| { | |
| "best_metric": 32.955865272938446, | |
| "best_model_checkpoint": "./checkpoint-7000", | |
| "epoch": 8.02742857142857, | |
| "global_step": 7000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 6.5745, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 4.839, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 3.9088, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 3.3404, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 2.9648, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.9e-06, | |
| "loss": 2.5936, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 2.2292, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 1.8805, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.4e-06, | |
| "loss": 1.6378, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 1.3965, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 1.3637, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 5.9e-06, | |
| "loss": 1.2695, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 1.1668, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 6.9e-06, | |
| "loss": 1.0783, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 7.4e-06, | |
| "loss": 1.0515, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.9765, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.848, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.8255, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.8107, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.7365, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.96923076923077e-06, | |
| "loss": 0.663, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.930769230769232e-06, | |
| "loss": 0.6438, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.892307692307693e-06, | |
| "loss": 0.5998, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.853846153846155e-06, | |
| "loss": 0.592, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.815384615384617e-06, | |
| "loss": 0.582, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.776923076923078e-06, | |
| "loss": 0.4367, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.738461538461538e-06, | |
| "loss": 0.4615, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.7e-06, | |
| "loss": 0.4568, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.661538461538462e-06, | |
| "loss": 0.4175, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.623076923076925e-06, | |
| "loss": 0.4241, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.584615384615385e-06, | |
| "loss": 0.3883, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.546153846153847e-06, | |
| "loss": 0.3812, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.50769230769231e-06, | |
| "loss": 0.348, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.46923076923077e-06, | |
| "loss": 0.3327, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.43076923076923e-06, | |
| "loss": 0.1867, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.392307692307693e-06, | |
| "loss": 0.2568, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.353846153846155e-06, | |
| "loss": 0.2686, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.315384615384617e-06, | |
| "loss": 0.2308, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.276923076923077e-06, | |
| "loss": 0.2571, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.23846153846154e-06, | |
| "loss": 0.2448, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "eval_loss": 0.8498038649559021, | |
| "eval_runtime": 169.1225, | |
| "eval_samples_per_second": 3.039, | |
| "eval_steps_per_second": 0.384, | |
| "eval_wer": 41.753774680603954, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.200000000000002e-06, | |
| "loss": 0.242, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.161538461538462e-06, | |
| "loss": 0.2153, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.123076923076923e-06, | |
| "loss": 0.2002, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.084615384615385e-06, | |
| "loss": 0.1988, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.046153846153847e-06, | |
| "loss": 0.1947, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.007692307692309e-06, | |
| "loss": 0.1906, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.96923076923077e-06, | |
| "loss": 0.1808, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.930769230769232e-06, | |
| "loss": 0.1715, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 8.892307692307694e-06, | |
| "loss": 0.1545, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 8.853846153846154e-06, | |
| "loss": 0.1664, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 8.815384615384615e-06, | |
| "loss": 0.1048, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 8.776923076923077e-06, | |
| "loss": 0.1138, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.73846153846154e-06, | |
| "loss": 0.1086, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.700000000000001e-06, | |
| "loss": 0.1088, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 8.661538461538462e-06, | |
| "loss": 0.0997, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 8.623076923076924e-06, | |
| "loss": 0.1017, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 8.584615384615386e-06, | |
| "loss": 0.09, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 8.546153846153847e-06, | |
| "loss": 0.0917, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 8.507692307692307e-06, | |
| "loss": 0.0742, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 8.46923076923077e-06, | |
| "loss": 0.0847, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 8.430769230769231e-06, | |
| "loss": 0.0739, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 8.392307692307694e-06, | |
| "loss": 0.0736, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 8.353846153846154e-06, | |
| "loss": 0.0692, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 8.315384615384616e-06, | |
| "loss": 0.0804, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 8.276923076923078e-06, | |
| "loss": 0.0581, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 8.238461538461539e-06, | |
| "loss": 0.068, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 8.2e-06, | |
| "loss": 0.0483, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 8.161538461538462e-06, | |
| "loss": 0.0489, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 8.123076923076924e-06, | |
| "loss": 0.036, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8.084615384615386e-06, | |
| "loss": 0.0334, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8.046153846153846e-06, | |
| "loss": 0.0407, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 8.007692307692309e-06, | |
| "loss": 0.0412, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.96923076923077e-06, | |
| "loss": 0.0479, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.930769230769231e-06, | |
| "loss": 0.0453, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 7.892307692307692e-06, | |
| "loss": 0.0411, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 7.853846153846154e-06, | |
| "loss": 0.0369, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 7.815384615384616e-06, | |
| "loss": 0.0399, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 7.776923076923078e-06, | |
| "loss": 0.0354, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 7.738461538461539e-06, | |
| "loss": 0.0366, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 7.7e-06, | |
| "loss": 0.0367, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "eval_loss": 0.8609296679496765, | |
| "eval_runtime": 162.7001, | |
| "eval_samples_per_second": 3.159, | |
| "eval_steps_per_second": 0.4, | |
| "eval_wer": 35.77235772357724, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 7.661538461538463e-06, | |
| "loss": 0.0355, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 7.623076923076924e-06, | |
| "loss": 0.0343, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 7.584615384615385e-06, | |
| "loss": 0.0256, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 7.546153846153846e-06, | |
| "loss": 0.0254, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 7.507692307692308e-06, | |
| "loss": 0.0243, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 7.46923076923077e-06, | |
| "loss": 0.0221, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 7.430769230769232e-06, | |
| "loss": 0.0282, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 7.392307692307693e-06, | |
| "loss": 0.0203, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 7.353846153846154e-06, | |
| "loss": 0.0284, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 7.3153846153846166e-06, | |
| "loss": 0.0208, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 7.276923076923077e-06, | |
| "loss": 0.0235, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 7.238461538461538e-06, | |
| "loss": 0.0283, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 7.2000000000000005e-06, | |
| "loss": 0.0179, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 7.161538461538462e-06, | |
| "loss": 0.0219, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 7.123076923076924e-06, | |
| "loss": 0.017, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 7.084615384615385e-06, | |
| "loss": 0.0154, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 7.046153846153847e-06, | |
| "loss": 0.0222, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 7.007692307692309e-06, | |
| "loss": 0.0178, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 6.96923076923077e-06, | |
| "loss": 0.0204, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 6.930769230769231e-06, | |
| "loss": 0.0216, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.892307692307693e-06, | |
| "loss": 0.0158, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 6.853846153846154e-06, | |
| "loss": 0.0149, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 6.815384615384616e-06, | |
| "loss": 0.0124, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 6.776923076923078e-06, | |
| "loss": 0.0156, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 6.738461538461539e-06, | |
| "loss": 0.0127, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 6.700000000000001e-06, | |
| "loss": 0.0164, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 6.661538461538462e-06, | |
| "loss": 0.0175, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 6.623076923076923e-06, | |
| "loss": 0.0138, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 6.584615384615385e-06, | |
| "loss": 0.0186, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 6.546153846153846e-06, | |
| "loss": 0.0125, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 6.5076923076923085e-06, | |
| "loss": 0.0127, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 6.46923076923077e-06, | |
| "loss": 0.0154, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 6.430769230769231e-06, | |
| "loss": 0.014, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 6.392307692307693e-06, | |
| "loss": 0.0128, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 6.353846153846155e-06, | |
| "loss": 0.0139, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 6.315384615384615e-06, | |
| "loss": 0.0133, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 6.276923076923077e-06, | |
| "loss": 0.01, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 6.238461538461539e-06, | |
| "loss": 0.0108, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 6.200000000000001e-06, | |
| "loss": 0.011, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 6.161538461538462e-06, | |
| "loss": 0.0095, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "eval_loss": 0.9109113812446594, | |
| "eval_runtime": 166.4325, | |
| "eval_samples_per_second": 3.088, | |
| "eval_steps_per_second": 0.391, | |
| "eval_wer": 34.930313588850176, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 6.123076923076923e-06, | |
| "loss": 0.0138, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 6.084615384615386e-06, | |
| "loss": 0.0111, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 6.046153846153847e-06, | |
| "loss": 0.0165, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 6.007692307692307e-06, | |
| "loss": 0.0104, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 5.9692307692307695e-06, | |
| "loss": 0.0091, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 5.930769230769231e-06, | |
| "loss": 0.0078, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 5.892307692307693e-06, | |
| "loss": 0.0073, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 5.853846153846154e-06, | |
| "loss": 0.0075, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 5.815384615384616e-06, | |
| "loss": 0.01, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 5.776923076923078e-06, | |
| "loss": 0.0095, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 5.738461538461539e-06, | |
| "loss": 0.0072, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 5.7e-06, | |
| "loss": 0.0063, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 5.661538461538462e-06, | |
| "loss": 0.0104, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 5.623076923076923e-06, | |
| "loss": 0.0077, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 5.584615384615385e-06, | |
| "loss": 0.0072, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 5.546153846153847e-06, | |
| "loss": 0.0074, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 5.507692307692308e-06, | |
| "loss": 0.0039, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 5.46923076923077e-06, | |
| "loss": 0.0049, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 5.430769230769231e-06, | |
| "loss": 0.0062, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 5.392307692307692e-06, | |
| "loss": 0.0035, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 5.353846153846154e-06, | |
| "loss": 0.0074, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 5.315384615384615e-06, | |
| "loss": 0.0079, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 5.2769230769230775e-06, | |
| "loss": 0.0061, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 5.238461538461539e-06, | |
| "loss": 0.0073, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 5.2e-06, | |
| "loss": 0.0067, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 5.161538461538462e-06, | |
| "loss": 0.0056, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 5.123076923076924e-06, | |
| "loss": 0.0056, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 5.084615384615384e-06, | |
| "loss": 0.0057, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 5.046153846153846e-06, | |
| "loss": 0.0057, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 5.007692307692308e-06, | |
| "loss": 0.0041, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 4.96923076923077e-06, | |
| "loss": 0.004, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 4.930769230769231e-06, | |
| "loss": 0.0066, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 4.892307692307693e-06, | |
| "loss": 0.0047, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 4.853846153846154e-06, | |
| "loss": 0.0037, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 4.815384615384616e-06, | |
| "loss": 0.0027, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 4.776923076923077e-06, | |
| "loss": 0.0039, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 4.738461538461539e-06, | |
| "loss": 0.0039, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 4.7e-06, | |
| "loss": 0.0036, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 4.661538461538462e-06, | |
| "loss": 0.0045, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.623076923076923e-06, | |
| "loss": 0.0048, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "eval_loss": 0.9601693153381348, | |
| "eval_runtime": 165.3505, | |
| "eval_samples_per_second": 3.109, | |
| "eval_steps_per_second": 0.393, | |
| "eval_wer": 34.34959349593496, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.5846153846153855e-06, | |
| "loss": 0.0015, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.546153846153846e-06, | |
| "loss": 0.0046, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 4.507692307692308e-06, | |
| "loss": 0.0034, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 4.4692307692307695e-06, | |
| "loss": 0.0021, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 4.430769230769232e-06, | |
| "loss": 0.0021, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 4.392307692307693e-06, | |
| "loss": 0.0027, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 4.353846153846154e-06, | |
| "loss": 0.0029, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 4.315384615384616e-06, | |
| "loss": 0.0029, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.276923076923078e-06, | |
| "loss": 0.0029, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.238461538461539e-06, | |
| "loss": 0.0023, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 4.2000000000000004e-06, | |
| "loss": 0.0015, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.161538461538462e-06, | |
| "loss": 0.0013, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.123076923076924e-06, | |
| "loss": 0.0023, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.084615384615385e-06, | |
| "loss": 0.0013, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 4.0461538461538466e-06, | |
| "loss": 0.0022, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 4.007692307692308e-06, | |
| "loss": 0.0024, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 3.96923076923077e-06, | |
| "loss": 0.003, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 3.930769230769231e-06, | |
| "loss": 0.0019, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 3.892307692307693e-06, | |
| "loss": 0.0017, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 3.853846153846154e-06, | |
| "loss": 0.0031, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 3.815384615384616e-06, | |
| "loss": 0.0029, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 3.7769230769230775e-06, | |
| "loss": 0.0013, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 3.7384615384615384e-06, | |
| "loss": 0.0023, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 3.7e-06, | |
| "loss": 0.0022, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 3.661538461538462e-06, | |
| "loss": 0.0012, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 3.6230769230769236e-06, | |
| "loss": 0.0017, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 3.5846153846153845e-06, | |
| "loss": 0.0015, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 3.5461538461538463e-06, | |
| "loss": 0.0015, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 3.507692307692308e-06, | |
| "loss": 0.0018, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 3.4692307692307698e-06, | |
| "loss": 0.0008, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 3.4307692307692307e-06, | |
| "loss": 0.0024, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 3.3923076923076924e-06, | |
| "loss": 0.0009, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 3.353846153846154e-06, | |
| "loss": 0.0011, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 3.315384615384616e-06, | |
| "loss": 0.0007, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 3.276923076923077e-06, | |
| "loss": 0.0022, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 3.2384615384615385e-06, | |
| "loss": 0.0019, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 3.2000000000000003e-06, | |
| "loss": 0.0019, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 3.161538461538462e-06, | |
| "loss": 0.0012, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 3.123076923076923e-06, | |
| "loss": 0.0008, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 3.0846153846153847e-06, | |
| "loss": 0.0009, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "eval_loss": 1.0040911436080933, | |
| "eval_runtime": 164.566, | |
| "eval_samples_per_second": 3.123, | |
| "eval_steps_per_second": 0.395, | |
| "eval_wer": 33.21718931475029, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 3.0461538461538464e-06, | |
| "loss": 0.0009, | |
| "step": 5025 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 3.007692307692308e-06, | |
| "loss": 0.0006, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 2.969230769230769e-06, | |
| "loss": 0.0013, | |
| "step": 5075 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 2.930769230769231e-06, | |
| "loss": 0.0009, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.8923076923076925e-06, | |
| "loss": 0.0005, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 2.8538461538461543e-06, | |
| "loss": 0.0006, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 2.815384615384615e-06, | |
| "loss": 0.0029, | |
| "step": 5175 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 2.776923076923077e-06, | |
| "loss": 0.0009, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 2.7384615384615387e-06, | |
| "loss": 0.0006, | |
| "step": 5225 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 2.7000000000000004e-06, | |
| "loss": 0.0017, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 2.6615384615384613e-06, | |
| "loss": 0.0016, | |
| "step": 5275 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.623076923076923e-06, | |
| "loss": 0.0005, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.584615384615385e-06, | |
| "loss": 0.0007, | |
| "step": 5325 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.5461538461538465e-06, | |
| "loss": 0.0013, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 2.507692307692308e-06, | |
| "loss": 0.0023, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 2.469230769230769e-06, | |
| "loss": 0.0006, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 2.430769230769231e-06, | |
| "loss": 0.0005, | |
| "step": 5425 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 2.3923076923076922e-06, | |
| "loss": 0.0011, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 2.353846153846154e-06, | |
| "loss": 0.0012, | |
| "step": 5475 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 2.3153846153846153e-06, | |
| "loss": 0.001, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 2.276923076923077e-06, | |
| "loss": 0.0005, | |
| "step": 5525 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 2.2384615384615384e-06, | |
| "loss": 0.0006, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 2.2e-06, | |
| "loss": 0.0008, | |
| "step": 5575 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 2.1615384615384614e-06, | |
| "loss": 0.0005, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 2.123076923076923e-06, | |
| "loss": 0.0007, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 2.0846153846153845e-06, | |
| "loss": 0.0031, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 2.0461538461538462e-06, | |
| "loss": 0.0004, | |
| "step": 5675 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 2.0076923076923076e-06, | |
| "loss": 0.0007, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 1.9692307692307693e-06, | |
| "loss": 0.0004, | |
| "step": 5725 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 1.930769230769231e-06, | |
| "loss": 0.0004, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 1.8923076923076924e-06, | |
| "loss": 0.0005, | |
| "step": 5775 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 1.8538461538461541e-06, | |
| "loss": 0.0004, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 1.8153846153846154e-06, | |
| "loss": 0.0004, | |
| "step": 5825 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 1.7769230769230772e-06, | |
| "loss": 0.0003, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 1.7384615384615385e-06, | |
| "loss": 0.0004, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 1.7000000000000002e-06, | |
| "loss": 0.0006, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 1.6615384615384616e-06, | |
| "loss": 0.0003, | |
| "step": 5925 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 1.6230769230769233e-06, | |
| "loss": 0.0004, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 1.5846153846153846e-06, | |
| "loss": 0.0003, | |
| "step": 5975 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 1.5461538461538464e-06, | |
| "loss": 0.0003, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "eval_loss": 1.036159634590149, | |
| "eval_runtime": 161.2884, | |
| "eval_samples_per_second": 3.187, | |
| "eval_steps_per_second": 0.403, | |
| "eval_wer": 33.10104529616725, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 1.5076923076923077e-06, | |
| "loss": 0.0005, | |
| "step": 6025 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 1.4692307692307694e-06, | |
| "loss": 0.0004, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 1.4307692307692308e-06, | |
| "loss": 0.0007, | |
| "step": 6075 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 1.3923076923076925e-06, | |
| "loss": 0.0005, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 1.3538461538461538e-06, | |
| "loss": 0.0003, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 1.3153846153846156e-06, | |
| "loss": 0.0004, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 1.2769230769230769e-06, | |
| "loss": 0.0003, | |
| "step": 6175 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 1.2384615384615386e-06, | |
| "loss": 0.0004, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 1.2000000000000002e-06, | |
| "loss": 0.0003, | |
| "step": 6225 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 1.1615384615384617e-06, | |
| "loss": 0.001, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 1.1230769230769232e-06, | |
| "loss": 0.0003, | |
| "step": 6275 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 1.0846153846153848e-06, | |
| "loss": 0.0006, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 1.0461538461538463e-06, | |
| "loss": 0.0003, | |
| "step": 6325 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 1.0076923076923078e-06, | |
| "loss": 0.0003, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 9.692307692307693e-07, | |
| "loss": 0.0003, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 9.307692307692308e-07, | |
| "loss": 0.001, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 8.923076923076923e-07, | |
| "loss": 0.0003, | |
| "step": 6425 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 8.538461538461538e-07, | |
| "loss": 0.0005, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 8.153846153846154e-07, | |
| "loss": 0.0003, | |
| "step": 6475 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 7.769230769230769e-07, | |
| "loss": 0.0007, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 7.384615384615385e-07, | |
| "loss": 0.0003, | |
| "step": 6525 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 7.000000000000001e-07, | |
| "loss": 0.0002, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 6.615384615384616e-07, | |
| "loss": 0.0003, | |
| "step": 6575 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 6.230769230769231e-07, | |
| "loss": 0.0003, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 5.846153846153847e-07, | |
| "loss": 0.0004, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 5.461538461538462e-07, | |
| "loss": 0.0003, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 5.076923076923077e-07, | |
| "loss": 0.0003, | |
| "step": 6675 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 4.6923076923076927e-07, | |
| "loss": 0.0003, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 4.307692307692308e-07, | |
| "loss": 0.0003, | |
| "step": 6725 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 3.9230769230769233e-07, | |
| "loss": 0.0002, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 3.5384615384615386e-07, | |
| "loss": 0.0002, | |
| "step": 6775 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 3.153846153846154e-07, | |
| "loss": 0.0002, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 2.76923076923077e-07, | |
| "loss": 0.0002, | |
| "step": 6825 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 2.384615384615385e-07, | |
| "loss": 0.0002, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 2.0000000000000002e-07, | |
| "loss": 0.0011, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 1.6153846153846155e-07, | |
| "loss": 0.0002, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 1.2307692307692308e-07, | |
| "loss": 0.0003, | |
| "step": 6925 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 8.461538461538461e-08, | |
| "loss": 0.0002, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 4.615384615384616e-08, | |
| "loss": 0.0002, | |
| "step": 6975 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 7.692307692307693e-09, | |
| "loss": 0.0006, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "eval_loss": 1.043182611465454, | |
| "eval_runtime": 167.8722, | |
| "eval_samples_per_second": 3.062, | |
| "eval_steps_per_second": 0.387, | |
| "eval_wer": 32.955865272938446, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "step": 7000, | |
| "total_flos": 1.1427509028519936e+20, | |
| "train_loss": 0.19533738821432262, | |
| "train_runtime": 27496.3254, | |
| "train_samples_per_second": 4.073, | |
| "train_steps_per_second": 0.255 | |
| } | |
| ], | |
| "max_steps": 7000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 1.1427509028519936e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |