12.04, |
|
"learning_rate": 1.8715789473684212e-06, |
|
"loss": 0.0007, |
|
"step": 8225 |
|
}, |
|
{ |
|
"epoch": 12.04, |
|
"learning_rate": 1.845263157894737e-06, |
|
"loss": 0.0005, |
|
"step": 8250 |
|
}, |
|
{ |
|
"epoch": 12.04, |
|
"learning_rate": 1.8189473684210528e-06, |
|
"loss": 0.0008, |
|
"step": 8275 |
|
}, |
|
{ |
|
"epoch": 12.05, |
|
"learning_rate": 1.7926315789473686e-06, |
|
"loss": 0.0005, |
|
"step": 8300 |
|
}, |
|
{ |
|
"epoch": 12.05, |
|
"learning_rate": 1.7663157894736843e-06, |
|
"loss": 0.0008, |
|
"step": 8325 |
|
}, |
|
{ |
|
"epoch": 12.05, |
|
"learning_rate": 1.74e-06, |
|
"loss": 0.0016, |
|
"step": 8350 |
|
}, |
|
{ |
|
"epoch": 12.05, |
|
"learning_rate": 1.713684210526316e-06, |
|
"loss": 0.0009, |
|
"step": 8375 |
|
}, |
|
{ |
|
"epoch": 12.06, |
|
"learning_rate": 1.6873684210526317e-06, |
|
"loss": 0.0005, |
|
"step": 8400 |
|
}, |
|
{ |
|
"epoch": 12.06, |
|
"learning_rate": 1.6610526315789473e-06, |
|
"loss": 0.0004, |
|
"step": 8425 |
|
}, |
|
{ |
|
"epoch": 12.06, |
|
"learning_rate": 1.6357894736842107e-06, |
|
"loss": 0.0012, |
|
"step": 8450 |
|
}, |
|
{ |
|
"epoch": 12.06, |
|
"learning_rate": 1.6094736842105265e-06, |
|
"loss": 0.0004, |
|
"step": 8475 |
|
}, |
|
{ |
|
"epoch": 13.0, |
|
"learning_rate": 1.5831578947368423e-06, |
|
"loss": 0.0004, |
|
"step": 8500 |
|
}, |
|
{ |
|
"epoch": 13.0, |
|
"learning_rate": 1.556842105263158e-06, |
|
"loss": 0.0006, |
|
"step": 8525 |
|
}, |
|
{ |
|
"epoch": 13.01, |
|
"learning_rate": 1.5305263157894737e-06, |
|
"loss": 0.0004, |
|
"step": 8550 |
|
}, |
|
{ |
|
"epoch": 13.01, |
|
"learning_rate": 1.5042105263157898e-06, |
|
"loss": 0.0004, |
|
"step": 8575 |
|
}, |
|
{ |
|
"epoch": 13.01, |
|
"learning_rate": 1.4778947368421054e-06, |
|
"loss": 0.0005, |
|
"step": 8600 |
|
}, |
|
{ |
|
"epoch": 13.01, |
|
"learning_rate": 1.4515789473684212e-06, |
|
"loss": 0.001, |
|
"step": 8625 |
|
}, |
|
{ |
|
"epoch": 13.02, |
|
"learning_rate": 1.425263157894737e-06, |
|
"loss": 0.0005, |
|
"step": 8650 |
|
}, |
|
{ |
|
"epoch": 13.02, |
|
"learning_rate": 1.3989473684210528e-06, |
|
"loss": 0.0005, |
|
"step": 8675 |
|
}, |
|
{ |
|
"epoch": 13.02, |
|
"learning_rate": 1.3726315789473684e-06, |
|
"loss": 0.0004, |
|
"step": 8700 |
|
}, |
|
{ |
|
"epoch": 13.02, |
|
"learning_rate": 1.3463157894736842e-06, |
|
"loss": 0.0005, |
|
"step": 8725 |
|
}, |
|
{ |
|
"epoch": 13.03, |
|
"learning_rate": 1.32e-06, |
|
"loss": 0.0004, |
|
"step": 8750 |
|
}, |
|
{ |
|
"epoch": 13.03, |
|
"learning_rate": 1.2936842105263159e-06, |
|
"loss": 0.0004, |
|
"step": 8775 |
|
}, |
|
{ |
|
"epoch": 13.03, |
|
"learning_rate": 1.2673684210526315e-06, |
|
"loss": 0.0004, |
|
"step": 8800 |
|
}, |
|
{ |
|
"epoch": 13.03, |
|
"learning_rate": 1.2410526315789475e-06, |
|
"loss": 0.001, |
|
"step": 8825 |
|
}, |
|
{ |
|
"epoch": 13.04, |
|
"learning_rate": 1.2147368421052633e-06, |
|
"loss": 0.0003, |
|
"step": 8850 |
|
}, |
|
{ |
|
"epoch": 13.04, |
|
"learning_rate": 1.188421052631579e-06, |
|
"loss": 0.0006, |
|
"step": 8875 |
|
}, |
|
{ |
|
"epoch": 13.04, |
|
"learning_rate": 1.1621052631578948e-06, |
|
"loss": 0.0004, |
|
"step": 8900 |
|
}, |
|
{ |
|
"epoch": 13.04, |
|
"learning_rate": 1.1357894736842106e-06, |
|
"loss": 0.0006, |
|
"step": 8925 |
|
}, |
|
{ |
|
"epoch": 13.05, |
|
"learning_rate": 1.1094736842105264e-06, |
|
"loss": 0.0004, |
|
"step": 8950 |
|
}, |
|
{ |
|
"epoch": 13.05, |
|
"learning_rate": 1.0831578947368422e-06, |
|
"loss": 0.0004, |
|
"step": 8975 |
|
}, |
|
{ |
|
"epoch": 13.05, |
|
"learning_rate": 1.0568421052631578e-06, |
|
"loss": 0.0006, |
|
"step": 9000 |
|
}, |
|
{ |
|
"epoch": 13.05, |
|
"eval_loss": 0.6343392729759216, |
|
"eval_runtime": 1576.7922, |
|
"eval_samples_per_second": 6.621, |
|
"eval_steps_per_second": 3.311, |
|
"eval_wer": 56.26133333333333, |
|
"step": 9000 |
|
}, |
|
{ |
|
"epoch": 13.05, |
|
"learning_rate": 1.0305263157894739e-06, |
|
"loss": 0.0006, |
|
"step": 9025 |
|
}, |
|
{ |
|
"epoch": 13.06, |
|
"learning_rate": 1.0042105263157897e-06, |
|
"loss": 0.0007, |
|
"step": 9050 |
|
}, |
|
{ |
|
"epoch": 13.06, |
|
"learning_rate": 9.778947368421053e-07, |
|
"loss": 0.0003, |
|
"step": 9075 |
|
}, |
|
{ |
|
"epoch": 13.06, |
|
"learning_rate": 9.515789473684212e-07, |
|
"loss": 0.0003, |
|
"step": 9100 |
|
}, |
|
{ |
|
"epoch": 13.06, |
|
"learning_rate": 9.252631578947368e-07, |
|
"loss": 0.0004, |
|
"step": 9125 |
|
}, |
|
{ |
|
"epoch": 14.0, |
|
"learning_rate": 8.989473684210527e-07, |
|
"loss": 0.0004, |
|
"step": 9150 |
|
}, |
|
{ |
|
"epoch": 14.0, |
|
"learning_rate": 8.726315789473686e-07, |
|
"loss": 0.0006, |
|
"step": 9175 |
|
}, |
|
{ |
|
"epoch": 14.01, |
|
"learning_rate": 8.463157894736843e-07, |
|
"loss": 0.0004, |
|
"step": 9200 |
|
}, |
|
{ |
|
"epoch": 14.01, |
|
"learning_rate": 8.200000000000001e-07, |
|
"loss": 0.0004, |
|
"step": 9225 |
|
}, |
|
{ |
|
"epoch": 14.01, |
|
"learning_rate": 7.936842105263158e-07, |
|
"loss": 0.0004, |
|
"step": 9250 |
|
}, |
|
{ |
|
"epoch": 14.01, |
|
"learning_rate": 7.673684210526316e-07, |
|
"loss": 0.0006, |
|
"step": 9275 |
|
}, |
|
{ |
|
"epoch": 14.02, |
|
"learning_rate": 7.410526315789475e-07, |
|
"loss": 0.0004, |
|
"step": 9300 |
|
}, |
|
{ |
|
"epoch": 14.02, |
|
"learning_rate": 7.147368421052632e-07, |
|
"loss": 0.0004, |
|
"step": 9325 |
|
}, |
|
{ |
|
"epoch": 14.02, |
|
"learning_rate": 6.884210526315791e-07, |
|
"loss": 0.0004, |
|
"step": 9350 |
|
}, |
|
{ |
|
"epoch": 14.02, |
|
"learning_rate": 6.621052631578948e-07, |
|
"loss": 0.0004, |
|
"step": 9375 |
|
}, |
|
{ |
|
"epoch": 14.03, |
|
"learning_rate": 6.357894736842106e-07, |
|
"loss": 0.0003, |
|
"step": 9400 |
|
}, |
|
{ |
|
"epoch": 14.03, |
|
"learning_rate": 6.094736842105263e-07, |
|
"loss": 0.0003, |
|
"step": 9425 |
|
}, |
|
{ |
|
"epoch": 14.03, |
|
"learning_rate": 5.831578947368421e-07, |
|
"loss": 0.0004, |
|
"step": 9450 |
|
}, |
|
{ |
|
"epoch": 14.03, |
|
"learning_rate": 5.56842105263158e-07, |
|
"loss": 0.0007, |
|
"step": 9475 |
|
}, |
|
{ |
|
"epoch": 14.04, |
|
"learning_rate": 5.305263157894737e-07, |
|
"loss": 0.0003, |
|
"step": 9500 |
|
}, |
|
{ |
|
"epoch": 14.04, |
|
"learning_rate": 5.042105263157895e-07, |
|
"loss": 0.0005, |
|
"step": 9525 |
|
}, |
|
{ |
|
"epoch": 14.04, |
|
"learning_rate": 4.778947368421053e-07, |
|
"loss": 0.0004, |
|
"step": 9550 |
|
}, |
|
{ |
|
"epoch": 14.04, |
|
"learning_rate": 4.5157894736842107e-07, |
|
"loss": 0.0004, |
|
"step": 9575 |
|
}, |
|
{ |
|
"epoch": 14.05, |
|
"learning_rate": 4.2526315789473684e-07, |
|
"loss": 0.0003, |
|
"step": 9600 |
|
}, |
|
{ |
|
"epoch": 14.05, |
|
"learning_rate": 3.9894736842105266e-07, |
|
"loss": 0.0003, |
|
"step": 9625 |
|
}, |
|
{ |
|
"epoch": 14.05, |
|
"learning_rate": 3.726315789473685e-07, |
|
"loss": 0.0005, |
|
"step": 9650 |
|
}, |
|
{ |
|
"epoch": 14.05, |
|
"learning_rate": 3.4631578947368424e-07, |
|
"loss": 0.0004, |
|
"step": 9675 |
|
}, |
|
{ |
|
"epoch": 14.06, |
|
"learning_rate": 3.2e-07, |
|
"loss": 0.0004, |
|
"step": 9700 |
|
}, |
|
{ |
|
"epoch": 14.06, |
|
"learning_rate": 2.936842105263158e-07, |
|
"loss": 0.0006, |
|
"step": 9725 |
|
}, |
|
{ |
|
"epoch": 14.06, |
|
"learning_rate": 2.6736842105263164e-07, |
|
"loss": 0.0003, |
|
"step": 9750 |
|
}, |
|
{ |
|
"epoch": 14.06, |
|
"learning_rate": 2.410526315789474e-07, |
|
"loss": 0.0003, |
|
"step": 9775 |
|
}, |
|
{ |
|
"epoch": 15.0, |
|
"learning_rate": 2.1473684210526317e-07, |
|
"loss": 0.0004, |
|
"step": 9800 |
|
}, |
|
{ |
|
"epoch": 15.0, |
|
"learning_rate": 1.8842105263157897e-07, |
|
"loss": 0.0008, |
|
"step": 9825 |
|
}, |
|
{ |
|
"epoch": 15.01, |
|
"learning_rate": 1.6210526315789476e-07, |
|
"loss": 0.0003, |
|
"step": 9850 |
|
}, |
|
{ |
|
"epoch": 15.01, |
|
"learning_rate": 1.3578947368421055e-07, |
|
"loss": 0.0004, |
|
"step": 9875 |
|
}, |
|
{ |
|
"epoch": 15.01, |
|
"learning_rate": 1.0947368421052632e-07, |
|
"loss": 0.0004, |
|
"step": 9900 |
|
}, |
|
{ |
|
"epoch": 15.01, |
|
"learning_rate": 8.315789473684211e-08, |
|
"loss": 0.0004, |
|
"step": 9925 |
|
}, |
|
{ |
|
"epoch": 15.02, |
|
"learning_rate": 5.68421052631579e-08, |
|
"loss": 0.0006, |
|
"step": 9950 |
|
}, |
|
{ |
|
"epoch": 15.02, |
|
"learning_rate": 3.0526315789473686e-08, |
|
"loss": 0.0004, |
|
"step": 9975 |
|
}, |
|
{ |
|
"epoch": 15.02, |
|
"learning_rate": 4.210526315789474e-09, |
|
"loss": 0.0003, |
|
"step": 10000 |
|
}, |
|
{ |
|
"epoch": 15.02, |
|
"eval_loss": 0.6387761831283569, |
|
"eval_runtime": 1577.8375, |
|
"eval_samples_per_second": 6.617, |
|
"eval_steps_per_second": 3.308, |
|
"eval_wer": 56.228, |
|
"step": 10000 |
|
}, |
|
{ |
|
"epoch": 15.02, |
|
"step": 10000, |
|
"total_flos": 9.2260752482304e+19, |
|
"train_loss": 0.05549127077427693, |
|
"train_runtime": 51468.4598, |
|
"train_samples_per_second": 6.217, |
|
"train_steps_per_second": 0.194 |
|
} |
|
], |
|
"max_steps": 10000, |
|
"num_train_epochs": 9223372036854775807, |
|
"total_flos": 9.2260752482304e+19, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|