| { | |
| "best_metric": 5.382267550063277, | |
| "best_model_checkpoint": "./checkpoint-4000", | |
| "epoch": 81.00775, | |
| "global_step": 4000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 2.3949, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 2.4253, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 1.9231, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 1.3843, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 1.1276, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 2.9e-06, | |
| "loss": 0.6593, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.4776, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 0.5082, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 4.4e-06, | |
| "loss": 0.4011, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.3199, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.3085, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.2264, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.2319, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.1764, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.1834, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.1444, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.115, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.1339, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.0896, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.0814, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_loss": 0.19153684377670288, | |
| "eval_runtime": 163.655, | |
| "eval_samples_per_second": 4.619, | |
| "eval_steps_per_second": 0.147, | |
| "eval_wer": 6.6701406982803535, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 9.942857142857145e-06, | |
| "loss": 0.0659, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 9.871428571428572e-06, | |
| "loss": 0.0515, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 9.800000000000001e-06, | |
| "loss": 0.0431, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 9.72857142857143e-06, | |
| "loss": 0.0468, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 9.657142857142859e-06, | |
| "loss": 0.0303, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 9.585714285714286e-06, | |
| "loss": 0.0249, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 9.514285714285715e-06, | |
| "loss": 0.0227, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 9.442857142857144e-06, | |
| "loss": 0.0193, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 9.371428571428572e-06, | |
| "loss": 0.0149, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 9.3e-06, | |
| "loss": 0.0138, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 9.22857142857143e-06, | |
| "loss": 0.0109, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 9.157142857142857e-06, | |
| "loss": 0.0098, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 9.085714285714286e-06, | |
| "loss": 0.0096, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 9.014285714285715e-06, | |
| "loss": 0.008, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 8.942857142857142e-06, | |
| "loss": 0.0075, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 8.871428571428571e-06, | |
| "loss": 0.0059, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 8.8e-06, | |
| "loss": 0.006, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 8.72857142857143e-06, | |
| "loss": 0.0047, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 8.657142857142858e-06, | |
| "loss": 0.0044, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 8.585714285714286e-06, | |
| "loss": 0.0045, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_loss": 0.18161511421203613, | |
| "eval_runtime": 162.7276, | |
| "eval_samples_per_second": 4.646, | |
| "eval_steps_per_second": 0.147, | |
| "eval_wer": 5.508821558847614, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 8.514285714285715e-06, | |
| "loss": 0.0042, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 8.442857142857144e-06, | |
| "loss": 0.0038, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 8.371428571428573e-06, | |
| "loss": 0.0033, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 8.3e-06, | |
| "loss": 0.0029, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 8.22857142857143e-06, | |
| "loss": 0.003, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 8.157142857142858e-06, | |
| "loss": 0.0025, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 8.085714285714287e-06, | |
| "loss": 0.003, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 8.014285714285715e-06, | |
| "loss": 0.0027, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 7.942857142857144e-06, | |
| "loss": 0.0025, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 25.01, | |
| "learning_rate": 7.871428571428573e-06, | |
| "loss": 0.0023, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "learning_rate": 7.800000000000002e-06, | |
| "loss": 0.0022, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 26.01, | |
| "learning_rate": 7.72857142857143e-06, | |
| "loss": 0.002, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "learning_rate": 7.657142857142858e-06, | |
| "loss": 0.0018, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 27.01, | |
| "learning_rate": 7.585714285714286e-06, | |
| "loss": 0.0019, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "learning_rate": 7.514285714285715e-06, | |
| "loss": 0.0018, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 28.01, | |
| "learning_rate": 7.442857142857144e-06, | |
| "loss": 0.0017, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "learning_rate": 7.371428571428571e-06, | |
| "loss": 0.0017, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 29.01, | |
| "learning_rate": 7.3e-06, | |
| "loss": 0.0016, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "learning_rate": 7.2285714285714294e-06, | |
| "loss": 0.0015, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 30.01, | |
| "learning_rate": 7.1571428571428584e-06, | |
| "loss": 0.0016, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 30.01, | |
| "eval_loss": 0.19241470098495483, | |
| "eval_runtime": 158.0572, | |
| "eval_samples_per_second": 4.783, | |
| "eval_steps_per_second": 0.152, | |
| "eval_wer": 5.501377205389712, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "learning_rate": 7.085714285714286e-06, | |
| "loss": 0.0014, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 31.01, | |
| "learning_rate": 7.014285714285715e-06, | |
| "loss": 0.0015, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "learning_rate": 6.942857142857144e-06, | |
| "loss": 0.0013, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 32.01, | |
| "learning_rate": 6.871428571428572e-06, | |
| "loss": 0.0014, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "learning_rate": 6.800000000000001e-06, | |
| "loss": 0.0012, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 33.01, | |
| "learning_rate": 6.728571428571429e-06, | |
| "loss": 0.0014, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "learning_rate": 6.657142857142857e-06, | |
| "loss": 0.0012, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 34.01, | |
| "learning_rate": 6.585714285714286e-06, | |
| "loss": 0.0015, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "learning_rate": 6.514285714285715e-06, | |
| "loss": 0.0013, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 35.01, | |
| "learning_rate": 6.442857142857143e-06, | |
| "loss": 0.0011, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "learning_rate": 6.371428571428572e-06, | |
| "loss": 0.0011, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 36.01, | |
| "learning_rate": 6.300000000000001e-06, | |
| "loss": 0.001, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "learning_rate": 6.22857142857143e-06, | |
| "loss": 0.0011, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 37.01, | |
| "learning_rate": 6.157142857142858e-06, | |
| "loss": 0.001, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "learning_rate": 6.085714285714286e-06, | |
| "loss": 0.0011, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 38.01, | |
| "learning_rate": 6.014285714285715e-06, | |
| "loss": 0.0011, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "learning_rate": 5.942857142857143e-06, | |
| "loss": 0.001, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 39.01, | |
| "learning_rate": 5.871428571428572e-06, | |
| "loss": 0.0009, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "learning_rate": 5.8e-06, | |
| "loss": 0.0009, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "learning_rate": 5.7285714285714285e-06, | |
| "loss": 0.0009, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 40.01, | |
| "eval_loss": 0.19593994319438934, | |
| "eval_runtime": 173.4896, | |
| "eval_samples_per_second": 4.358, | |
| "eval_steps_per_second": 0.138, | |
| "eval_wer": 5.5609320330529295, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "learning_rate": 5.6571428571428576e-06, | |
| "loss": 0.001, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 41.01, | |
| "learning_rate": 5.5857142857142866e-06, | |
| "loss": 0.0008, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "learning_rate": 5.514285714285714e-06, | |
| "loss": 0.0009, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 42.01, | |
| "learning_rate": 5.442857142857143e-06, | |
| "loss": 0.0008, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "learning_rate": 5.371428571428572e-06, | |
| "loss": 0.0008, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 43.01, | |
| "learning_rate": 5.300000000000001e-06, | |
| "loss": 0.0008, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "learning_rate": 5.22857142857143e-06, | |
| "loss": 0.0008, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 44.01, | |
| "learning_rate": 5.157142857142857e-06, | |
| "loss": 0.0008, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 5.085714285714286e-06, | |
| "loss": 0.0007, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 45.01, | |
| "learning_rate": 5.014285714285715e-06, | |
| "loss": 0.0008, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 4.9428571428571435e-06, | |
| "loss": 0.0007, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 46.01, | |
| "learning_rate": 4.871428571428572e-06, | |
| "loss": 0.0007, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 4.800000000000001e-06, | |
| "loss": 0.0007, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 47.01, | |
| "learning_rate": 4.728571428571429e-06, | |
| "loss": 0.0007, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 4.657142857142857e-06, | |
| "loss": 0.0007, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 48.01, | |
| "learning_rate": 4.585714285714286e-06, | |
| "loss": 0.0007, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 4.514285714285714e-06, | |
| "loss": 0.0007, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 49.01, | |
| "learning_rate": 4.442857142857143e-06, | |
| "loss": 0.0007, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 50.01, | |
| "learning_rate": 4.371428571428572e-06, | |
| "loss": 0.0007, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "learning_rate": 4.3e-06, | |
| "loss": 0.0006, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "eval_loss": 0.19894644618034363, | |
| "eval_runtime": 195.7084, | |
| "eval_samples_per_second": 3.863, | |
| "eval_steps_per_second": 0.123, | |
| "eval_wer": 5.419489317352788, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 51.01, | |
| "learning_rate": 4.228571428571429e-06, | |
| "loss": 0.0006, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "learning_rate": 4.1571428571428575e-06, | |
| "loss": 0.0006, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 52.01, | |
| "learning_rate": 4.0857142857142865e-06, | |
| "loss": 0.0006, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "learning_rate": 4.014285714285715e-06, | |
| "loss": 0.0006, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 53.01, | |
| "learning_rate": 3.942857142857143e-06, | |
| "loss": 0.0006, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "learning_rate": 3.871428571428572e-06, | |
| "loss": 0.0006, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 54.01, | |
| "learning_rate": 3.8000000000000005e-06, | |
| "loss": 0.0006, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "learning_rate": 3.7285714285714286e-06, | |
| "loss": 0.0006, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 55.01, | |
| "learning_rate": 3.6571428571428576e-06, | |
| "loss": 0.0006, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "learning_rate": 3.5857142857142862e-06, | |
| "loss": 0.0006, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 56.01, | |
| "learning_rate": 3.5142857142857144e-06, | |
| "loss": 0.0005, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "learning_rate": 3.4428571428571434e-06, | |
| "loss": 0.0006, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 57.01, | |
| "learning_rate": 3.3714285714285716e-06, | |
| "loss": 0.0005, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "learning_rate": 3.3000000000000006e-06, | |
| "loss": 0.0005, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 58.01, | |
| "learning_rate": 3.2285714285714288e-06, | |
| "loss": 0.0005, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "learning_rate": 3.1571428571428573e-06, | |
| "loss": 0.0005, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 59.01, | |
| "learning_rate": 3.085714285714286e-06, | |
| "loss": 0.0005, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "learning_rate": 3.0142857142857145e-06, | |
| "loss": 0.0005, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 60.01, | |
| "learning_rate": 2.9428571428571427e-06, | |
| "loss": 0.0005, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 61.0, | |
| "learning_rate": 2.8714285714285717e-06, | |
| "loss": 0.0005, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 61.0, | |
| "eval_loss": 0.20141343772411346, | |
| "eval_runtime": 152.0323, | |
| "eval_samples_per_second": 4.973, | |
| "eval_steps_per_second": 0.158, | |
| "eval_wer": 5.441822377726494, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 61.01, | |
| "learning_rate": 2.8000000000000003e-06, | |
| "loss": 0.0005, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 62.0, | |
| "learning_rate": 2.728571428571429e-06, | |
| "loss": 0.0005, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 62.01, | |
| "learning_rate": 2.6571428571428575e-06, | |
| "loss": 0.0005, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 63.0, | |
| "learning_rate": 2.5857142857142856e-06, | |
| "loss": 0.0005, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 63.01, | |
| "learning_rate": 2.5142857142857147e-06, | |
| "loss": 0.0005, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 64.0, | |
| "learning_rate": 2.442857142857143e-06, | |
| "loss": 0.0005, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 64.01, | |
| "learning_rate": 2.371428571428572e-06, | |
| "loss": 0.0005, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 65.0, | |
| "learning_rate": 2.3000000000000004e-06, | |
| "loss": 0.0005, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 65.01, | |
| "learning_rate": 2.228571428571429e-06, | |
| "loss": 0.0005, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 66.0, | |
| "learning_rate": 2.157142857142857e-06, | |
| "loss": 0.0005, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 66.01, | |
| "learning_rate": 2.0857142857142858e-06, | |
| "loss": 0.0005, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 67.0, | |
| "learning_rate": 2.0142857142857144e-06, | |
| "loss": 0.0004, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 67.01, | |
| "learning_rate": 1.942857142857143e-06, | |
| "loss": 0.0005, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "learning_rate": 1.8714285714285715e-06, | |
| "loss": 0.0005, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 68.01, | |
| "learning_rate": 1.8000000000000001e-06, | |
| "loss": 0.0004, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 69.0, | |
| "learning_rate": 1.7285714285714287e-06, | |
| "loss": 0.0004, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 69.01, | |
| "learning_rate": 1.657142857142857e-06, | |
| "loss": 0.0005, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 70.0, | |
| "learning_rate": 1.5857142857142857e-06, | |
| "loss": 0.0004, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 70.01, | |
| "learning_rate": 1.5142857142857145e-06, | |
| "loss": 0.0004, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 71.01, | |
| "learning_rate": 1.442857142857143e-06, | |
| "loss": 0.0004, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 71.01, | |
| "eval_loss": 0.20298483967781067, | |
| "eval_runtime": 166.2049, | |
| "eval_samples_per_second": 4.549, | |
| "eval_steps_per_second": 0.144, | |
| "eval_wer": 5.367378843147472, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 71.01, | |
| "learning_rate": 1.3714285714285717e-06, | |
| "loss": 0.0004, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 72.01, | |
| "learning_rate": 1.3e-06, | |
| "loss": 0.0004, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 72.01, | |
| "learning_rate": 1.2285714285714286e-06, | |
| "loss": 0.0004, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 73.01, | |
| "learning_rate": 1.1571428571428572e-06, | |
| "loss": 0.0004, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 73.01, | |
| "learning_rate": 1.0857142857142858e-06, | |
| "loss": 0.0004, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 74.01, | |
| "learning_rate": 1.0142857142857144e-06, | |
| "loss": 0.0004, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 74.01, | |
| "learning_rate": 9.42857142857143e-07, | |
| "loss": 0.0004, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 75.01, | |
| "learning_rate": 8.714285714285716e-07, | |
| "loss": 0.0004, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 76.0, | |
| "learning_rate": 8.000000000000001e-07, | |
| "loss": 0.0004, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 76.01, | |
| "learning_rate": 7.285714285714287e-07, | |
| "loss": 0.0004, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 77.0, | |
| "learning_rate": 6.571428571428571e-07, | |
| "loss": 0.0004, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 77.01, | |
| "learning_rate": 5.857142857142857e-07, | |
| "loss": 0.0004, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "learning_rate": 5.142857142857143e-07, | |
| "loss": 0.0004, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 78.01, | |
| "learning_rate": 4.4285714285714286e-07, | |
| "loss": 0.0004, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "learning_rate": 3.7142857142857145e-07, | |
| "loss": 0.0004, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 79.01, | |
| "learning_rate": 3.0000000000000004e-07, | |
| "loss": 0.0004, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "learning_rate": 2.285714285714286e-07, | |
| "loss": 0.0004, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 80.01, | |
| "learning_rate": 1.5714285714285717e-07, | |
| "loss": 0.0004, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 81.0, | |
| "learning_rate": 8.571428571428573e-08, | |
| "loss": 0.0004, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 81.01, | |
| "learning_rate": 1.4285714285714288e-08, | |
| "loss": 0.0004, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 81.01, | |
| "eval_loss": 0.20338791608810425, | |
| "eval_runtime": 123.1003, | |
| "eval_samples_per_second": 6.141, | |
| "eval_steps_per_second": 0.195, | |
| "eval_wer": 5.382267550063277, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 81.01, | |
| "step": 4000, | |
| "total_flos": 6.25457547804672e+18, | |
| "train_loss": 0.08644696959014982, | |
| "train_runtime": 45590.1016, | |
| "train_samples_per_second": 5.615, | |
| "train_steps_per_second": 0.088 | |
| } | |
| ], | |
| "max_steps": 4000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 6.25457547804672e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |