| { | |
| "best_metric": 14.674060048688126, | |
| "best_model_checkpoint": "./checkpoint-1000", | |
| "epoch": 199.005, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.6000000000000004e-07, | |
| "loss": 1.272, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.600000000000001e-07, | |
| "loss": 0.8924, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.46e-06, | |
| "loss": 0.5164, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 1.9600000000000003e-06, | |
| "loss": 0.3132, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 2.46e-06, | |
| "loss": 0.1865, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 2.96e-06, | |
| "loss": 0.1016, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 3.46e-06, | |
| "loss": 0.0495, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 3.96e-06, | |
| "loss": 0.022, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 4.4600000000000005e-06, | |
| "loss": 0.0105, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 4.960000000000001e-06, | |
| "loss": 0.0059, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 5.460000000000001e-06, | |
| "loss": 0.0047, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.9600000000000005e-06, | |
| "loss": 0.0034, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 6.460000000000001e-06, | |
| "loss": 0.0028, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 6.96e-06, | |
| "loss": 0.0029, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 7.4600000000000006e-06, | |
| "loss": 0.0018, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 7.960000000000002e-06, | |
| "loss": 0.0015, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 8.46e-06, | |
| "loss": 0.0012, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 8.96e-06, | |
| "loss": 0.001, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 9.460000000000001e-06, | |
| "loss": 0.0008, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 9.960000000000001e-06, | |
| "loss": 0.0007, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 9.94888888888889e-06, | |
| "loss": 0.0007, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 9.893333333333334e-06, | |
| "loss": 0.0006, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 9.837777777777778e-06, | |
| "loss": 0.0005, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 9.782222222222222e-06, | |
| "loss": 0.0005, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 9.726666666666668e-06, | |
| "loss": 0.0005, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "learning_rate": 9.671111111111112e-06, | |
| "loss": 0.0004, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "learning_rate": 9.615555555555558e-06, | |
| "loss": 0.0004, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "learning_rate": 9.56e-06, | |
| "loss": 0.0004, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "learning_rate": 9.504444444444446e-06, | |
| "loss": 0.0004, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "learning_rate": 9.44888888888889e-06, | |
| "loss": 0.0003, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "learning_rate": 9.393333333333334e-06, | |
| "loss": 0.0003, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "learning_rate": 9.33777777777778e-06, | |
| "loss": 0.0003, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "learning_rate": 9.282222222222222e-06, | |
| "loss": 0.0003, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 33.01, | |
| "learning_rate": 9.226666666666668e-06, | |
| "loss": 0.0003, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 34.01, | |
| "learning_rate": 9.171111111111112e-06, | |
| "loss": 0.0003, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "learning_rate": 9.115555555555556e-06, | |
| "loss": 0.0003, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 36.01, | |
| "learning_rate": 9.060000000000001e-06, | |
| "loss": 0.0002, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 37.01, | |
| "learning_rate": 9.004444444444445e-06, | |
| "loss": 0.0002, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "learning_rate": 8.94888888888889e-06, | |
| "loss": 0.0002, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "learning_rate": 8.893333333333333e-06, | |
| "loss": 0.0002, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "eval_loss": 0.4552850127220154, | |
| "eval_runtime": 1454.4803, | |
| "eval_samples_per_second": 0.637, | |
| "eval_steps_per_second": 0.02, | |
| "eval_wer": 14.674060048688126, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "learning_rate": 8.83777777777778e-06, | |
| "loss": 0.0002, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "learning_rate": 8.782222222222223e-06, | |
| "loss": 0.0002, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 42.01, | |
| "learning_rate": 8.726666666666667e-06, | |
| "loss": 0.0002, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 43.01, | |
| "learning_rate": 8.671111111111113e-06, | |
| "loss": 0.0002, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "learning_rate": 8.615555555555555e-06, | |
| "loss": 0.0002, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 8.560000000000001e-06, | |
| "loss": 0.0002, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 8.504444444444445e-06, | |
| "loss": 0.0002, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 8.448888888888889e-06, | |
| "loss": 0.0002, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 8.393333333333335e-06, | |
| "loss": 0.0002, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 8.337777777777777e-06, | |
| "loss": 0.0002, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "learning_rate": 8.282222222222223e-06, | |
| "loss": 0.0002, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 51.01, | |
| "learning_rate": 8.226666666666667e-06, | |
| "loss": 0.0002, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 52.01, | |
| "learning_rate": 8.171111111111113e-06, | |
| "loss": 0.0002, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 53.01, | |
| "learning_rate": 8.115555555555557e-06, | |
| "loss": 0.0001, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 54.01, | |
| "learning_rate": 8.06e-06, | |
| "loss": 0.0001, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 55.01, | |
| "learning_rate": 8.004444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 56.01, | |
| "learning_rate": 7.948888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 57.01, | |
| "learning_rate": 7.893333333333335e-06, | |
| "loss": 0.0001, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 58.01, | |
| "learning_rate": 7.837777777777779e-06, | |
| "loss": 0.0001, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 59.01, | |
| "learning_rate": 7.782222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 60.01, | |
| "learning_rate": 7.726666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 61.01, | |
| "learning_rate": 7.67111111111111e-06, | |
| "loss": 0.0001, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 62.01, | |
| "learning_rate": 7.6155555555555564e-06, | |
| "loss": 0.0001, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 63.01, | |
| "learning_rate": 7.5600000000000005e-06, | |
| "loss": 0.0001, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 64.0, | |
| "learning_rate": 7.504444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 65.0, | |
| "learning_rate": 7.44888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 66.0, | |
| "learning_rate": 7.393333333333333e-06, | |
| "loss": 0.0001, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 67.0, | |
| "learning_rate": 7.337777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "learning_rate": 7.282222222222222e-06, | |
| "loss": 0.0001, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 69.0, | |
| "learning_rate": 7.226666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 70.0, | |
| "learning_rate": 7.171111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 71.0, | |
| "learning_rate": 7.115555555555557e-06, | |
| "loss": 0.0001, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 72.0, | |
| "learning_rate": 7.06e-06, | |
| "loss": 0.0001, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 73.0, | |
| "learning_rate": 7.004444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "learning_rate": 6.948888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 75.0, | |
| "learning_rate": 6.893333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 76.0, | |
| "learning_rate": 6.837777777777779e-06, | |
| "loss": 0.0001, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 77.0, | |
| "learning_rate": 6.782222222222222e-06, | |
| "loss": 0.0001, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "learning_rate": 6.726666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "learning_rate": 6.671111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "eval_loss": 0.5023436546325684, | |
| "eval_runtime": 1459.7613, | |
| "eval_samples_per_second": 0.635, | |
| "eval_steps_per_second": 0.02, | |
| "eval_wer": 14.940041475069876, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "learning_rate": 6.615555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 81.0, | |
| "learning_rate": 6.560000000000001e-06, | |
| "loss": 0.0001, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 82.0, | |
| "learning_rate": 6.504444444444446e-06, | |
| "loss": 0.0001, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 83.0, | |
| "learning_rate": 6.448888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 84.0, | |
| "learning_rate": 6.393333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 85.0, | |
| "learning_rate": 6.3377777777777786e-06, | |
| "loss": 0.0001, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 86.0, | |
| "learning_rate": 6.282222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 87.0, | |
| "learning_rate": 6.2266666666666675e-06, | |
| "loss": 0.0001, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 88.0, | |
| "learning_rate": 6.171111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 89.0, | |
| "learning_rate": 6.1155555555555555e-06, | |
| "loss": 0.0001, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 90.0, | |
| "learning_rate": 6.0600000000000004e-06, | |
| "loss": 0.0001, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 91.0, | |
| "learning_rate": 6.004444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 92.0, | |
| "learning_rate": 5.948888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 93.0, | |
| "learning_rate": 5.893333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 94.0, | |
| "learning_rate": 5.837777777777777e-06, | |
| "loss": 0.0001, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 95.0, | |
| "learning_rate": 5.782222222222222e-06, | |
| "loss": 0.0001, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 96.0, | |
| "learning_rate": 5.726666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 97.0, | |
| "learning_rate": 5.671111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 98.0, | |
| "learning_rate": 5.615555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 99.0, | |
| "learning_rate": 5.560000000000001e-06, | |
| "loss": 0.0001, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 100.0, | |
| "learning_rate": 5.504444444444444e-06, | |
| "loss": 0.0001, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 101.0, | |
| "learning_rate": 5.448888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 102.0, | |
| "learning_rate": 5.393333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 103.0, | |
| "learning_rate": 5.337777777777779e-06, | |
| "loss": 0.0001, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 104.0, | |
| "learning_rate": 5.282222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 105.0, | |
| "learning_rate": 5.226666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 106.0, | |
| "learning_rate": 5.171111111111111e-06, | |
| "loss": 0.0001, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 107.0, | |
| "learning_rate": 5.115555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 108.0, | |
| "learning_rate": 5.060000000000001e-06, | |
| "loss": 0.0001, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 109.0, | |
| "learning_rate": 5.004444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 110.0, | |
| "learning_rate": 4.94888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 111.0, | |
| "learning_rate": 4.893333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 112.0, | |
| "learning_rate": 4.837777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 113.0, | |
| "learning_rate": 4.7822222222222226e-06, | |
| "loss": 0.0, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 114.0, | |
| "learning_rate": 4.7266666666666674e-06, | |
| "loss": 0.0, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 115.0, | |
| "learning_rate": 4.6711111111111115e-06, | |
| "loss": 0.0, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 116.0, | |
| "learning_rate": 4.6155555555555555e-06, | |
| "loss": 0.0, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 117.0, | |
| "learning_rate": 4.56e-06, | |
| "loss": 0.0, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 118.0, | |
| "learning_rate": 4.504444444444444e-06, | |
| "loss": 0.0, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 119.0, | |
| "learning_rate": 4.448888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 119.0, | |
| "eval_loss": 0.5316691994667053, | |
| "eval_runtime": 1455.451, | |
| "eval_samples_per_second": 0.637, | |
| "eval_steps_per_second": 0.02, | |
| "eval_wer": 15.160941303759806, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 120.0, | |
| "learning_rate": 4.393333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 121.0, | |
| "learning_rate": 4.337777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 122.0, | |
| "learning_rate": 4.282222222222222e-06, | |
| "loss": 0.0, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 123.0, | |
| "learning_rate": 4.226666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 124.0, | |
| "learning_rate": 4.171111111111111e-06, | |
| "loss": 0.0, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 125.0, | |
| "learning_rate": 4.115555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 126.0, | |
| "learning_rate": 4.060000000000001e-06, | |
| "loss": 0.0, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 127.0, | |
| "learning_rate": 4.004444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 128.0, | |
| "learning_rate": 3.948888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 129.0, | |
| "learning_rate": 3.893333333333333e-06, | |
| "loss": 0.0, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 130.0, | |
| "learning_rate": 3.837777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 131.0, | |
| "learning_rate": 3.782222222222223e-06, | |
| "loss": 0.0, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 132.0, | |
| "learning_rate": 3.726666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 133.0, | |
| "learning_rate": 3.6711111111111113e-06, | |
| "loss": 0.0, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 134.0, | |
| "learning_rate": 3.615555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 135.0, | |
| "learning_rate": 3.5600000000000002e-06, | |
| "loss": 0.0, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 136.0, | |
| "learning_rate": 3.5044444444444447e-06, | |
| "loss": 0.0, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 137.0, | |
| "learning_rate": 3.4488888888888896e-06, | |
| "loss": 0.0, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 138.0, | |
| "learning_rate": 3.3933333333333336e-06, | |
| "loss": 0.0, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 139.0, | |
| "learning_rate": 3.337777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 140.0, | |
| "learning_rate": 3.282222222222223e-06, | |
| "loss": 0.0, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 141.0, | |
| "learning_rate": 3.226666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 142.0, | |
| "learning_rate": 3.1711111111111114e-06, | |
| "loss": 0.0, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 143.0, | |
| "learning_rate": 3.1155555555555555e-06, | |
| "loss": 0.0, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 144.0, | |
| "learning_rate": 3.0600000000000003e-06, | |
| "loss": 0.0, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 145.0, | |
| "learning_rate": 3.004444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 146.0, | |
| "learning_rate": 2.948888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 147.0, | |
| "learning_rate": 2.8933333333333337e-06, | |
| "loss": 0.0, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 148.0, | |
| "learning_rate": 2.837777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 149.0, | |
| "learning_rate": 2.7822222222222222e-06, | |
| "loss": 0.0, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 150.0, | |
| "learning_rate": 2.726666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 151.0, | |
| "learning_rate": 2.6711111111111116e-06, | |
| "loss": 0.0, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 152.0, | |
| "learning_rate": 2.6155555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 153.0, | |
| "learning_rate": 2.56e-06, | |
| "loss": 0.0, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 154.0, | |
| "learning_rate": 2.504444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 155.0, | |
| "learning_rate": 2.448888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 156.0, | |
| "learning_rate": 2.3933333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 157.0, | |
| "learning_rate": 2.337777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 158.0, | |
| "learning_rate": 2.2822222222222223e-06, | |
| "loss": 0.0, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 159.0, | |
| "learning_rate": 2.226666666666667e-06, | |
| "loss": 0.0, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 159.0, | |
| "eval_loss": 0.5512966513633728, | |
| "eval_runtime": 1459.1411, | |
| "eval_samples_per_second": 0.635, | |
| "eval_steps_per_second": 0.02, | |
| "eval_wer": 15.201514741682445, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 160.0, | |
| "learning_rate": 2.1711111111111113e-06, | |
| "loss": 0.0, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 161.0, | |
| "learning_rate": 2.1155555555555557e-06, | |
| "loss": 0.0, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 162.0, | |
| "learning_rate": 2.06e-06, | |
| "loss": 0.0, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 163.0, | |
| "learning_rate": 2.0044444444444446e-06, | |
| "loss": 0.0, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 164.0, | |
| "learning_rate": 1.948888888888889e-06, | |
| "loss": 0.0, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 165.0, | |
| "learning_rate": 1.8933333333333333e-06, | |
| "loss": 0.0, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 166.0, | |
| "learning_rate": 1.837777777777778e-06, | |
| "loss": 0.0, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 167.0, | |
| "learning_rate": 1.7822222222222225e-06, | |
| "loss": 0.0, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 168.0, | |
| "learning_rate": 1.7266666666666667e-06, | |
| "loss": 0.0, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 169.0, | |
| "learning_rate": 1.6711111111111112e-06, | |
| "loss": 0.0, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 170.0, | |
| "learning_rate": 1.6155555555555559e-06, | |
| "loss": 0.0, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 171.0, | |
| "learning_rate": 1.56e-06, | |
| "loss": 0.0, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 172.0, | |
| "learning_rate": 1.5044444444444446e-06, | |
| "loss": 0.0, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 173.0, | |
| "learning_rate": 1.4488888888888892e-06, | |
| "loss": 0.0, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 174.0, | |
| "learning_rate": 1.3933333333333335e-06, | |
| "loss": 0.0, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 175.0, | |
| "learning_rate": 1.337777777777778e-06, | |
| "loss": 0.0, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 176.0, | |
| "learning_rate": 1.2822222222222222e-06, | |
| "loss": 0.0, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 177.0, | |
| "learning_rate": 1.2266666666666666e-06, | |
| "loss": 0.0, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 178.0, | |
| "learning_rate": 1.171111111111111e-06, | |
| "loss": 0.0, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 179.0, | |
| "learning_rate": 1.1155555555555558e-06, | |
| "loss": 0.0, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 180.0, | |
| "learning_rate": 1.06e-06, | |
| "loss": 0.0, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 181.0, | |
| "learning_rate": 1.0044444444444445e-06, | |
| "loss": 0.0, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 182.0, | |
| "learning_rate": 9.488888888888889e-07, | |
| "loss": 0.0, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 183.0, | |
| "learning_rate": 8.933333333333334e-07, | |
| "loss": 0.0, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 184.0, | |
| "learning_rate": 8.37777777777778e-07, | |
| "loss": 0.0, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 185.0, | |
| "learning_rate": 7.822222222222223e-07, | |
| "loss": 0.0, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 186.0, | |
| "learning_rate": 7.266666666666668e-07, | |
| "loss": 0.0, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 187.0, | |
| "learning_rate": 6.711111111111111e-07, | |
| "loss": 0.0, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 188.0, | |
| "learning_rate": 6.155555555555556e-07, | |
| "loss": 0.0, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 189.0, | |
| "learning_rate": 5.6e-07, | |
| "loss": 0.0, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 190.0, | |
| "learning_rate": 5.044444444444445e-07, | |
| "loss": 0.0, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 191.0, | |
| "learning_rate": 4.488888888888889e-07, | |
| "loss": 0.0, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 192.0, | |
| "learning_rate": 3.9333333333333336e-07, | |
| "loss": 0.0, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 193.0, | |
| "learning_rate": 3.3777777777777777e-07, | |
| "loss": 0.0, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 194.0, | |
| "learning_rate": 2.822222222222222e-07, | |
| "loss": 0.0, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 195.0, | |
| "learning_rate": 2.266666666666667e-07, | |
| "loss": 0.0, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 196.0, | |
| "learning_rate": 1.7111111111111114e-07, | |
| "loss": 0.0, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 197.0, | |
| "learning_rate": 1.1555555555555556e-07, | |
| "loss": 0.0, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 198.0, | |
| "learning_rate": 6.000000000000001e-08, | |
| "loss": 0.0, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "learning_rate": 4.444444444444445e-09, | |
| "loss": 0.0, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "eval_loss": 0.5593463182449341, | |
| "eval_runtime": 1455.1907, | |
| "eval_samples_per_second": 0.637, | |
| "eval_steps_per_second": 0.02, | |
| "eval_wer": 15.206022901451627, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "step": 5000, | |
| "total_flos": 1.6125557538816e+20, | |
| "train_loss": 0.017043720645061693, | |
| "train_runtime": 46777.3097, | |
| "train_samples_per_second": 3.42, | |
| "train_steps_per_second": 0.107 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 1.6125557538816e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |