| { |
| "best_metric": 13.119933160141512, |
| "best_model_checkpoint": "./checkpoint-1000", |
| "epoch": 17.031, |
| "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": 4.2000000000000006e-07, |
| "loss": 1.8289, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 9.200000000000001e-07, |
| "loss": 1.3412, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 1.42e-06, |
| "loss": 0.9356, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 1.9200000000000003e-06, |
| "loss": 0.582, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 2.42e-06, |
| "loss": 0.6274, |
| "step": 125 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 2.92e-06, |
| "loss": 0.6242, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 3.4200000000000007e-06, |
| "loss": 0.3186, |
| "step": 175 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 3.920000000000001e-06, |
| "loss": 0.3184, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 4.42e-06, |
| "loss": 0.3175, |
| "step": 225 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 4.92e-06, |
| "loss": 0.4728, |
| "step": 250 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5.420000000000001e-06, |
| "loss": 0.2793, |
| "step": 275 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 5.92e-06, |
| "loss": 0.2236, |
| "step": 300 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 6.42e-06, |
| "loss": 0.2599, |
| "step": 325 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 6.92e-06, |
| "loss": 0.2334, |
| "step": 350 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 7.420000000000001e-06, |
| "loss": 0.4297, |
| "step": 375 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 7.92e-06, |
| "loss": 0.2255, |
| "step": 400 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 8.42e-06, |
| "loss": 0.2135, |
| "step": 425 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 8.920000000000001e-06, |
| "loss": 0.3233, |
| "step": 450 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 9.42e-06, |
| "loss": 0.1885, |
| "step": 475 |
| }, |
| { |
| "epoch": 2.01, |
| "learning_rate": 9.920000000000002e-06, |
| "loss": 0.167, |
| "step": 500 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 9.940000000000001e-06, |
| "loss": 0.1503, |
| "step": 525 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 9.86857142857143e-06, |
| "loss": 0.1713, |
| "step": 550 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 9.797142857142858e-06, |
| "loss": 0.2312, |
| "step": 575 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 9.725714285714287e-06, |
| "loss": 0.335, |
| "step": 600 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 9.654285714285716e-06, |
| "loss": 0.3067, |
| "step": 625 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 9.582857142857143e-06, |
| "loss": 0.3003, |
| "step": 650 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 9.511428571428572e-06, |
| "loss": 0.1926, |
| "step": 675 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 9.440000000000001e-06, |
| "loss": 0.1874, |
| "step": 700 |
| }, |
| { |
| "epoch": 3.01, |
| "learning_rate": 9.368571428571428e-06, |
| "loss": 0.243, |
| "step": 725 |
| }, |
| { |
| "epoch": 3.02, |
| "learning_rate": 9.297142857142857e-06, |
| "loss": 0.1099, |
| "step": 750 |
| }, |
| { |
| "epoch": 3.02, |
| "learning_rate": 9.225714285714286e-06, |
| "loss": 0.1232, |
| "step": 775 |
| }, |
| { |
| "epoch": 3.03, |
| "learning_rate": 9.154285714285715e-06, |
| "loss": 0.0992, |
| "step": 800 |
| }, |
| { |
| "epoch": 3.04, |
| "learning_rate": 9.082857142857143e-06, |
| "loss": 0.1248, |
| "step": 825 |
| }, |
| { |
| "epoch": 3.04, |
| "learning_rate": 9.011428571428572e-06, |
| "loss": 0.1755, |
| "step": 850 |
| }, |
| { |
| "epoch": 3.05, |
| "learning_rate": 8.94e-06, |
| "loss": 0.2459, |
| "step": 875 |
| }, |
| { |
| "epoch": 3.05, |
| "learning_rate": 8.86857142857143e-06, |
| "loss": 0.1562, |
| "step": 900 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 8.797142857142857e-06, |
| "loss": 0.1336, |
| "step": 925 |
| }, |
| { |
| "epoch": 4.01, |
| "learning_rate": 8.725714285714286e-06, |
| "loss": 0.0611, |
| "step": 950 |
| }, |
| { |
| "epoch": 4.02, |
| "learning_rate": 8.654285714285715e-06, |
| "loss": 0.0909, |
| "step": 975 |
| }, |
| { |
| "epoch": 4.02, |
| "learning_rate": 8.582857142857144e-06, |
| "loss": 0.1789, |
| "step": 1000 |
| }, |
| { |
| "epoch": 4.02, |
| "eval_loss": 0.34212934970855713, |
| "eval_runtime": 458.604, |
| "eval_samples_per_second": 5.58, |
| "eval_steps_per_second": 0.174, |
| "eval_wer": 13.119933160141512, |
| "step": 1000 |
| }, |
| { |
| "epoch": 4.03, |
| "learning_rate": 8.511428571428571e-06, |
| "loss": 0.2191, |
| "step": 1025 |
| }, |
| { |
| "epoch": 4.03, |
| "learning_rate": 8.44e-06, |
| "loss": 0.1451, |
| "step": 1050 |
| }, |
| { |
| "epoch": 4.04, |
| "learning_rate": 8.36857142857143e-06, |
| "loss": 0.0928, |
| "step": 1075 |
| }, |
| { |
| "epoch": 4.05, |
| "learning_rate": 8.297142857142859e-06, |
| "loss": 0.0828, |
| "step": 1100 |
| }, |
| { |
| "epoch": 4.05, |
| "learning_rate": 8.225714285714288e-06, |
| "loss": 0.0929, |
| "step": 1125 |
| }, |
| { |
| "epoch": 5.0, |
| "learning_rate": 8.154285714285715e-06, |
| "loss": 0.0772, |
| "step": 1150 |
| }, |
| { |
| "epoch": 5.01, |
| "learning_rate": 8.082857142857144e-06, |
| "loss": 0.0491, |
| "step": 1175 |
| }, |
| { |
| "epoch": 5.01, |
| "learning_rate": 8.011428571428573e-06, |
| "loss": 0.1263, |
| "step": 1200 |
| }, |
| { |
| "epoch": 5.02, |
| "learning_rate": 7.94e-06, |
| "loss": 0.091, |
| "step": 1225 |
| }, |
| { |
| "epoch": 5.03, |
| "learning_rate": 7.86857142857143e-06, |
| "loss": 0.1252, |
| "step": 1250 |
| }, |
| { |
| "epoch": 5.03, |
| "learning_rate": 7.797142857142858e-06, |
| "loss": 0.091, |
| "step": 1275 |
| }, |
| { |
| "epoch": 5.04, |
| "learning_rate": 7.725714285714286e-06, |
| "loss": 0.0565, |
| "step": 1300 |
| }, |
| { |
| "epoch": 5.05, |
| "learning_rate": 7.654285714285715e-06, |
| "loss": 0.0435, |
| "step": 1325 |
| }, |
| { |
| "epoch": 5.05, |
| "learning_rate": 7.5828571428571444e-06, |
| "loss": 0.0468, |
| "step": 1350 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 7.511428571428572e-06, |
| "loss": 0.0782, |
| "step": 1375 |
| }, |
| { |
| "epoch": 6.01, |
| "learning_rate": 7.440000000000001e-06, |
| "loss": 0.0529, |
| "step": 1400 |
| }, |
| { |
| "epoch": 6.01, |
| "learning_rate": 7.36857142857143e-06, |
| "loss": 0.0318, |
| "step": 1425 |
| }, |
| { |
| "epoch": 6.02, |
| "learning_rate": 7.297142857142858e-06, |
| "loss": 0.0302, |
| "step": 1450 |
| }, |
| { |
| "epoch": 6.03, |
| "learning_rate": 7.225714285714286e-06, |
| "loss": 0.0258, |
| "step": 1475 |
| }, |
| { |
| "epoch": 6.03, |
| "learning_rate": 7.154285714285715e-06, |
| "loss": 0.0261, |
| "step": 1500 |
| }, |
| { |
| "epoch": 6.04, |
| "learning_rate": 7.082857142857143e-06, |
| "loss": 0.0877, |
| "step": 1525 |
| }, |
| { |
| "epoch": 6.05, |
| "learning_rate": 7.011428571428572e-06, |
| "loss": 0.0821, |
| "step": 1550 |
| }, |
| { |
| "epoch": 6.05, |
| "learning_rate": 6.9400000000000005e-06, |
| "loss": 0.028, |
| "step": 1575 |
| }, |
| { |
| "epoch": 7.0, |
| "learning_rate": 6.868571428571429e-06, |
| "loss": 0.0243, |
| "step": 1600 |
| }, |
| { |
| "epoch": 7.01, |
| "learning_rate": 6.797142857142858e-06, |
| "loss": 0.0159, |
| "step": 1625 |
| }, |
| { |
| "epoch": 7.01, |
| "learning_rate": 6.725714285714287e-06, |
| "loss": 0.0553, |
| "step": 1650 |
| }, |
| { |
| "epoch": 7.02, |
| "learning_rate": 6.654285714285716e-06, |
| "loss": 0.0247, |
| "step": 1675 |
| }, |
| { |
| "epoch": 7.03, |
| "learning_rate": 6.582857142857143e-06, |
| "loss": 0.039, |
| "step": 1700 |
| }, |
| { |
| "epoch": 7.03, |
| "learning_rate": 6.511428571428572e-06, |
| "loss": 0.0484, |
| "step": 1725 |
| }, |
| { |
| "epoch": 7.04, |
| "learning_rate": 6.440000000000001e-06, |
| "loss": 0.0206, |
| "step": 1750 |
| }, |
| { |
| "epoch": 7.04, |
| "learning_rate": 6.368571428571429e-06, |
| "loss": 0.0142, |
| "step": 1775 |
| }, |
| { |
| "epoch": 7.05, |
| "learning_rate": 6.297142857142857e-06, |
| "loss": 0.0157, |
| "step": 1800 |
| }, |
| { |
| "epoch": 8.0, |
| "learning_rate": 6.225714285714286e-06, |
| "loss": 0.0152, |
| "step": 1825 |
| }, |
| { |
| "epoch": 8.01, |
| "learning_rate": 6.1542857142857145e-06, |
| "loss": 0.0184, |
| "step": 1850 |
| }, |
| { |
| "epoch": 8.01, |
| "learning_rate": 6.0828571428571435e-06, |
| "loss": 0.012, |
| "step": 1875 |
| }, |
| { |
| "epoch": 8.02, |
| "learning_rate": 6.011428571428572e-06, |
| "loss": 0.0103, |
| "step": 1900 |
| }, |
| { |
| "epoch": 8.03, |
| "learning_rate": 5.94e-06, |
| "loss": 0.0168, |
| "step": 1925 |
| }, |
| { |
| "epoch": 8.03, |
| "learning_rate": 5.868571428571429e-06, |
| "loss": 0.0309, |
| "step": 1950 |
| }, |
| { |
| "epoch": 8.04, |
| "learning_rate": 5.797142857142858e-06, |
| "loss": 0.0122, |
| "step": 1975 |
| }, |
| { |
| "epoch": 8.04, |
| "learning_rate": 5.725714285714287e-06, |
| "loss": 0.0264, |
| "step": 2000 |
| }, |
| { |
| "epoch": 8.04, |
| "eval_loss": 0.45788130164146423, |
| "eval_runtime": 445.9037, |
| "eval_samples_per_second": 5.739, |
| "eval_steps_per_second": 0.179, |
| "eval_wer": 13.515489353924885, |
| "step": 2000 |
| }, |
| { |
| "epoch": 8.05, |
| "learning_rate": 5.654285714285714e-06, |
| "loss": 0.0125, |
| "step": 2025 |
| }, |
| { |
| "epoch": 8.06, |
| "learning_rate": 5.582857142857143e-06, |
| "loss": 0.0088, |
| "step": 2050 |
| }, |
| { |
| "epoch": 9.01, |
| "learning_rate": 5.511428571428572e-06, |
| "loss": 0.0056, |
| "step": 2075 |
| }, |
| { |
| "epoch": 9.01, |
| "learning_rate": 5.4400000000000004e-06, |
| "loss": 0.0052, |
| "step": 2100 |
| }, |
| { |
| "epoch": 9.02, |
| "learning_rate": 5.368571428571429e-06, |
| "loss": 0.0089, |
| "step": 2125 |
| }, |
| { |
| "epoch": 9.02, |
| "learning_rate": 5.297142857142858e-06, |
| "loss": 0.0213, |
| "step": 2150 |
| }, |
| { |
| "epoch": 9.03, |
| "learning_rate": 5.225714285714286e-06, |
| "loss": 0.0087, |
| "step": 2175 |
| }, |
| { |
| "epoch": 9.04, |
| "learning_rate": 5.154285714285715e-06, |
| "loss": 0.0149, |
| "step": 2200 |
| }, |
| { |
| "epoch": 9.04, |
| "learning_rate": 5.082857142857144e-06, |
| "loss": 0.0055, |
| "step": 2225 |
| }, |
| { |
| "epoch": 9.05, |
| "learning_rate": 5.011428571428571e-06, |
| "loss": 0.0046, |
| "step": 2250 |
| }, |
| { |
| "epoch": 9.06, |
| "learning_rate": 4.94e-06, |
| "loss": 0.0056, |
| "step": 2275 |
| }, |
| { |
| "epoch": 10.01, |
| "learning_rate": 4.868571428571429e-06, |
| "loss": 0.0029, |
| "step": 2300 |
| }, |
| { |
| "epoch": 10.01, |
| "learning_rate": 4.797142857142857e-06, |
| "loss": 0.0068, |
| "step": 2325 |
| }, |
| { |
| "epoch": 10.02, |
| "learning_rate": 4.725714285714286e-06, |
| "loss": 0.0045, |
| "step": 2350 |
| }, |
| { |
| "epoch": 10.02, |
| "learning_rate": 4.6542857142857145e-06, |
| "loss": 0.0094, |
| "step": 2375 |
| }, |
| { |
| "epoch": 10.03, |
| "learning_rate": 4.5828571428571435e-06, |
| "loss": 0.0058, |
| "step": 2400 |
| }, |
| { |
| "epoch": 10.04, |
| "learning_rate": 4.511428571428572e-06, |
| "loss": 0.0067, |
| "step": 2425 |
| }, |
| { |
| "epoch": 10.04, |
| "learning_rate": 4.440000000000001e-06, |
| "loss": 0.0041, |
| "step": 2450 |
| }, |
| { |
| "epoch": 10.05, |
| "learning_rate": 4.368571428571429e-06, |
| "loss": 0.0034, |
| "step": 2475 |
| }, |
| { |
| "epoch": 10.05, |
| "learning_rate": 4.297142857142858e-06, |
| "loss": 0.0031, |
| "step": 2500 |
| }, |
| { |
| "epoch": 11.0, |
| "learning_rate": 4.225714285714286e-06, |
| "loss": 0.0059, |
| "step": 2525 |
| }, |
| { |
| "epoch": 11.01, |
| "learning_rate": 4.154285714285714e-06, |
| "loss": 0.003, |
| "step": 2550 |
| }, |
| { |
| "epoch": 11.02, |
| "learning_rate": 4.082857142857143e-06, |
| "loss": 0.0022, |
| "step": 2575 |
| }, |
| { |
| "epoch": 11.02, |
| "learning_rate": 4.011428571428571e-06, |
| "loss": 0.0038, |
| "step": 2600 |
| }, |
| { |
| "epoch": 11.03, |
| "learning_rate": 3.94e-06, |
| "loss": 0.0032, |
| "step": 2625 |
| }, |
| { |
| "epoch": 11.04, |
| "learning_rate": 3.8685714285714286e-06, |
| "loss": 0.002, |
| "step": 2650 |
| }, |
| { |
| "epoch": 11.04, |
| "learning_rate": 3.7971428571428576e-06, |
| "loss": 0.0023, |
| "step": 2675 |
| }, |
| { |
| "epoch": 11.05, |
| "learning_rate": 3.7257142857142857e-06, |
| "loss": 0.0022, |
| "step": 2700 |
| }, |
| { |
| "epoch": 11.05, |
| "learning_rate": 3.6542857142857148e-06, |
| "loss": 0.0071, |
| "step": 2725 |
| }, |
| { |
| "epoch": 12.0, |
| "learning_rate": 3.582857142857143e-06, |
| "loss": 0.0033, |
| "step": 2750 |
| }, |
| { |
| "epoch": 12.01, |
| "learning_rate": 3.511428571428572e-06, |
| "loss": 0.003, |
| "step": 2775 |
| }, |
| { |
| "epoch": 12.02, |
| "learning_rate": 3.44e-06, |
| "loss": 0.0041, |
| "step": 2800 |
| }, |
| { |
| "epoch": 12.02, |
| "learning_rate": 3.3685714285714287e-06, |
| "loss": 0.0022, |
| "step": 2825 |
| }, |
| { |
| "epoch": 12.03, |
| "learning_rate": 3.2971428571428577e-06, |
| "loss": 0.0017, |
| "step": 2850 |
| }, |
| { |
| "epoch": 12.03, |
| "learning_rate": 3.225714285714286e-06, |
| "loss": 0.0015, |
| "step": 2875 |
| }, |
| { |
| "epoch": 12.04, |
| "learning_rate": 3.154285714285715e-06, |
| "loss": 0.0021, |
| "step": 2900 |
| }, |
| { |
| "epoch": 12.05, |
| "learning_rate": 3.082857142857143e-06, |
| "loss": 0.0045, |
| "step": 2925 |
| }, |
| { |
| "epoch": 12.05, |
| "learning_rate": 3.0114285714285716e-06, |
| "loss": 0.0024, |
| "step": 2950 |
| }, |
| { |
| "epoch": 13.0, |
| "learning_rate": 2.9400000000000002e-06, |
| "loss": 0.0016, |
| "step": 2975 |
| }, |
| { |
| "epoch": 13.01, |
| "learning_rate": 2.868571428571429e-06, |
| "loss": 0.0023, |
| "step": 3000 |
| }, |
| { |
| "epoch": 13.01, |
| "eval_loss": 0.5478612184524536, |
| "eval_runtime": 437.7935, |
| "eval_samples_per_second": 5.845, |
| "eval_steps_per_second": 0.183, |
| "eval_wer": 13.65386874844976, |
| "step": 3000 |
| }, |
| { |
| "epoch": 13.02, |
| "learning_rate": 2.797142857142857e-06, |
| "loss": 0.0017, |
| "step": 3025 |
| }, |
| { |
| "epoch": 13.02, |
| "learning_rate": 2.725714285714286e-06, |
| "loss": 0.0013, |
| "step": 3050 |
| }, |
| { |
| "epoch": 13.03, |
| "learning_rate": 2.654285714285714e-06, |
| "loss": 0.0015, |
| "step": 3075 |
| }, |
| { |
| "epoch": 13.03, |
| "learning_rate": 2.582857142857143e-06, |
| "loss": 0.0025, |
| "step": 3100 |
| }, |
| { |
| "epoch": 13.04, |
| "learning_rate": 2.5114285714285718e-06, |
| "loss": 0.002, |
| "step": 3125 |
| }, |
| { |
| "epoch": 13.05, |
| "learning_rate": 2.4400000000000004e-06, |
| "loss": 0.0014, |
| "step": 3150 |
| }, |
| { |
| "epoch": 13.05, |
| "learning_rate": 2.3685714285714285e-06, |
| "loss": 0.0015, |
| "step": 3175 |
| }, |
| { |
| "epoch": 14.0, |
| "learning_rate": 2.297142857142857e-06, |
| "loss": 0.0033, |
| "step": 3200 |
| }, |
| { |
| "epoch": 14.01, |
| "learning_rate": 2.2257142857142857e-06, |
| "loss": 0.0015, |
| "step": 3225 |
| }, |
| { |
| "epoch": 14.01, |
| "learning_rate": 2.1542857142857147e-06, |
| "loss": 0.0013, |
| "step": 3250 |
| }, |
| { |
| "epoch": 14.02, |
| "learning_rate": 2.0828571428571433e-06, |
| "loss": 0.0012, |
| "step": 3275 |
| }, |
| { |
| "epoch": 14.03, |
| "learning_rate": 2.0114285714285715e-06, |
| "loss": 0.0011, |
| "step": 3300 |
| }, |
| { |
| "epoch": 14.03, |
| "learning_rate": 1.94e-06, |
| "loss": 0.0012, |
| "step": 3325 |
| }, |
| { |
| "epoch": 14.04, |
| "learning_rate": 1.8685714285714289e-06, |
| "loss": 0.0023, |
| "step": 3350 |
| }, |
| { |
| "epoch": 14.05, |
| "learning_rate": 1.7971428571428572e-06, |
| "loss": 0.0028, |
| "step": 3375 |
| }, |
| { |
| "epoch": 14.05, |
| "learning_rate": 1.7257142857142858e-06, |
| "loss": 0.0021, |
| "step": 3400 |
| }, |
| { |
| "epoch": 15.0, |
| "learning_rate": 1.6542857142857144e-06, |
| "loss": 0.002, |
| "step": 3425 |
| }, |
| { |
| "epoch": 15.01, |
| "learning_rate": 1.582857142857143e-06, |
| "loss": 0.0011, |
| "step": 3450 |
| }, |
| { |
| "epoch": 15.01, |
| "learning_rate": 1.5114285714285714e-06, |
| "loss": 0.0012, |
| "step": 3475 |
| }, |
| { |
| "epoch": 15.02, |
| "learning_rate": 1.44e-06, |
| "loss": 0.0012, |
| "step": 3500 |
| }, |
| { |
| "epoch": 15.03, |
| "learning_rate": 1.3685714285714286e-06, |
| "loss": 0.0028, |
| "step": 3525 |
| }, |
| { |
| "epoch": 15.03, |
| "learning_rate": 1.2971428571428574e-06, |
| "loss": 0.0019, |
| "step": 3550 |
| }, |
| { |
| "epoch": 15.04, |
| "learning_rate": 1.2257142857142857e-06, |
| "loss": 0.0019, |
| "step": 3575 |
| }, |
| { |
| "epoch": 15.04, |
| "learning_rate": 1.1542857142857143e-06, |
| "loss": 0.0012, |
| "step": 3600 |
| }, |
| { |
| "epoch": 15.05, |
| "learning_rate": 1.082857142857143e-06, |
| "loss": 0.0011, |
| "step": 3625 |
| }, |
| { |
| "epoch": 16.0, |
| "learning_rate": 1.0114285714285715e-06, |
| "loss": 0.0011, |
| "step": 3650 |
| }, |
| { |
| "epoch": 16.01, |
| "learning_rate": 9.400000000000001e-07, |
| "loss": 0.0012, |
| "step": 3675 |
| }, |
| { |
| "epoch": 16.01, |
| "learning_rate": 8.685714285714286e-07, |
| "loss": 0.0011, |
| "step": 3700 |
| }, |
| { |
| "epoch": 16.02, |
| "learning_rate": 7.971428571428572e-07, |
| "loss": 0.0014, |
| "step": 3725 |
| }, |
| { |
| "epoch": 16.03, |
| "learning_rate": 7.257142857142857e-07, |
| "loss": 0.0019, |
| "step": 3750 |
| }, |
| { |
| "epoch": 16.03, |
| "learning_rate": 6.542857142857144e-07, |
| "loss": 0.0011, |
| "step": 3775 |
| }, |
| { |
| "epoch": 16.04, |
| "learning_rate": 5.82857142857143e-07, |
| "loss": 0.0011, |
| "step": 3800 |
| }, |
| { |
| "epoch": 16.04, |
| "learning_rate": 5.114285714285714e-07, |
| "loss": 0.001, |
| "step": 3825 |
| }, |
| { |
| "epoch": 16.05, |
| "learning_rate": 4.4e-07, |
| "loss": 0.0011, |
| "step": 3850 |
| }, |
| { |
| "epoch": 16.06, |
| "learning_rate": 3.685714285714286e-07, |
| "loss": 0.0025, |
| "step": 3875 |
| }, |
| { |
| "epoch": 17.01, |
| "learning_rate": 2.9714285714285715e-07, |
| "loss": 0.001, |
| "step": 3900 |
| }, |
| { |
| "epoch": 17.01, |
| "learning_rate": 2.2571428571428574e-07, |
| "loss": 0.0009, |
| "step": 3925 |
| }, |
| { |
| "epoch": 17.02, |
| "learning_rate": 1.542857142857143e-07, |
| "loss": 0.0011, |
| "step": 3950 |
| }, |
| { |
| "epoch": 17.02, |
| "learning_rate": 8.285714285714285e-08, |
| "loss": 0.001, |
| "step": 3975 |
| }, |
| { |
| "epoch": 17.03, |
| "learning_rate": 1.142857142857143e-08, |
| "loss": 0.0011, |
| "step": 4000 |
| }, |
| { |
| "epoch": 17.03, |
| "eval_loss": 0.5635844469070435, |
| "eval_runtime": 433.7476, |
| "eval_samples_per_second": 5.9, |
| "eval_steps_per_second": 0.184, |
| "eval_wer": 13.464576180467619, |
| "step": 4000 |
| }, |
| { |
| "epoch": 17.03, |
| "step": 4000, |
| "total_flos": 7.371106012053504e+19, |
| "train_loss": 0.09987659510597587, |
| "train_runtime": 37988.6085, |
| "train_samples_per_second": 6.739, |
| "train_steps_per_second": 0.105 |
| } |
| ], |
| "max_steps": 4000, |
| "num_train_epochs": 9223372036854775807, |
| "total_flos": 7.371106012053504e+19, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
