| { | |
| "best_metric": 27.02397743300423, | |
| "best_model_checkpoint": "./checkpoint-1000", | |
| "epoch": 249.004, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 4.7408, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 2.5496, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 1.7526, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 1.4285, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 1.1887, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 2.9e-06, | |
| "loss": 1.0659, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.9853, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 0.8344, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 4.4e-06, | |
| "loss": 0.5356, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.3008, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.1907, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.1309, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.0875, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.0629, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.0415, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.032, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.0206, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.019, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.0165, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.0149, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "learning_rate": 9.955555555555556e-06, | |
| "loss": 0.0126, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "learning_rate": 9.9e-06, | |
| "loss": 0.0096, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "learning_rate": 9.844444444444446e-06, | |
| "loss": 0.0098, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "learning_rate": 9.78888888888889e-06, | |
| "loss": 0.0072, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "learning_rate": 9.733333333333334e-06, | |
| "loss": 0.0071, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "learning_rate": 9.677777777777778e-06, | |
| "loss": 0.0053, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "learning_rate": 9.622222222222222e-06, | |
| "loss": 0.0041, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "learning_rate": 9.566666666666668e-06, | |
| "loss": 0.0062, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "learning_rate": 9.511111111111112e-06, | |
| "loss": 0.0045, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "learning_rate": 9.455555555555557e-06, | |
| "loss": 0.0027, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "learning_rate": 9.4e-06, | |
| "loss": 0.0037, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "learning_rate": 9.344444444444446e-06, | |
| "loss": 0.0034, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "learning_rate": 9.28888888888889e-06, | |
| "loss": 0.0026, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "learning_rate": 9.233333333333334e-06, | |
| "loss": 0.0039, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "learning_rate": 9.17777777777778e-06, | |
| "loss": 0.0038, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "learning_rate": 9.122222222222223e-06, | |
| "loss": 0.0016, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "learning_rate": 9.066666666666667e-06, | |
| "loss": 0.0012, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "learning_rate": 9.011111111111111e-06, | |
| "loss": 0.0015, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "learning_rate": 8.955555555555555e-06, | |
| "loss": 0.0016, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "learning_rate": 8.900000000000001e-06, | |
| "loss": 0.0021, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "eval_loss": 0.42445117235183716, | |
| "eval_runtime": 436.164, | |
| "eval_samples_per_second": 0.504, | |
| "eval_steps_per_second": 0.016, | |
| "eval_wer": 27.02397743300423, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "learning_rate": 8.844444444444445e-06, | |
| "loss": 0.0011, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "learning_rate": 8.788888888888891e-06, | |
| "loss": 0.0005, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "learning_rate": 8.733333333333333e-06, | |
| "loss": 0.0003, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "learning_rate": 8.677777777777779e-06, | |
| "loss": 0.0006, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "learning_rate": 8.622222222222223e-06, | |
| "loss": 0.0007, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "learning_rate": 8.566666666666667e-06, | |
| "loss": 0.0016, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "learning_rate": 8.511111111111113e-06, | |
| "loss": 0.0018, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "learning_rate": 8.455555555555555e-06, | |
| "loss": 0.0017, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 61.0, | |
| "learning_rate": 8.400000000000001e-06, | |
| "loss": 0.0016, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 62.0, | |
| "learning_rate": 8.344444444444445e-06, | |
| "loss": 0.0011, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 63.0, | |
| "learning_rate": 8.288888888888889e-06, | |
| "loss": 0.0007, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 64.0, | |
| "learning_rate": 8.233333333333335e-06, | |
| "loss": 0.0004, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 66.0, | |
| "learning_rate": 8.177777777777779e-06, | |
| "loss": 0.0002, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 67.0, | |
| "learning_rate": 8.122222222222223e-06, | |
| "loss": 0.0002, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "learning_rate": 8.066666666666667e-06, | |
| "loss": 0.0002, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 69.0, | |
| "learning_rate": 8.011111111111113e-06, | |
| "loss": 0.0002, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 71.0, | |
| "learning_rate": 7.955555555555557e-06, | |
| "loss": 0.0002, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 72.0, | |
| "learning_rate": 7.9e-06, | |
| "loss": 0.0002, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 73.0, | |
| "learning_rate": 7.844444444444446e-06, | |
| "loss": 0.0002, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "learning_rate": 7.788888888888889e-06, | |
| "loss": 0.0002, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 76.0, | |
| "learning_rate": 7.733333333333334e-06, | |
| "loss": 0.0002, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 77.0, | |
| "learning_rate": 7.677777777777778e-06, | |
| "loss": 0.0002, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "learning_rate": 7.622222222222223e-06, | |
| "loss": 0.0002, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "learning_rate": 7.566666666666667e-06, | |
| "loss": 0.0002, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 81.0, | |
| "learning_rate": 7.511111111111111e-06, | |
| "loss": 0.0002, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 82.0, | |
| "learning_rate": 7.455555555555556e-06, | |
| "loss": 0.0002, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 83.0, | |
| "learning_rate": 7.4e-06, | |
| "loss": 0.0002, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 84.0, | |
| "learning_rate": 7.344444444444445e-06, | |
| "loss": 0.0002, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 86.0, | |
| "learning_rate": 7.28888888888889e-06, | |
| "loss": 0.0002, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 87.0, | |
| "learning_rate": 7.233333333333334e-06, | |
| "loss": 0.0002, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 88.0, | |
| "learning_rate": 7.177777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 89.0, | |
| "learning_rate": 7.122222222222222e-06, | |
| "loss": 0.0001, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 91.0, | |
| "learning_rate": 7.066666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 92.0, | |
| "learning_rate": 7.011111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 93.0, | |
| "learning_rate": 6.955555555555557e-06, | |
| "loss": 0.0001, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 94.0, | |
| "learning_rate": 6.9e-06, | |
| "loss": 0.0001, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 96.0, | |
| "learning_rate": 6.844444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 97.0, | |
| "learning_rate": 6.788888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 98.0, | |
| "learning_rate": 6.733333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 99.0, | |
| "learning_rate": 6.677777777777779e-06, | |
| "loss": 0.0001, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 99.0, | |
| "eval_loss": 0.7338100075721741, | |
| "eval_runtime": 418.9102, | |
| "eval_samples_per_second": 0.525, | |
| "eval_steps_per_second": 0.017, | |
| "eval_wer": 28.124118476727784, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 101.0, | |
| "learning_rate": 6.6222222222222236e-06, | |
| "loss": 0.0001, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 102.0, | |
| "learning_rate": 6.566666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 103.0, | |
| "learning_rate": 6.511111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 104.0, | |
| "learning_rate": 6.455555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 106.0, | |
| "learning_rate": 6.4000000000000006e-06, | |
| "loss": 0.0001, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 107.0, | |
| "learning_rate": 6.3444444444444454e-06, | |
| "loss": 0.0001, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 108.0, | |
| "learning_rate": 6.28888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 109.0, | |
| "learning_rate": 6.2333333333333335e-06, | |
| "loss": 0.0001, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 111.0, | |
| "learning_rate": 6.177777777777778e-06, | |
| "loss": 0.0001, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 112.0, | |
| "learning_rate": 6.1222222222222224e-06, | |
| "loss": 0.0001, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 113.0, | |
| "learning_rate": 6.066666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 114.0, | |
| "learning_rate": 6.011111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 116.0, | |
| "learning_rate": 5.955555555555555e-06, | |
| "loss": 0.0001, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 117.0, | |
| "learning_rate": 5.9e-06, | |
| "loss": 0.0001, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 118.0, | |
| "learning_rate": 5.844444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 119.0, | |
| "learning_rate": 5.788888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 121.0, | |
| "learning_rate": 5.733333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 122.0, | |
| "learning_rate": 5.677777777777779e-06, | |
| "loss": 0.0001, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 123.0, | |
| "learning_rate": 5.622222222222222e-06, | |
| "loss": 0.0001, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 124.0, | |
| "learning_rate": 5.566666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 126.0, | |
| "learning_rate": 5.511111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 127.0, | |
| "learning_rate": 5.455555555555556e-06, | |
| "loss": 0.0001, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 128.0, | |
| "learning_rate": 5.400000000000001e-06, | |
| "loss": 0.0001, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 129.0, | |
| "learning_rate": 5.344444444444446e-06, | |
| "loss": 0.0001, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 131.0, | |
| "learning_rate": 5.288888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 132.0, | |
| "learning_rate": 5.233333333333334e-06, | |
| "loss": 0.0001, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 133.0, | |
| "learning_rate": 5.177777777777779e-06, | |
| "loss": 0.0001, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 134.0, | |
| "learning_rate": 5.122222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 136.0, | |
| "learning_rate": 5.0666666666666676e-06, | |
| "loss": 0.0001, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 137.0, | |
| "learning_rate": 5.011111111111111e-06, | |
| "loss": 0.0001, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 138.0, | |
| "learning_rate": 4.9555555555555565e-06, | |
| "loss": 0.0001, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 139.0, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 0.0001, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 141.0, | |
| "learning_rate": 4.8444444444444446e-06, | |
| "loss": 0.0001, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 142.0, | |
| "learning_rate": 4.7888888888888894e-06, | |
| "loss": 0.0001, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 143.0, | |
| "learning_rate": 4.7333333333333335e-06, | |
| "loss": 0.0, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 144.0, | |
| "learning_rate": 4.677777777777778e-06, | |
| "loss": 0.0, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 146.0, | |
| "learning_rate": 4.622222222222222e-06, | |
| "loss": 0.0, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 147.0, | |
| "learning_rate": 4.566666666666667e-06, | |
| "loss": 0.0, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 148.0, | |
| "learning_rate": 4.511111111111111e-06, | |
| "loss": 0.0, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 149.0, | |
| "learning_rate": 4.455555555555555e-06, | |
| "loss": 0.0, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 149.0, | |
| "eval_loss": 0.8594304323196411, | |
| "eval_runtime": 412.7645, | |
| "eval_samples_per_second": 0.533, | |
| "eval_steps_per_second": 0.017, | |
| "eval_wer": 28.660084626234134, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 151.0, | |
| "learning_rate": 4.4e-06, | |
| "loss": 0.0, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 152.0, | |
| "learning_rate": 4.344444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 153.0, | |
| "learning_rate": 4.288888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 154.0, | |
| "learning_rate": 4.233333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 156.0, | |
| "learning_rate": 4.177777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 157.0, | |
| "learning_rate": 4.122222222222222e-06, | |
| "loss": 0.0, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 158.0, | |
| "learning_rate": 4.066666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 159.0, | |
| "learning_rate": 4.011111111111111e-06, | |
| "loss": 0.0, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 161.0, | |
| "learning_rate": 3.955555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 162.0, | |
| "learning_rate": 3.900000000000001e-06, | |
| "loss": 0.0, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 163.0, | |
| "learning_rate": 3.844444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 164.0, | |
| "learning_rate": 3.7888888888888893e-06, | |
| "loss": 0.0, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 166.0, | |
| "learning_rate": 3.7333333333333337e-06, | |
| "loss": 0.0, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 167.0, | |
| "learning_rate": 3.6777777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 168.0, | |
| "learning_rate": 3.6222222222222226e-06, | |
| "loss": 0.0, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 169.0, | |
| "learning_rate": 3.566666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 171.0, | |
| "learning_rate": 3.511111111111111e-06, | |
| "loss": 0.0, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 172.0, | |
| "learning_rate": 3.455555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 173.0, | |
| "learning_rate": 3.4000000000000005e-06, | |
| "loss": 0.0, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 174.0, | |
| "learning_rate": 3.3444444444444445e-06, | |
| "loss": 0.0, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 176.0, | |
| "learning_rate": 3.2888888888888894e-06, | |
| "loss": 0.0, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 177.0, | |
| "learning_rate": 3.2333333333333334e-06, | |
| "loss": 0.0, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 178.0, | |
| "learning_rate": 3.177777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 179.0, | |
| "learning_rate": 3.1222222222222228e-06, | |
| "loss": 0.0, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 181.0, | |
| "learning_rate": 3.066666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 182.0, | |
| "learning_rate": 3.0111111111111113e-06, | |
| "loss": 0.0, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 183.0, | |
| "learning_rate": 2.955555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 184.0, | |
| "learning_rate": 2.9e-06, | |
| "loss": 0.0, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 186.0, | |
| "learning_rate": 2.8444444444444446e-06, | |
| "loss": 0.0, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 187.0, | |
| "learning_rate": 2.788888888888889e-06, | |
| "loss": 0.0, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 188.0, | |
| "learning_rate": 2.7333333333333336e-06, | |
| "loss": 0.0, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 189.0, | |
| "learning_rate": 2.677777777777778e-06, | |
| "loss": 0.0, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 191.0, | |
| "learning_rate": 2.6222222222222225e-06, | |
| "loss": 0.0, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 192.0, | |
| "learning_rate": 2.566666666666667e-06, | |
| "loss": 0.0, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 193.0, | |
| "learning_rate": 2.5111111111111114e-06, | |
| "loss": 0.0, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 194.0, | |
| "learning_rate": 2.455555555555556e-06, | |
| "loss": 0.0, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 196.0, | |
| "learning_rate": 2.4000000000000003e-06, | |
| "loss": 0.0, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 197.0, | |
| "learning_rate": 2.3444444444444448e-06, | |
| "loss": 0.0, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 198.0, | |
| "learning_rate": 2.2888888888888892e-06, | |
| "loss": 0.0, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "learning_rate": 2.2333333333333333e-06, | |
| "loss": 0.0, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 199.0, | |
| "eval_loss": 0.9102850556373596, | |
| "eval_runtime": 430.0535, | |
| "eval_samples_per_second": 0.512, | |
| "eval_steps_per_second": 0.016, | |
| "eval_wer": 28.34978843441467, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 201.0, | |
| "learning_rate": 2.1777777777777777e-06, | |
| "loss": 0.0, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 202.0, | |
| "learning_rate": 2.1222222222222226e-06, | |
| "loss": 0.0, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 203.0, | |
| "learning_rate": 2.0666666666666666e-06, | |
| "loss": 0.0, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 204.0, | |
| "learning_rate": 2.011111111111111e-06, | |
| "loss": 0.0, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 206.0, | |
| "learning_rate": 1.955555555555556e-06, | |
| "loss": 0.0, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 207.0, | |
| "learning_rate": 1.9000000000000002e-06, | |
| "loss": 0.0, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 208.0, | |
| "learning_rate": 1.8444444444444445e-06, | |
| "loss": 0.0, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 209.0, | |
| "learning_rate": 1.788888888888889e-06, | |
| "loss": 0.0, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 211.0, | |
| "learning_rate": 1.7333333333333336e-06, | |
| "loss": 0.0, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 212.0, | |
| "learning_rate": 1.6777777777777779e-06, | |
| "loss": 0.0, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 213.0, | |
| "learning_rate": 1.6222222222222223e-06, | |
| "loss": 0.0, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 214.0, | |
| "learning_rate": 1.566666666666667e-06, | |
| "loss": 0.0, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 216.0, | |
| "learning_rate": 1.5111111111111112e-06, | |
| "loss": 0.0, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 217.0, | |
| "learning_rate": 1.4555555555555557e-06, | |
| "loss": 0.0, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 218.0, | |
| "learning_rate": 1.4000000000000001e-06, | |
| "loss": 0.0, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 219.0, | |
| "learning_rate": 1.3444444444444446e-06, | |
| "loss": 0.0, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 221.0, | |
| "learning_rate": 1.288888888888889e-06, | |
| "loss": 0.0, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 222.0, | |
| "learning_rate": 1.2333333333333335e-06, | |
| "loss": 0.0, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 223.0, | |
| "learning_rate": 1.1777777777777778e-06, | |
| "loss": 0.0, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 224.0, | |
| "learning_rate": 1.1222222222222222e-06, | |
| "loss": 0.0, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 226.0, | |
| "learning_rate": 1.066666666666667e-06, | |
| "loss": 0.0, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 227.0, | |
| "learning_rate": 1.0111111111111111e-06, | |
| "loss": 0.0, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 228.0, | |
| "learning_rate": 9.555555555555556e-07, | |
| "loss": 0.0, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 229.0, | |
| "learning_rate": 9.000000000000001e-07, | |
| "loss": 0.0, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 231.0, | |
| "learning_rate": 8.444444444444445e-07, | |
| "loss": 0.0, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 232.0, | |
| "learning_rate": 7.888888888888889e-07, | |
| "loss": 0.0, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 233.0, | |
| "learning_rate": 7.333333333333334e-07, | |
| "loss": 0.0, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 234.0, | |
| "learning_rate": 6.777777777777779e-07, | |
| "loss": 0.0, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 236.0, | |
| "learning_rate": 6.222222222222223e-07, | |
| "loss": 0.0, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 237.0, | |
| "learning_rate": 5.666666666666667e-07, | |
| "loss": 0.0, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 238.0, | |
| "learning_rate": 5.111111111111112e-07, | |
| "loss": 0.0, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 239.0, | |
| "learning_rate": 4.5555555555555563e-07, | |
| "loss": 0.0, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 241.0, | |
| "learning_rate": 4.0000000000000003e-07, | |
| "loss": 0.0, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 242.0, | |
| "learning_rate": 3.444444444444445e-07, | |
| "loss": 0.0, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 243.0, | |
| "learning_rate": 2.888888888888889e-07, | |
| "loss": 0.0, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 244.0, | |
| "learning_rate": 2.3333333333333336e-07, | |
| "loss": 0.0, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 246.0, | |
| "learning_rate": 1.777777777777778e-07, | |
| "loss": 0.0, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 247.0, | |
| "learning_rate": 1.2222222222222225e-07, | |
| "loss": 0.0, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 248.0, | |
| "learning_rate": 6.666666666666668e-08, | |
| "loss": 0.0, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 249.0, | |
| "learning_rate": 1.1111111111111112e-08, | |
| "loss": 0.0, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 249.0, | |
| "eval_loss": 0.9329229593276978, | |
| "eval_runtime": 412.6881, | |
| "eval_samples_per_second": 0.533, | |
| "eval_steps_per_second": 0.017, | |
| "eval_wer": 28.293370944992947, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 249.0, | |
| "step": 5000, | |
| "total_flos": 4.458644434944e+19, | |
| "train_loss": 0.08057039538093959, | |
| "train_runtime": 28298.6169, | |
| "train_samples_per_second": 5.654, | |
| "train_steps_per_second": 0.177 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 4.458644434944e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |