| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 59.0029, | |
| "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": 5e-06, | |
| "loss": 2.5766, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1e-05, | |
| "loss": 0.8604, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.5e-05, | |
| "loss": 0.6098, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2e-05, | |
| "loss": 0.5684, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.5e-05, | |
| "loss": 0.3789, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3e-05, | |
| "loss": 0.4968, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.5e-05, | |
| "loss": 0.4028, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4e-05, | |
| "loss": 0.3609, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 4.5e-05, | |
| "loss": 0.4086, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 5e-05, | |
| "loss": 0.3733, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 5.500000000000001e-05, | |
| "loss": 0.447, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 6e-05, | |
| "loss": 0.2135, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 6.500000000000001e-05, | |
| "loss": 0.283, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 7e-05, | |
| "loss": 0.2141, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 7.500000000000001e-05, | |
| "loss": 0.2942, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8e-05, | |
| "loss": 0.328, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8.5e-05, | |
| "loss": 0.2984, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9e-05, | |
| "loss": 0.3605, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.5e-05, | |
| "loss": 0.1459, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.98e-05, | |
| "loss": 0.1493, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "eval_loss": 0.39604946970939636, | |
| "eval_runtime": 330.6469, | |
| "eval_samples_per_second": 8.75, | |
| "eval_steps_per_second": 0.275, | |
| "eval_wer": 28.977890540050744, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.974736842105264e-05, | |
| "loss": 0.1667, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.948421052631579e-05, | |
| "loss": 0.2499, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.922105263157895e-05, | |
| "loss": 0.2577, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.89578947368421e-05, | |
| "loss": 0.2223, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.869473684210528e-05, | |
| "loss": 0.2246, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.843157894736843e-05, | |
| "loss": 0.1294, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 9.816842105263159e-05, | |
| "loss": 0.0672, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 9.790526315789475e-05, | |
| "loss": 0.1302, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 9.76421052631579e-05, | |
| "loss": 0.1537, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 9.737894736842106e-05, | |
| "loss": 0.1676, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 9.711578947368422e-05, | |
| "loss": 0.1513, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 9.685263157894737e-05, | |
| "loss": 0.1065, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 9.658947368421053e-05, | |
| "loss": 0.069, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 9.632631578947368e-05, | |
| "loss": 0.0628, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 9.606315789473684e-05, | |
| "loss": 0.0785, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 9.58e-05, | |
| "loss": 0.0986, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 9.553684210526315e-05, | |
| "loss": 0.0922, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 9.527368421052631e-05, | |
| "loss": 0.0885, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 9.501052631578948e-05, | |
| "loss": 0.0506, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 9.474736842105264e-05, | |
| "loss": 0.0383, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "eval_loss": 0.4041108787059784, | |
| "eval_runtime": 312.2575, | |
| "eval_samples_per_second": 9.265, | |
| "eval_steps_per_second": 0.291, | |
| "eval_wer": 26.513229430953245, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 9.44842105263158e-05, | |
| "loss": 0.041, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 9.422105263157895e-05, | |
| "loss": 0.0563, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 9.39578947368421e-05, | |
| "loss": 0.0556, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 9.369473684210526e-05, | |
| "loss": 0.0585, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 9.343157894736842e-05, | |
| "loss": 0.0521, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 9.316842105263158e-05, | |
| "loss": 0.0451, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 9.290526315789475e-05, | |
| "loss": 0.0278, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 9.26421052631579e-05, | |
| "loss": 0.0363, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 9.237894736842106e-05, | |
| "loss": 0.0375, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 9.211578947368422e-05, | |
| "loss": 0.0406, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 9.185263157894737e-05, | |
| "loss": 0.0402, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 9.158947368421054e-05, | |
| "loss": 0.0326, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 9.13263157894737e-05, | |
| "loss": 0.0339, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 9.106315789473686e-05, | |
| "loss": 0.0152, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 9.080000000000001e-05, | |
| "loss": 0.0312, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 9.053684210526317e-05, | |
| "loss": 0.0276, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 9.027368421052632e-05, | |
| "loss": 0.0338, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 9.001052631578948e-05, | |
| "loss": 0.0305, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 8.974736842105264e-05, | |
| "loss": 0.0244, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 8.94842105263158e-05, | |
| "loss": 0.0264, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "eval_loss": 0.42740288376808167, | |
| "eval_runtime": 308.9915, | |
| "eval_samples_per_second": 9.363, | |
| "eval_steps_per_second": 0.295, | |
| "eval_wer": 25.58898151504168, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 8.922105263157895e-05, | |
| "loss": 0.0141, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 8.895789473684211e-05, | |
| "loss": 0.024, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 8.869473684210526e-05, | |
| "loss": 0.0259, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 8.843157894736842e-05, | |
| "loss": 0.0284, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 8.816842105263158e-05, | |
| "loss": 0.0248, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 8.790526315789475e-05, | |
| "loss": 0.0229, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 8.76421052631579e-05, | |
| "loss": 0.0128, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 8.737894736842106e-05, | |
| "loss": 0.0097, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 8.711578947368422e-05, | |
| "loss": 0.015, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 8.685263157894737e-05, | |
| "loss": 0.0238, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 8.658947368421053e-05, | |
| "loss": 0.0186, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 8.632631578947369e-05, | |
| "loss": 0.017, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 8.606315789473684e-05, | |
| "loss": 0.0143, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 8.58e-05, | |
| "loss": 0.0085, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 8.553684210526315e-05, | |
| "loss": 0.0094, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 8.527368421052631e-05, | |
| "loss": 0.0128, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 8.501052631578947e-05, | |
| "loss": 0.0116, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 8.474736842105262e-05, | |
| "loss": 0.0105, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 8.44842105263158e-05, | |
| "loss": 0.0145, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 8.422105263157895e-05, | |
| "loss": 0.0155, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "eval_loss": 0.44370800256729126, | |
| "eval_runtime": 309.4934, | |
| "eval_samples_per_second": 9.348, | |
| "eval_steps_per_second": 0.294, | |
| "eval_wer": 24.77346864806089, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 8.395789473684211e-05, | |
| "loss": 0.0061, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 8.369473684210526e-05, | |
| "loss": 0.0102, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 8.343157894736843e-05, | |
| "loss": 0.0105, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 8.316842105263159e-05, | |
| "loss": 0.0136, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 8.290526315789475e-05, | |
| "loss": 0.0111, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 8.26421052631579e-05, | |
| "loss": 0.0109, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 8.237894736842106e-05, | |
| "loss": 0.0071, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 8.211578947368422e-05, | |
| "loss": 0.0038, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 8.185263157894737e-05, | |
| "loss": 0.0098, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 8.158947368421053e-05, | |
| "loss": 0.0139, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 8.132631578947369e-05, | |
| "loss": 0.0096, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 8.106315789473684e-05, | |
| "loss": 0.0139, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 8.080000000000001e-05, | |
| "loss": 0.0109, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 8.053684210526317e-05, | |
| "loss": 0.0044, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 8.027368421052633e-05, | |
| "loss": 0.0054, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 8.001052631578948e-05, | |
| "loss": 0.0043, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 7.974736842105264e-05, | |
| "loss": 0.0072, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 7.94842105263158e-05, | |
| "loss": 0.0052, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 7.922105263157895e-05, | |
| "loss": 0.0066, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 7.895789473684211e-05, | |
| "loss": 0.0041, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "eval_loss": 0.4453926980495453, | |
| "eval_runtime": 317.0387, | |
| "eval_samples_per_second": 9.125, | |
| "eval_steps_per_second": 0.287, | |
| "eval_wer": 25.045306270387822, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 7.869473684210526e-05, | |
| "loss": 0.0064, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 7.843157894736842e-05, | |
| "loss": 0.0042, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 7.816842105263158e-05, | |
| "loss": 0.008, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 7.790526315789473e-05, | |
| "loss": 0.0044, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 7.764210526315789e-05, | |
| "loss": 0.0063, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 7.737894736842105e-05, | |
| "loss": 0.0087, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 7.711578947368422e-05, | |
| "loss": 0.0075, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 7.685263157894737e-05, | |
| "loss": 0.0071, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 7.658947368421053e-05, | |
| "loss": 0.003, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 7.632631578947369e-05, | |
| "loss": 0.0045, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 7.607368421052632e-05, | |
| "loss": 0.0089, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 7.581052631578947e-05, | |
| "loss": 0.0063, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 7.554736842105263e-05, | |
| "loss": 0.008, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 7.528421052631579e-05, | |
| "loss": 0.0064, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 7.502105263157894e-05, | |
| "loss": 0.0068, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 7.47578947368421e-05, | |
| "loss": 0.0053, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 7.449473684210526e-05, | |
| "loss": 0.0043, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 7.423157894736843e-05, | |
| "loss": 0.0099, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 7.396842105263158e-05, | |
| "loss": 0.0036, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 7.370526315789474e-05, | |
| "loss": 0.0044, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "eval_loss": 0.4444020986557007, | |
| "eval_runtime": 314.034, | |
| "eval_samples_per_second": 9.212, | |
| "eval_steps_per_second": 0.29, | |
| "eval_wer": 23.976078289235232, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 7.34421052631579e-05, | |
| "loss": 0.0073, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 7.317894736842105e-05, | |
| "loss": 0.0035, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 7.291578947368421e-05, | |
| "loss": 0.0039, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 7.265263157894737e-05, | |
| "loss": 0.005, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 7.238947368421052e-05, | |
| "loss": 0.0074, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 7.212631578947369e-05, | |
| "loss": 0.0055, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 7.186315789473685e-05, | |
| "loss": 0.0045, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 7.16e-05, | |
| "loss": 0.0081, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 7.133684210526316e-05, | |
| "loss": 0.0026, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 7.107368421052632e-05, | |
| "loss": 0.0032, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 7.081052631578948e-05, | |
| "loss": 0.0039, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 7.054736842105265e-05, | |
| "loss": 0.0097, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 7.02842105263158e-05, | |
| "loss": 0.0086, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 7.002105263157896e-05, | |
| "loss": 0.0063, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 6.975789473684211e-05, | |
| "loss": 0.0074, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 6.949473684210527e-05, | |
| "loss": 0.005, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 6.923157894736843e-05, | |
| "loss": 0.0065, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 6.896842105263158e-05, | |
| "loss": 0.0097, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 6.870526315789474e-05, | |
| "loss": 0.0058, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 6.84421052631579e-05, | |
| "loss": 0.0044, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "eval_loss": 0.4394459128379822, | |
| "eval_runtime": 312.7908, | |
| "eval_samples_per_second": 9.249, | |
| "eval_steps_per_second": 0.291, | |
| "eval_wer": 23.486770569046755, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 6.817894736842105e-05, | |
| "loss": 0.0036, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 6.791578947368421e-05, | |
| "loss": 0.0038, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 6.765263157894737e-05, | |
| "loss": 0.0046, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 6.738947368421052e-05, | |
| "loss": 0.0046, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 6.712631578947369e-05, | |
| "loss": 0.0067, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 6.686315789473685e-05, | |
| "loss": 0.0044, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 6.66e-05, | |
| "loss": 0.0048, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 6.633684210526316e-05, | |
| "loss": 0.0058, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 6.607368421052632e-05, | |
| "loss": 0.004, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 6.581052631578948e-05, | |
| "loss": 0.0067, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 6.554736842105263e-05, | |
| "loss": 0.0059, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 6.528421052631579e-05, | |
| "loss": 0.0061, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 6.502105263157895e-05, | |
| "loss": 0.0052, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 6.47578947368421e-05, | |
| "loss": 0.0033, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 6.449473684210526e-05, | |
| "loss": 0.0024, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 6.423157894736841e-05, | |
| "loss": 0.0022, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 6.396842105263157e-05, | |
| "loss": 0.0019, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 6.370526315789474e-05, | |
| "loss": 0.0051, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 6.34421052631579e-05, | |
| "loss": 0.0024, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 6.317894736842105e-05, | |
| "loss": 0.0022, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "eval_loss": 0.4414944350719452, | |
| "eval_runtime": 311.785, | |
| "eval_samples_per_second": 9.279, | |
| "eval_steps_per_second": 0.292, | |
| "eval_wer": 22.852482783617255, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 6.291578947368421e-05, | |
| "loss": 0.0033, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 6.265263157894738e-05, | |
| "loss": 0.0012, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 6.238947368421054e-05, | |
| "loss": 0.001, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 6.21263157894737e-05, | |
| "loss": 0.0013, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 6.186315789473685e-05, | |
| "loss": 0.0025, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 6.16e-05, | |
| "loss": 0.0024, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 6.133684210526316e-05, | |
| "loss": 0.0019, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 6.107368421052632e-05, | |
| "loss": 0.0025, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 6.0810526315789476e-05, | |
| "loss": 0.0007, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "learning_rate": 6.054736842105263e-05, | |
| "loss": 0.0014, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "learning_rate": 6.028421052631579e-05, | |
| "loss": 0.0024, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 25.01, | |
| "learning_rate": 6.002105263157895e-05, | |
| "loss": 0.0017, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 25.01, | |
| "learning_rate": 5.9757894736842116e-05, | |
| "loss": 0.0046, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 25.01, | |
| "learning_rate": 5.949473684210527e-05, | |
| "loss": 0.0037, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 25.02, | |
| "learning_rate": 5.923157894736843e-05, | |
| "loss": 0.0013, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "learning_rate": 5.8968421052631585e-05, | |
| "loss": 0.0041, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "learning_rate": 5.870526315789474e-05, | |
| "loss": 0.0036, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "learning_rate": 5.84421052631579e-05, | |
| "loss": 0.0031, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "learning_rate": 5.8178947368421054e-05, | |
| "loss": 0.0051, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "learning_rate": 5.791578947368421e-05, | |
| "loss": 0.0034, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "eval_loss": 0.4601580798625946, | |
| "eval_runtime": 304.9721, | |
| "eval_samples_per_second": 9.486, | |
| "eval_steps_per_second": 0.298, | |
| "eval_wer": 23.649873142442914, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "learning_rate": 5.765263157894737e-05, | |
| "loss": 0.0059, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 26.02, | |
| "learning_rate": 5.7389473684210524e-05, | |
| "loss": 0.0033, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "learning_rate": 5.712631578947368e-05, | |
| "loss": 0.0015, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "learning_rate": 5.6863157894736844e-05, | |
| "loss": 0.0054, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 27.01, | |
| "learning_rate": 5.66e-05, | |
| "loss": 0.0031, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 27.01, | |
| "learning_rate": 5.6336842105263163e-05, | |
| "loss": 0.0021, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 27.01, | |
| "learning_rate": 5.607368421052632e-05, | |
| "loss": 0.0056, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 27.01, | |
| "learning_rate": 5.5810526315789476e-05, | |
| "loss": 0.0028, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 27.02, | |
| "learning_rate": 5.554736842105264e-05, | |
| "loss": 0.0009, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "learning_rate": 5.5284210526315796e-05, | |
| "loss": 0.0023, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "learning_rate": 5.502105263157895e-05, | |
| "loss": 0.0023, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 5.475789473684211e-05, | |
| "loss": 0.0033, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 5.4494736842105266e-05, | |
| "loss": 0.0037, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 5.423157894736842e-05, | |
| "loss": 0.001, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 5.396842105263158e-05, | |
| "loss": 0.0014, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 28.02, | |
| "learning_rate": 5.3705263157894735e-05, | |
| "loss": 0.0007, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "learning_rate": 5.344210526315789e-05, | |
| "loss": 0.0041, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "learning_rate": 5.317894736842105e-05, | |
| "loss": 0.0028, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "learning_rate": 5.291578947368422e-05, | |
| "loss": 0.0021, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "learning_rate": 5.2652631578947375e-05, | |
| "loss": 0.0027, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "eval_loss": 0.4576888084411621, | |
| "eval_runtime": 315.3533, | |
| "eval_samples_per_second": 9.174, | |
| "eval_steps_per_second": 0.289, | |
| "eval_wer": 23.378035520115983, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "learning_rate": 5.238947368421053e-05, | |
| "loss": 0.0022, | |
| "step": 5025 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "learning_rate": 5.212631578947369e-05, | |
| "loss": 0.0038, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "learning_rate": 5.1863157894736844e-05, | |
| "loss": 0.0012, | |
| "step": 5075 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "learning_rate": 5.16e-05, | |
| "loss": 0.0035, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 30.01, | |
| "learning_rate": 5.1336842105263164e-05, | |
| "loss": 0.0036, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 30.01, | |
| "learning_rate": 5.107368421052632e-05, | |
| "loss": 0.0029, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 30.01, | |
| "learning_rate": 5.081052631578948e-05, | |
| "loss": 0.0034, | |
| "step": 5175 | |
| }, | |
| { | |
| "epoch": 30.01, | |
| "learning_rate": 5.054736842105263e-05, | |
| "loss": 0.0076, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 30.02, | |
| "learning_rate": 5.028421052631579e-05, | |
| "loss": 0.0038, | |
| "step": 5225 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "learning_rate": 5.0021052631578946e-05, | |
| "loss": 0.0011, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "learning_rate": 4.975789473684211e-05, | |
| "loss": 0.003, | |
| "step": 5275 | |
| }, | |
| { | |
| "epoch": 31.01, | |
| "learning_rate": 4.9494736842105266e-05, | |
| "loss": 0.0055, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 31.01, | |
| "learning_rate": 4.923157894736842e-05, | |
| "loss": 0.0061, | |
| "step": 5325 | |
| }, | |
| { | |
| "epoch": 31.01, | |
| "learning_rate": 4.896842105263158e-05, | |
| "loss": 0.0051, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 31.01, | |
| "learning_rate": 4.8705263157894736e-05, | |
| "loss": 0.0034, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 31.02, | |
| "learning_rate": 4.84421052631579e-05, | |
| "loss": 0.002, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "learning_rate": 4.8178947368421055e-05, | |
| "loss": 0.0027, | |
| "step": 5425 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "learning_rate": 4.791578947368421e-05, | |
| "loss": 0.0013, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "learning_rate": 4.765263157894737e-05, | |
| "loss": 0.0068, | |
| "step": 5475 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "learning_rate": 4.7389473684210525e-05, | |
| "loss": 0.0072, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "eval_loss": 0.45727378129959106, | |
| "eval_runtime": 313.1992, | |
| "eval_samples_per_second": 9.237, | |
| "eval_steps_per_second": 0.291, | |
| "eval_wer": 23.396158028271113, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "learning_rate": 4.712631578947369e-05, | |
| "loss": 0.0043, | |
| "step": 5525 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "learning_rate": 4.6863157894736845e-05, | |
| "loss": 0.0025, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 32.02, | |
| "learning_rate": 4.660000000000001e-05, | |
| "loss": 0.0017, | |
| "step": 5575 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "learning_rate": 4.6336842105263164e-05, | |
| "loss": 0.0062, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "learning_rate": 4.607368421052632e-05, | |
| "loss": 0.0044, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 33.01, | |
| "learning_rate": 4.581052631578948e-05, | |
| "loss": 0.0024, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 33.01, | |
| "learning_rate": 4.5547368421052634e-05, | |
| "loss": 0.0044, | |
| "step": 5675 | |
| }, | |
| { | |
| "epoch": 33.01, | |
| "learning_rate": 4.528421052631579e-05, | |
| "loss": 0.004, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 33.01, | |
| "learning_rate": 4.502105263157895e-05, | |
| "loss": 0.0059, | |
| "step": 5725 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "learning_rate": 4.475789473684211e-05, | |
| "loss": 0.0034, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "learning_rate": 4.4494736842105267e-05, | |
| "loss": 0.0024, | |
| "step": 5775 | |
| }, | |
| { | |
| "epoch": 34.01, | |
| "learning_rate": 4.423157894736842e-05, | |
| "loss": 0.0009, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 34.01, | |
| "learning_rate": 4.396842105263158e-05, | |
| "loss": 0.0012, | |
| "step": 5825 | |
| }, | |
| { | |
| "epoch": 34.01, | |
| "learning_rate": 4.3705263157894736e-05, | |
| "loss": 0.001, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 34.01, | |
| "learning_rate": 4.344210526315789e-05, | |
| "loss": 0.0002, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 34.02, | |
| "learning_rate": 4.317894736842105e-05, | |
| "loss": 0.0016, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "learning_rate": 4.291578947368421e-05, | |
| "loss": 0.0013, | |
| "step": 5925 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "learning_rate": 4.265263157894737e-05, | |
| "loss": 0.0002, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "learning_rate": 4.238947368421053e-05, | |
| "loss": 0.0002, | |
| "step": 5975 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "learning_rate": 4.212631578947369e-05, | |
| "loss": 0.0002, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "eval_loss": 0.46734702587127686, | |
| "eval_runtime": 309.9516, | |
| "eval_samples_per_second": 9.334, | |
| "eval_steps_per_second": 0.294, | |
| "eval_wer": 23.106197897789055, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "learning_rate": 4.1863157894736845e-05, | |
| "loss": 0.0001, | |
| "step": 6025 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "learning_rate": 4.16e-05, | |
| "loss": 0.0002, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 35.02, | |
| "learning_rate": 4.133684210526316e-05, | |
| "loss": 0.0011, | |
| "step": 6075 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "learning_rate": 4.107368421052632e-05, | |
| "loss": 0.0006, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "learning_rate": 4.081052631578948e-05, | |
| "loss": 0.0001, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 36.01, | |
| "learning_rate": 4.0547368421052634e-05, | |
| "loss": 0.0001, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 36.01, | |
| "learning_rate": 4.028421052631579e-05, | |
| "loss": 0.0001, | |
| "step": 6175 | |
| }, | |
| { | |
| "epoch": 36.01, | |
| "learning_rate": 4.002105263157895e-05, | |
| "loss": 0.0001, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 36.01, | |
| "learning_rate": 3.9757894736842104e-05, | |
| "loss": 0.0001, | |
| "step": 6225 | |
| }, | |
| { | |
| "epoch": 36.02, | |
| "learning_rate": 3.949473684210527e-05, | |
| "loss": 0.0009, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "learning_rate": 3.9231578947368423e-05, | |
| "loss": 0.0008, | |
| "step": 6275 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "learning_rate": 3.896842105263158e-05, | |
| "loss": 0.0001, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 37.01, | |
| "learning_rate": 3.8705263157894736e-05, | |
| "loss": 0.0001, | |
| "step": 6325 | |
| }, | |
| { | |
| "epoch": 37.01, | |
| "learning_rate": 3.844210526315789e-05, | |
| "loss": 0.0001, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 37.01, | |
| "learning_rate": 3.8178947368421056e-05, | |
| "loss": 0.0001, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 37.01, | |
| "learning_rate": 3.791578947368421e-05, | |
| "loss": 0.0001, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "learning_rate": 3.7652631578947376e-05, | |
| "loss": 0.0018, | |
| "step": 6425 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "learning_rate": 3.738947368421053e-05, | |
| "loss": 0.0001, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "learning_rate": 3.712631578947369e-05, | |
| "loss": 0.0001, | |
| "step": 6475 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "learning_rate": 3.6863157894736845e-05, | |
| "loss": 0.0001, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "eval_loss": 0.472318172454834, | |
| "eval_runtime": 306.523, | |
| "eval_samples_per_second": 9.438, | |
| "eval_steps_per_second": 0.297, | |
| "eval_wer": 22.997462848858284, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "learning_rate": 3.66e-05, | |
| "loss": 0.0001, | |
| "step": 6525 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "learning_rate": 3.633684210526316e-05, | |
| "loss": 0.0001, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 38.02, | |
| "learning_rate": 3.6073684210526315e-05, | |
| "loss": 0.0009, | |
| "step": 6575 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "learning_rate": 3.581052631578948e-05, | |
| "loss": 0.0004, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "learning_rate": 3.5547368421052635e-05, | |
| "loss": 0.0001, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "learning_rate": 3.528421052631579e-05, | |
| "loss": 0.0001, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "learning_rate": 3.502105263157895e-05, | |
| "loss": 0.0001, | |
| "step": 6675 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "learning_rate": 3.4757894736842104e-05, | |
| "loss": 0.0001, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "learning_rate": 3.449473684210526e-05, | |
| "loss": 0.0001, | |
| "step": 6725 | |
| }, | |
| { | |
| "epoch": 39.02, | |
| "learning_rate": 3.423157894736842e-05, | |
| "loss": 0.0006, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "learning_rate": 3.396842105263158e-05, | |
| "loss": 0.0001, | |
| "step": 6775 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "learning_rate": 3.370526315789474e-05, | |
| "loss": 0.0001, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "learning_rate": 3.34421052631579e-05, | |
| "loss": 0.0001, | |
| "step": 6825 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "learning_rate": 3.317894736842106e-05, | |
| "loss": 0.0001, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "learning_rate": 3.291578947368421e-05, | |
| "loss": 0.0001, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "learning_rate": 3.265263157894737e-05, | |
| "loss": 0.0001, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 40.02, | |
| "learning_rate": 3.238947368421053e-05, | |
| "loss": 0.0004, | |
| "step": 6925 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "learning_rate": 3.212631578947369e-05, | |
| "loss": 0.0001, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "learning_rate": 3.1863157894736846e-05, | |
| "loss": 0.0001, | |
| "step": 6975 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "learning_rate": 3.16e-05, | |
| "loss": 0.0001, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "eval_loss": 0.4770253896713257, | |
| "eval_runtime": 310.7781, | |
| "eval_samples_per_second": 9.309, | |
| "eval_steps_per_second": 0.293, | |
| "eval_wer": 23.088075389633925, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "learning_rate": 3.133684210526316e-05, | |
| "loss": 0.0001, | |
| "step": 7025 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "learning_rate": 3.1073684210526315e-05, | |
| "loss": 0.0001, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "learning_rate": 3.081052631578947e-05, | |
| "loss": 0.0001, | |
| "step": 7075 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "learning_rate": 3.0547368421052635e-05, | |
| "loss": 0.0001, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "learning_rate": 3.028421052631579e-05, | |
| "loss": 0.0, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 42.01, | |
| "learning_rate": 3.0021052631578948e-05, | |
| "loss": 0.0001, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 42.01, | |
| "learning_rate": 2.9757894736842108e-05, | |
| "loss": 0.0001, | |
| "step": 7175 | |
| }, | |
| { | |
| "epoch": 42.01, | |
| "learning_rate": 2.9494736842105264e-05, | |
| "loss": 0.0001, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 42.01, | |
| "learning_rate": 2.923157894736842e-05, | |
| "loss": 0.0, | |
| "step": 7225 | |
| }, | |
| { | |
| "epoch": 42.02, | |
| "learning_rate": 2.8968421052631577e-05, | |
| "loss": 0.0001, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "learning_rate": 2.870526315789474e-05, | |
| "loss": 0.0001, | |
| "step": 7275 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "learning_rate": 2.8442105263157897e-05, | |
| "loss": 0.0, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 43.01, | |
| "learning_rate": 2.8178947368421054e-05, | |
| "loss": 0.0001, | |
| "step": 7325 | |
| }, | |
| { | |
| "epoch": 43.01, | |
| "learning_rate": 2.791578947368421e-05, | |
| "loss": 0.0, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 43.01, | |
| "learning_rate": 2.765263157894737e-05, | |
| "loss": 0.0001, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 43.01, | |
| "learning_rate": 2.7389473684210527e-05, | |
| "loss": 0.0, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 43.02, | |
| "learning_rate": 2.7126315789473683e-05, | |
| "loss": 0.0001, | |
| "step": 7425 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "learning_rate": 2.6863157894736846e-05, | |
| "loss": 0.0001, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "learning_rate": 2.6600000000000003e-05, | |
| "loss": 0.0, | |
| "step": 7475 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "learning_rate": 2.633684210526316e-05, | |
| "loss": 0.0, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "eval_loss": 0.4806711971759796, | |
| "eval_runtime": 316.5914, | |
| "eval_samples_per_second": 9.138, | |
| "eval_steps_per_second": 0.287, | |
| "eval_wer": 23.051830373323668, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "learning_rate": 2.6073684210526316e-05, | |
| "loss": 0.0, | |
| "step": 7525 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "learning_rate": 2.5810526315789472e-05, | |
| "loss": 0.0, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "learning_rate": 2.5547368421052632e-05, | |
| "loss": 0.0001, | |
| "step": 7575 | |
| }, | |
| { | |
| "epoch": 44.02, | |
| "learning_rate": 2.528421052631579e-05, | |
| "loss": 0.0001, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "learning_rate": 2.5021052631578952e-05, | |
| "loss": 0.0, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "learning_rate": 2.475789473684211e-05, | |
| "loss": 0.0, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 2.4494736842105265e-05, | |
| "loss": 0.0, | |
| "step": 7675 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 2.423157894736842e-05, | |
| "loss": 0.0, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 2.3968421052631578e-05, | |
| "loss": 0.0, | |
| "step": 7725 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 2.3705263157894738e-05, | |
| "loss": 0.0001, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "learning_rate": 2.3442105263157894e-05, | |
| "loss": 0.0001, | |
| "step": 7775 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "learning_rate": 2.3178947368421054e-05, | |
| "loss": 0.0, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 2.2915789473684214e-05, | |
| "loss": 0.0, | |
| "step": 7825 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 2.265263157894737e-05, | |
| "loss": 0.0, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 2.2389473684210527e-05, | |
| "loss": 0.0, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 2.2126315789473683e-05, | |
| "loss": 0.0, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 46.02, | |
| "learning_rate": 2.1863157894736843e-05, | |
| "loss": 0.0001, | |
| "step": 7925 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "learning_rate": 2.16e-05, | |
| "loss": 0.0001, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "learning_rate": 2.1336842105263156e-05, | |
| "loss": 0.0, | |
| "step": 7975 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 2.1073684210526316e-05, | |
| "loss": 0.0, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "eval_loss": 0.4834863841533661, | |
| "eval_runtime": 307.4455, | |
| "eval_samples_per_second": 9.41, | |
| "eval_steps_per_second": 0.296, | |
| "eval_wer": 22.961217832548027, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 2.0810526315789476e-05, | |
| "loss": 0.0, | |
| "step": 8025 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 2.0547368421052633e-05, | |
| "loss": 0.0, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 2.0284210526315792e-05, | |
| "loss": 0.0, | |
| "step": 8075 | |
| }, | |
| { | |
| "epoch": 47.02, | |
| "learning_rate": 2.002105263157895e-05, | |
| "loss": 0.0001, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "learning_rate": 1.9757894736842105e-05, | |
| "loss": 0.0, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "learning_rate": 1.9494736842105262e-05, | |
| "loss": 0.0, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 1.9231578947368422e-05, | |
| "loss": 0.0, | |
| "step": 8175 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 1.896842105263158e-05, | |
| "loss": 0.0, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 1.8705263157894738e-05, | |
| "loss": 0.0, | |
| "step": 8225 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 1.8442105263157898e-05, | |
| "loss": 0.0, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 48.02, | |
| "learning_rate": 1.8178947368421055e-05, | |
| "loss": 0.0001, | |
| "step": 8275 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "learning_rate": 1.791578947368421e-05, | |
| "loss": 0.0, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "learning_rate": 1.765263157894737e-05, | |
| "loss": 0.0, | |
| "step": 8325 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 1.7389473684210527e-05, | |
| "loss": 0.0, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 1.7126315789473684e-05, | |
| "loss": 0.0, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 1.686315789473684e-05, | |
| "loss": 0.0, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 1.66e-05, | |
| "loss": 0.0001, | |
| "step": 8425 | |
| }, | |
| { | |
| "epoch": 49.02, | |
| "learning_rate": 1.633684210526316e-05, | |
| "loss": 0.0001, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 50.0, | |
| "learning_rate": 1.6073684210526317e-05, | |
| "loss": 0.0, | |
| "step": 8475 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "learning_rate": 1.5810526315789477e-05, | |
| "loss": 0.0, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "eval_loss": 0.4857272207736969, | |
| "eval_runtime": 309.4149, | |
| "eval_samples_per_second": 9.35, | |
| "eval_steps_per_second": 0.294, | |
| "eval_wer": 22.92497281623777, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "learning_rate": 1.5547368421052633e-05, | |
| "loss": 0.0, | |
| "step": 8525 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "learning_rate": 1.528421052631579e-05, | |
| "loss": 0.0, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "learning_rate": 1.5021052631578946e-05, | |
| "loss": 0.0, | |
| "step": 8575 | |
| }, | |
| { | |
| "epoch": 50.02, | |
| "learning_rate": 1.4757894736842106e-05, | |
| "loss": 0.0001, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "learning_rate": 1.4494736842105264e-05, | |
| "loss": 0.0001, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "learning_rate": 1.423157894736842e-05, | |
| "loss": 0.0, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 51.01, | |
| "learning_rate": 1.396842105263158e-05, | |
| "loss": 0.0, | |
| "step": 8675 | |
| }, | |
| { | |
| "epoch": 51.01, | |
| "learning_rate": 1.3705263157894737e-05, | |
| "loss": 0.0, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 51.01, | |
| "learning_rate": 1.3442105263157895e-05, | |
| "loss": 0.0, | |
| "step": 8725 | |
| }, | |
| { | |
| "epoch": 51.01, | |
| "learning_rate": 1.3178947368421055e-05, | |
| "loss": 0.0, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 51.02, | |
| "learning_rate": 1.2915789473684212e-05, | |
| "loss": 0.0001, | |
| "step": 8775 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "learning_rate": 1.2652631578947368e-05, | |
| "loss": 0.0, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "learning_rate": 1.2389473684210528e-05, | |
| "loss": 0.0, | |
| "step": 8825 | |
| }, | |
| { | |
| "epoch": 52.01, | |
| "learning_rate": 1.2126315789473684e-05, | |
| "loss": 0.0, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 52.01, | |
| "learning_rate": 1.1863157894736843e-05, | |
| "loss": 0.0, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 52.01, | |
| "learning_rate": 1.16e-05, | |
| "loss": 0.0, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 52.01, | |
| "learning_rate": 1.1336842105263159e-05, | |
| "loss": 0.0, | |
| "step": 8925 | |
| }, | |
| { | |
| "epoch": 52.02, | |
| "learning_rate": 1.1073684210526317e-05, | |
| "loss": 0.0001, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "learning_rate": 1.0810526315789474e-05, | |
| "loss": 0.0, | |
| "step": 8975 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "learning_rate": 1.0547368421052632e-05, | |
| "loss": 0.0, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "eval_loss": 0.48744601011276245, | |
| "eval_runtime": 314.152, | |
| "eval_samples_per_second": 9.209, | |
| "eval_steps_per_second": 0.29, | |
| "eval_wer": 22.90685030808264, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 53.01, | |
| "learning_rate": 1.028421052631579e-05, | |
| "loss": 0.0, | |
| "step": 9025 | |
| }, | |
| { | |
| "epoch": 53.01, | |
| "learning_rate": 1.0021052631578948e-05, | |
| "loss": 0.0, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 53.01, | |
| "learning_rate": 9.757894736842106e-06, | |
| "loss": 0.0, | |
| "step": 9075 | |
| }, | |
| { | |
| "epoch": 53.01, | |
| "learning_rate": 9.494736842105263e-06, | |
| "loss": 0.0, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 53.02, | |
| "learning_rate": 9.231578947368421e-06, | |
| "loss": 0.0001, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "learning_rate": 8.96842105263158e-06, | |
| "loss": 0.0, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "learning_rate": 8.705263157894737e-06, | |
| "loss": 0.0, | |
| "step": 9175 | |
| }, | |
| { | |
| "epoch": 54.01, | |
| "learning_rate": 8.442105263157896e-06, | |
| "loss": 0.0, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 54.01, | |
| "learning_rate": 8.178947368421052e-06, | |
| "loss": 0.0, | |
| "step": 9225 | |
| }, | |
| { | |
| "epoch": 54.01, | |
| "learning_rate": 7.915789473684212e-06, | |
| "loss": 0.0, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 54.01, | |
| "learning_rate": 7.652631578947368e-06, | |
| "loss": 0.0001, | |
| "step": 9275 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "learning_rate": 7.389473684210527e-06, | |
| "loss": 0.0, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "learning_rate": 7.126315789473685e-06, | |
| "loss": 0.0, | |
| "step": 9325 | |
| }, | |
| { | |
| "epoch": 55.01, | |
| "learning_rate": 6.863157894736842e-06, | |
| "loss": 0.0, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 55.01, | |
| "learning_rate": 6.6e-06, | |
| "loss": 0.0, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 55.01, | |
| "learning_rate": 6.336842105263158e-06, | |
| "loss": 0.0, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 55.01, | |
| "learning_rate": 6.073684210526316e-06, | |
| "loss": 0.0, | |
| "step": 9425 | |
| }, | |
| { | |
| "epoch": 55.02, | |
| "learning_rate": 5.810526315789474e-06, | |
| "loss": 0.0001, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "learning_rate": 5.547368421052631e-06, | |
| "loss": 0.0, | |
| "step": 9475 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "learning_rate": 5.2842105263157896e-06, | |
| "loss": 0.0, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "eval_loss": 0.48867565393447876, | |
| "eval_runtime": 310.2558, | |
| "eval_samples_per_second": 9.325, | |
| "eval_steps_per_second": 0.293, | |
| "eval_wer": 22.90685030808264, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 56.01, | |
| "learning_rate": 5.021052631578948e-06, | |
| "loss": 0.0, | |
| "step": 9525 | |
| }, | |
| { | |
| "epoch": 56.01, | |
| "learning_rate": 4.757894736842106e-06, | |
| "loss": 0.0, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 56.01, | |
| "learning_rate": 4.494736842105263e-06, | |
| "loss": 0.0, | |
| "step": 9575 | |
| }, | |
| { | |
| "epoch": 56.01, | |
| "learning_rate": 4.2315789473684215e-06, | |
| "loss": 0.0, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 56.02, | |
| "learning_rate": 3.968421052631579e-06, | |
| "loss": 0.0001, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "learning_rate": 3.7052631578947374e-06, | |
| "loss": 0.0, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "learning_rate": 3.442105263157895e-06, | |
| "loss": 0.0, | |
| "step": 9675 | |
| }, | |
| { | |
| "epoch": 57.01, | |
| "learning_rate": 3.178947368421053e-06, | |
| "loss": 0.0, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 57.01, | |
| "learning_rate": 2.9157894736842107e-06, | |
| "loss": 0.0, | |
| "step": 9725 | |
| }, | |
| { | |
| "epoch": 57.01, | |
| "learning_rate": 2.6526315789473685e-06, | |
| "loss": 0.0, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 57.01, | |
| "learning_rate": 2.3894736842105266e-06, | |
| "loss": 0.0, | |
| "step": 9775 | |
| }, | |
| { | |
| "epoch": 57.02, | |
| "learning_rate": 2.1263157894736844e-06, | |
| "loss": 0.0001, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "learning_rate": 1.8631578947368424e-06, | |
| "loss": 0.0, | |
| "step": 9825 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "learning_rate": 1.6000000000000001e-06, | |
| "loss": 0.0, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 58.01, | |
| "learning_rate": 1.3368421052631581e-06, | |
| "loss": 0.0, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 58.01, | |
| "learning_rate": 1.0736842105263159e-06, | |
| "loss": 0.0, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 58.01, | |
| "learning_rate": 8.105263157894737e-07, | |
| "loss": 0.0, | |
| "step": 9925 | |
| }, | |
| { | |
| "epoch": 58.01, | |
| "learning_rate": 5.473684210526316e-07, | |
| "loss": 0.0001, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "learning_rate": 2.8421052631578953e-07, | |
| "loss": 0.0, | |
| "step": 9975 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "learning_rate": 2.105263157894737e-08, | |
| "loss": 0.0, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "eval_loss": 0.48912885785102844, | |
| "eval_runtime": 314.605, | |
| "eval_samples_per_second": 9.196, | |
| "eval_steps_per_second": 0.289, | |
| "eval_wer": 22.87060529177238, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "step": 10000, | |
| "total_flos": 1.84268992739328e+20, | |
| "train_loss": 0.03550489917879458, | |
| "train_runtime": 42960.0867, | |
| "train_samples_per_second": 14.898, | |
| "train_steps_per_second": 0.233 | |
| } | |
| ], | |
| "max_steps": 10000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 1.84268992739328e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |